app.py

 #!/usr/bin/env python -W ignore::DeprecationWarning

import os

import ast
import pathlib
import pandas as pd
import numpy as np
import random
import itertools
from tqdm import tqdm
from skimage import measure
from scipy import stats

def warn(*args, **kwargs):
    pass
import warnings
warnings.warn = warn

import logging
log = logging.getLogger('werkzeug')
log.setLevel(logging.ERROR)
logging.captureWarnings(True)

import inputfuncs
import ccprocess
import libcc

import segmfuncs
import parcelfuncs

import webbrowser
from threading import Timer

import dash
import dash_table
import dash_core_components as dcc
import dash_html_components as html
import dash_bootstrap_components as dbc
import dash_daq as daq
from dash.dependencies import Input, Output, State, ALL, MATCH
from dash.exceptions import PreventUpdate

from dash_extensions import Download
from dash_extensions.snippets import send_data_frame

import plotly.io as pio
import plotly.figure_factory as ff
import plotly.express as px
import plotly.graph_objs as go
from plotly.subplots import make_subplots

class Error(Exception):
    pass

theme = 'plotly'

print(' ')

# GENERAL DEFINITIONS -------------------------------------------------------------------------

dict_segmentation_functions = {'ROQS': segmfuncs.segm_roqs,
                             'Watershed': segmfuncs.segm_watershed,
                             'Imported Masks': segmfuncs.segm_mask}

dict_parcellation_functions = {'Witelson': parcelfuncs.parc_witelson,
                             'Hofer': parcelfuncs.parc_hofer,
                             'Chao': parcelfuncs.parc_chao,
                             'Cover': parcelfuncs.parc_cover,
                             'Freesurfer': parcelfuncs.parc_freesurfer}

dict_3d_segmentation_functions = {'Watershed3d': segmfuncs.segm_watershed_3d}

scalar_list = ['FA', 'MD', 'RD', 'AD']
colors_list = px.colors.qualitative.Plotly

dict_parcellation_methods = {'Witelson': 'witelson', 'Hofer & Frahm': 'hofer', 'Chao et al':'chao', 'Cover et al': 'cover', 'Freesurfer':'freesurfer'}
dict_segmentation_methods = {'ROQS': 'roqs', 'Watershed': 'watershed'}
dict_3d_segmentation_methods = {'Watershed3d':'watershed3d'}

# DATA IMPORTING -----------------------------------------------------------------------------

# Arg parser
opts = inputfuncs.get_parser().parse_args()

if opts.staple is True:
    dict_segmentation_functions['STAPLE'] = segmfuncs.segm_staple
    dict_segmentation_methods['STAPLE'] = 'staple'

df_categories = pd.DataFrame()
df_numerical = pd.DataFrame()

# Read external data
if opts.ext_data is not None:
    
    external_data_path = opts.ext_data
    external_data = pd.read_excel(external_data_path, dtype={'Subjects':'object'})
    external_data = external_data.set_index('Subjects')

    # Clear NaNs on index
    external_data = external_data[external_data.index.notnull()]
    
    # Remove unnamed columns 
    external_data = external_data.loc[:,~external_data.columns.str.match("Unnamed")]

    df_categories = external_data.select_dtypes(include=['object'])
    df_numerical = external_data.select_dtypes(include=['number'])
    
    col_categories = ['Method'] + list(df_categories.columns)    

# Get indicated directories
path_dict = {}
if opts.folders is not None:
    for directory in opts.folders:
        if directory is not None:
            if inputfuncs.check_directory(directory, opts.basename):
                path_dict[os.path.basename(directory)] = os.path.join(directory, '')

# Import the subjects inside the parents folders
group_dict = {}
if opts.parents is not None:
    for parent in opts.parents:
        if parent is not None:
            
            directory_dict, dict_folders = inputfuncs.import_parent(parent, opts.basename)
            path_dict.update(directory_dict)

            # Create dict with subjects as keys and group (parents names) as values
            group_dict.update(dict_folders)

df_group = pd.DataFrame.from_dict(group_dict, orient='index', columns=["Folder"])
df_categories = pd.concat([df_categories, df_group], axis = 1)

# Warning for no imported subjects
if len(path_dict.values()) == 0:
    print('Error: No subjects were imported')
    print('Terminating program.\n')
    raise SystemExit(0)

# Check if we are importing masks
if opts.maskname is not None:
    mask_basename = opts.maskname
    dict_segmentation_methods['Imported Masks'] ='imported_mask'
else:
    mask_basename = None

# DATA PROCESSING -----------------------------------------------------------------------------

# Create dataframe for each segmentation method

scalar_statistics_names = ['FA','FA StdDev','MD','MD StdDev','RD','RD StdDev','AD','AD StdDev']
scalar_midline_names = list(range(0,200))

loaded_subjects = []

dict_segmentation_masks = {}
dict_scalar_maps = {}
dict_scalar_statistics = {}
dict_scalar_midlines = {}
dict_error_prob = {}
dict_parcellations_masks = {}
dict_parcellations_statistics = {}
dict_thickness = {}
dict_removed_subjects = {}
dict_scalar_outliers = {}

# Segment and get info

for subject_path in tqdm(path_dict.values()):
    
    for segmentation_method in dict_segmentation_methods.keys():    
        
        if segmentation_method not in dict_scalar_statistics.keys(): 
            dict_segmentation_masks[segmentation_method] = {}
            dict_scalar_maps[segmentation_method] = {}
            dict_scalar_statistics[segmentation_method] = {}
            dict_scalar_midlines[segmentation_method] = {}
            dict_error_prob[segmentation_method] = {}
            dict_parcellations_masks[segmentation_method] = {}
            dict_thickness[segmentation_method] = {}
            dict_removed_subjects[segmentation_method] = []
            dict_scalar_outliers[segmentation_method] = []


        # Get data path info
        folderpath = subject_path + 'inCCsight/'
        filename = 'segm_' + dict_segmentation_methods[segmentation_method] + '_data.npy'
        subject_name = os.path.basename(os.path.dirname(subject_path))

        # Process/Load data
        if segmentation_method == 'Imported Masks':
            if ccprocess.check_mask(subject_path, mask_basename) is False:
                continue

        data_tuple = ccprocess.segment(subject_path, 
                                       segmentation_method, 
                                       dict_segmentation_functions, 
                                       dict_parcellation_functions, 
                                       opts.basename,
                                       mask_basename)


        if data_tuple is None:
            continue
        #except:
        #    print('> Warning: Segmentation failed for subject {} with method {}'.format(subject_name, segmentation_method))
        #    continue

        segmentation_mask, scalar_maps, scalar_statistics, scalar_midlines, error_prob, parcellations_masks = data_tuple

        # Get thickness
        try:
            thick, _, _ = libcc.thickness(segmentation_mask, 200)
        except:
            thick = np.empty(200)
        
        # Assemble dictionaries
        dict_segmentation_masks[segmentation_method][subject_name] = segmentation_mask
        dict_scalar_maps[segmentation_method][subject_name] = scalar_maps
        dict_scalar_statistics[segmentation_method][subject_name] = scalar_statistics
        dict_scalar_midlines[segmentation_method][subject_name] = scalar_midlines
        dict_error_prob[segmentation_method][subject_name] = error_prob
        dict_parcellations_masks[segmentation_method][subject_name] = parcellations_masks
        dict_thickness[segmentation_method][subject_name] = thick
        
        # Save array with subject keys
        loaded_subjects.append(subject_name)
    
loaded_subjects = list(set(loaded_subjects))
loaded_subjects.sort()

for segmentation_method in dict_segmentation_methods.keys():

    # Convert to pandas dataframe
    dict_scalar_statistics[segmentation_method] = pd.DataFrame.from_dict(dict_scalar_statistics[segmentation_method], 
                                                                         orient='index', 
                                                                         columns=scalar_statistics_names)
    dict_scalar_midlines[segmentation_method] = pd.DataFrame.from_dict(dict_scalar_midlines[segmentation_method], 
                                                                       orient='index')
    dict_thickness[segmentation_method] = pd.DataFrame.from_dict(dict_thickness[segmentation_method],
                                                                 orient='index')
    dict_error_prob[segmentation_method] = pd.DataFrame.from_dict(dict_error_prob[segmentation_method], columns=['error_prob'], orient='index')
    
    dict_parcellations_statistics[segmentation_method] = inputfuncs.parcellations_dfs_dicts(dict_scalar_maps[segmentation_method], dict_parcellations_masks[segmentation_method], segmentation_method)

    # Get FA and/or other scalars ouliers
    for scalar in ['FA']:
        df = dict_scalar_statistics[segmentation_method][scalar]
        outliers = df[~df.between(df.quantile(.1), df.quantile(.9))].index
        dict_scalar_outliers[segmentation_method] += list(outliers)

# VISUALIZATION -------------------------------------------------------------------------------

app = dash.Dash(__name__, 
               meta_tags=[{"name": "viewport", "content": "width=device-width, initial-scale=.8, maximum-scale=.8"}],
               external_stylesheets = [dbc.themes.BOOTSTRAP],
               prevent_initial_callbacks=True)
server = app.server
app.config["suppress_callback_exceptions"] = True
app.title = 'inCCsight'


# ------------------------------- BUILD FUNCS  -----------------------------------------------

def build_banner():
    return html.Div(
        id="banner",
        className="twelve columns",
        children=[
            html.Div(
                className='twelve columns',
                style=dict(display='flex', justifyContent='flex-start'), 
                children=[
                    html.Img(src=app.get_asset_url("unicampw.png"), style=dict(height='9rem', marginBottom='2rem', padding='1rem')),
                    html.Img(src=app.get_asset_url("miclab.png"), style=dict(height='9rem', marginBottom='2rem', padding='1rem')),
                ]),
            html.Div(
                className='twelve columns',
                style=dict(display='flex', justifyContent='center'),
                children=[
                    html.Img(src=app.get_asset_url("inccsight.png"), style=dict(height='25rem')),
                ]),
        ],
    )

def build_graph_title(title):
    return html.P(className="graph-title", children=title)

# DataViz ------------------------------------------------------------------------------------

def build_group_segm_boxplot(mode='Method', segmentation_method='ROQS', extra_dims=list(df_numerical.columns)):

    std_colors = pio.templates[theme]['layout']['colorway']
        
    if mode == 'Method':
    
        scalar_names = ['FA', 'MD', 'RD', 'AD'] + extra_dims
    
        subplots = make_subplots(rows=1, cols=len(scalar_names), subplot_titles=scalar_names)
    
        for i, scalar in enumerate(scalar_names):
            
            for j, segmentation_method in enumerate(dict_segmentation_methods.keys()):
                
                if len(extra_dims) == 0:
                    df = pd.DataFrame()
                else:
                    df = df_numerical[extra_dims]

                df = pd.concat([df, dict_scalar_statistics[segmentation_method]], axis=1)
                df = df.drop(dict_removed_subjects[segmentation_method])
                
                subplots.add_trace(go.Box(y=df[scalar], name=segmentation_method, legendgroup=segmentation_method, hovertext=df.index, marker=dict(color=std_colors[j])), row=1, col=i+1)

                if i == 0:
                    subplots.data[-1].update(name=segmentation_method, legendgroup=segmentation_method)
                else:
                    subplots.data[-1].update(showlegend=False)

    else:

        scalar_names = ['FA', 'MD', 'RD', 'AD'] + extra_dims
        subplots = make_subplots(rows=1, cols=len(scalar_names), subplot_titles=scalar_names)
        
        if len(extra_dims) == 0:
            df = pd.DataFrame()
        else:
            df = df_numerical[extra_dims]
    
        df = pd.concat([df, df_categories[mode], dict_scalar_statistics[segmentation_method]], axis=1)
        df = df.drop(dict_removed_subjects[segmentation_method])
        df = df.dropna(axis=0)
        
        categories = set(df[mode])
    
        for i, scalar in enumerate(scalar_names):
            
            for j, category in enumerate(categories):
                
                subplots.add_trace(go.Box(y=df[df[mode] == category][scalar], name=category, legendgroup=category, hovertext=df.index, marker=dict(color=std_colors[j])), row=1, col=i+1)

                if i == 0:
                    subplots.data[-1].update(name=category, legendgroup=category)
                else:
                    subplots.data[-1].update(showlegend=False)

    subplots.update_layout(height=400, paper_bgcolor='rgba(0,0,0,0)', legend_orientation="h")
    subplots.update_layout(font=dict(family="Open Sans, sans-serif", size=12), margin=dict(r=0, l=0, t=60))
    
    return subplots

def build_parcel_boxplot(scalar='FA', mode='Method', segmentation_method='ROQS', parcellation_method='Witelson'):

    std_colors = pio.templates[theme]['layout']['colorway']
    
    list_regions = ['P1', 'P2', 'P3', 'P4', 'P5']
    subplots = make_subplots(rows=1, cols=5, subplot_titles=list_regions)
    
    if mode == 'Method':
        
        for i, region in enumerate(list_regions):
            
            for j, segmentation_method in enumerate(dict_segmentation_methods.keys()):
                df = dict_parcellations_statistics[segmentation_method][parcellation_method][region][scalar]
                df = df.drop(dict_removed_subjects[segmentation_method])
                df = df.dropna(axis=0)
                subplots.add_trace(go.Box(y=df, name=segmentation_method, legendgroup=segmentation_method, hovertext=df.index, marker=dict(color=std_colors[j])), row=1, col=i+1)

                if i == 0:
                    subplots.data[-1].update(name=segmentation_method, legendgroup=segmentation_method)
                else:
                    subplots.data[-1].update(showlegend=False)

    else:
    
        categories = list(set(df_categories[mode]))

        df = pd.DataFrame()

        for category in categories:
            df_aux = pd.DataFrame()
            for region in list_regions:
                df_aux = pd.concat([df_aux, dict_parcellations_statistics[segmentation_method][parcellation_method][region][scalar][df_categories[mode] == category]], axis=1)
            df_aux[mode] = category
            df = pd.concat([df, df_aux], axis=0)
        
        df = df.drop(dict_removed_subjects[segmentation_method])
        df = df.dropna(axis=0)

        names = list_regions + [mode]
        df.columns = names

        categories = set(df[mode])
        
        for i, region in enumerate(list_regions):
            
            for j, category in enumerate(categories):
                
                subplots.add_trace(go.Box(y=df[df[mode] == category][region], name=category, legendgroup=category, hovertext=df.index, marker=dict(color=std_colors[j])), row=1, col=i+1)

                if i == 0:
                    subplots.data[-1].update(name=category, legendgroup=category)
                else:
                    subplots.data[-1].update(showlegend=False)

    subplots.update_layout(height=400, paper_bgcolor='rgba(0,0,0,0)', legend_orientation="h")
    subplots.update_layout(font=dict(family="Open Sans, sans-serif", size=12), margin=dict(r=0, l=0, t=60))
    
    return subplots

def build_segm_scatterplot(mode='Method', segmentation_method = 'ROQS', scalar_x = 'FA', scalar_y = 'MD', trendline=None):

    df = pd.DataFrame()

    if mode == 'Method':

        for segmentation_method in dict_segmentation_methods.keys():
            df_aux = dict_scalar_statistics[segmentation_method]
            df_aux['Method'] = segmentation_method 
            df = pd.concat([df, df_aux], axis=0)

    else:
        df = pd.concat([df_categories[mode], dict_scalar_statistics[segmentation_method]], axis=1)
    df = df.join(df_numerical, how='outer')
        
    df = df.drop(dict_removed_subjects[segmentation_method])
    df = df.dropna(axis=0)

    print(df)

    fig = px.scatter(df, 
                    x=scalar_x, 
                    y=scalar_y, 
                    color=mode,
                    marginal_y="violin", 
                    marginal_x="histogram",
                    hover_name=df.index,
                    trendline=trendline)

    fig.update_layout(height=800, paper_bgcolor='rgba(0,0,0,0)')
    fig.update_layout(legend=dict(orientation="h", yanchor="bottom", y=1.02, xanchor='right', x=1))
    fig.update_layout(font=dict(family="Open Sans, sans-serif", size=12))
    
    return fig

def build_segm_scattermatrix(mode='Method', segmentation_method = 'ROQS', extra_dims = list(df_numerical.columns)):

    dimensions = ['FA','MD','RD','AD']
    df = pd.DataFrame()

    if mode == 'Method':
        for segmentation_method in dict_segmentation_methods.keys():
            df_aux = dict_scalar_statistics[segmentation_method]
            df_aux['Method'] = segmentation_method 
            df = pd.concat([df, df_aux], axis=0)
    
    else:
        df = pd.concat([df_categories[mode], dict_scalar_statistics[segmentation_method]], axis=1)
    
        if len(extra_dims) > 0:
            dimensions = ['FA','MD','RD','AD'] + extra_dims
            df = pd.concat([df, df_numerical], axis=1)

    df = df.drop(dict_removed_subjects[segmentation_method])
    df = df.dropna(axis=0)

    fig = px.scatter_matrix(df, 
                            dimensions=dimensions,
                            color=mode,
                            hover_name=df.index)

    if mode == 'Method':
        n_cats = len(dict_segmentation_methods.keys())
    else:
        n_cats = len(set(df_categories[mode].dropna(axis=0)))

    fig.update_layout(legend=dict(orientation="h", yanchor="bottom", y=1.02, xanchor='right', x=1))
    fig.update_layout(height=500+250*n_cats, paper_bgcolor='rgba(0,0,0,0)')
    fig.update_layout(font=dict(family="Open Sans, sans-serif", size=12), margin=dict(r=0, l=0))

    return fig

def build_midline_plot(scalar='FA', mode='Method', segmentation_method='ROQS'):

    df = pd.DataFrame()
    
    if mode == 'Method':
        for segmentation_method in dict_segmentation_methods.keys():
            
            if scalar in ['FA', 'MD', 'AD', 'RD']:
                df_aux = pd.DataFrame.from_dict(dict(dict_scalar_midlines[segmentation_method][scalar]), orient='index', columns=scalar_midline_names)
            elif scalar == 'Thickness':
                df_aux = dict_thickness[segmentation_method]
            
            df_aux['Method'] = segmentation_method    
            df = pd.concat([df, df_aux], axis=0)
        df = df.drop(dict_removed_subjects[segmentation_method])
        df = df.dropna(axis=0)
            
        df_grouped = df.groupby('Method').mean().transpose()
        df_melt = pd.melt(df_grouped.reset_index(), id_vars='index', value_vars=set(df[mode]))
    
    else:
        if scalar in ['FA', 'MD', 'AD', 'RD']:
            df_aux = pd.DataFrame.from_dict(dict(dict_scalar_midlines[segmentation_method][scalar]), orient='index', columns=scalar_midline_names)
        elif scalar == 'Thickness':
            df_aux = dict_thickness[segmentation_method]
            
        df = pd.concat([df_categories[mode], df_aux], axis=1)
        df = df.drop(dict_removed_subjects[segmentation_method])
        df = df.dropna(axis=0)

        df_grouped = df.groupby(mode).mean().transpose()
        df_melt = pd.melt(df_grouped.reset_index(), id_vars='index', value_vars=set(df[mode]))
        
    fig = px.line(df_melt, x='index', y='value', color=mode)

    fig.update_layout(legend=dict(orientation="h", yanchor="bottom", y=1.02, xanchor='right', x=1))
    fig.update_layout(height=400, paper_bgcolor='rgba(0,0,0,0)')    
    fig.update_layout(font=dict(family="Open Sans, sans-serif", size=12))
    
    fig.update_layout(xaxis_title='Points along CC body', yaxis_title=scalar, legend_title=mode)
    
    return fig

def build_bubble_grouped(mode='Method', segmentation_method='ROQS', scalar='Thickness', size=True):
    
    def build_bubble_plot(scalar='FA', segmentation_method='Watershed', size = True, category_index = None):

        df_pts = pd.read_pickle('./assets/bubble_plot_xy.pkl')
        segm_contour = np.load('./assets/bubble_plot_contour.npy')

        if scalar == 'Thickness':
            df_aux = dict_thickness[segmentation_method][list(np.linspace(0,195,40))+[199]]    
        else:
            df_aux = pd.DataFrame.from_dict(dict(dict_scalar_midlines[segmentation_method][scalar]), orient='index', columns=scalar_midline_names)[list(np.linspace(0,195,40))+[199]]
        
        df_aux = df_aux.drop(dict_removed_subjects[segmentation_method])

        if category_index is not None:
            df_aux = df_aux.loc[np.intersect1d(df_aux.index, category_index)]

        df_aux = df_aux.dropna(axis=0)

        df_pts[scalar] = df_aux.mean().reset_index()[0]

        if size == True:
            fig = px.scatter(df_pts, x="x", y="y", color = scalar, size = df_pts[scalar])
            fig.update_traces(marker=dict(sizeref = 2. * max(df_pts[scalar]) / (45 ** 2)))
        else:
            fig = px.scatter(df_pts, x="x", y="y", color=scalar)
            fig.update_traces(marker=dict(size=25))

        fig.update_layout(paper_bgcolor='rgba(0,0,0,0)', legend_orientation="h")
        fig.update_layout(font=dict(family="Open Sans, sans-serif", size=12))

        fig.add_trace(go.Contour(z=segm_contour, contours=dict(start=0, end=70, size=70, coloring='none'),
                                 showscale=False, line_width=3, line=dict(color='rgba(0,0,0,.5)', dash='dot')))

        return fig
    
    if mode == 'Method':
        
        n_cats = len(dict_segmentation_methods.keys())
        fig = make_subplots(rows=n_cats, cols=1, vertical_spacing=0.1/n_cats)
        
        for i, segmentation_method in enumerate(dict_segmentation_methods):
            fig.add_trace(build_bubble_plot(scalar=scalar, segmentation_method=segmentation_method, size=size)['data'][0], row=i+1, col=1)
            fig.update_yaxes(title_text="{} for<br>{} method".format(scalar, segmentation_method), row=i+1, col=1)
        
    else:
        
        df = df_categories[mode]
        df = df.drop(dict_removed_subjects[segmentation_method])
        df = df.dropna(axis=0)

        n_cats = len(set(df))
        fig = make_subplots(rows=n_cats, cols=1, vertical_spacing=0.1/n_cats)
        
        for i, category in enumerate(set(df)):
            category_index = df_categories.loc[df_categories[mode] == category].index
            fig.add_trace(build_bubble_plot(scalar=scalar, segmentation_method=segmentation_method, size=size, category_index=category_index)['data'][0], row=i+1, col=1)
            fig.update_yaxes(title_text="{} for<br>{} category".format(scalar, category), row=i+1, col=1)
    
    fig.update_xaxes(title_text="Points along CC body", row=n_cats, col=1)
    fig.update_layout(height=250*n_cats, paper_bgcolor='rgba(0,0,0,0)') 
    fig.update_layout(font=dict(family="Open Sans, sans-serif", size=12), margin=dict(r=0, l=0, t=60))
    
    return fig

def build_bubble_grouped_pvalue(mode='Method', segmentation_method='ROQS', scalar='Thickness', threshold=0.05):
    
    def build_bubble_pvalue(pvalue, threshold=0.05, size=False, gray=False):

        df_pts = pd.read_pickle('./assets/bubble_plot_xy.pkl')
        segm_contour = np.load('./assets/bubble_plot_contour.npy')
        marker_color = 'rgba(100,100,100,0.5)'

        df_pts['p-value'] = pvalue

        if gray:
            fig = px.scatter(df_pts, x="x", y="y", hover_data=['p-value'])
            fig.update_traces(marker=(dict(color=marker_color)))
        else:
            fig = px.scatter(df_pts.loc[df_pts['p-value'] < threshold], x="x", y="y", color = 'p-value')
        fig.update_traces(marker=dict(size=25))

        fig.update_layout(paper_bgcolor='rgba(0,0,0,0)', legend_orientation="h")    
        fig.update_layout(font=dict(family="Open Sans, sans-serif", size=12))

        return fig
    
    if mode == 'Method':
        categories = list(itertools.combinations(dict_segmentation_methods.keys(), 2))
        n_cats = len(categories)

        if n_cats == 0:
            return empty_figure_with_text('Not enough categories to calculate p-values.')
        
        fig = make_subplots(rows=n_cats, cols=1)
        
        for i, category in enumerate(categories):
            
            if scalar == 'Thickness':
                pvalue = stats.ttest_ind(dict_thickness[category[0]], dict_thickness[category[1]]).pvalue
            else:
                pvalue = stats.ttest_ind(pd.DataFrame.from_dict(dict(dict_scalar_midlines[category[0]][scalar]), orient='index', columns=scalar_midline_names),
                                   pd.DataFrame.from_dict(dict(dict_scalar_midlines[category[1]][scalar]), orient='index', columns=scalar_midline_names)).pvalue

            pvalue = np.take(pvalue, list(np.linspace(0,195,40)))

            new_gray_fig = build_bubble_pvalue(pvalue, gray=True)
            
            new_fig = build_bubble_pvalue(pvalue, threshold)['data']
            
            if len(new_fig) > 0:
                fig.add_trace(new_gray_fig['data'][0], row=i+1, col=1)
                fig.add_trace(new_fig[0], row=i+1, col=1)
            else:
                fig.add_trace(new_gray_fig['data'][0], row=i+1, col=1)
                
            fig.update_yaxes(title_text="{} x {}".format(category[0], category[1]), row=i+1, col=1)
            
    else:
        
        df = df_categories[mode]
        df = df.dropna(axis=0)
        categories = list(itertools.combinations(set(df), 2))
        n_cats = len(categories)
        
        if n_cats == 0:
            return empty_figure_with_text('Not enough categories to calculate p-values.')

        fig = make_subplots(rows=n_cats, cols=1, x_title='Statistic Meaningful Differences (p < 0.05)')
        
        for i, category in enumerate(categories):
            
            if scalar == 'Thickness':
                pvalue = stats.ttest_ind(dict_thickness[segmentation_method].loc[df_categories[mode] == category[0]], 
                                   dict_thickness[segmentation_method].loc[df_categories[mode] == category[1]]).pvalue
            else:
                pvalue = stats.ttest_ind(pd.DataFrame.from_dict(dict(dict_scalar_midlines[segmentation_method][scalar]), orient='index', columns=scalar_midline_names).loc[df_categories[mode] == category[0]],
                                   pd.DataFrame.from_dict(dict(dict_scalar_midlines[segmentation_method][scalar]), orient='index', columns=scalar_midline_names).loc[df_categories[mode] == category[1]]).pvalue
    
            pvalue = np.take(pvalue, list(np.linspace(0,195,40)))

            new_gray_fig = build_bubble_pvalue(pvalue, gray=True)
            
            new_fig = build_bubble_pvalue(pvalue, threshold)['data']
            
            if len(new_fig) > 0:
                fig.add_trace(new_gray_fig['data'][0], row=i+1, col=1)
                fig.add_trace(new_fig[0], row=i+1, col=1)
            else:
                fig.add_trace(new_gray_fig['data'][0], row=i+1, col=1)
                
            fig.update_yaxes(title_text="{} x {}".format(category[0], category[1]), row=i+1, col=1)

    fig.update_layout(height=400*n_cats) 
    fig.update_xaxes(title_text="Points along CC body", row=n_cats, col=1)
    fig.update_layout(paper_bgcolor='rgba(0,0,0,0)', legend_orientation="h")    
    fig.update_layout(font=dict(family="Open Sans, sans-serif", size=12))

    return fig

# Dropdowns --------------------------------------------------------------------------------

def build_midlineplot_dropdown():

    options = [{'label': scalar, 'value': scalar} for scalar in scalar_list+['Thickness']]

    layout = html.Div([
                html.Div([
                    html.H6('Scalar:', className='table-options-title', style={'padding':'0px 10px 0px 10px'}),
                    dcc.Dropdown(id='dropdown-midline-scalars',
                                 options=options,
                                 multi=False,
                                 value='FA',
                                 style={'width':'120px'}),
                ], className='row', style=dict(display='flex', justifyContent='left', verticalAlign='center')),
            ]
        )
    return layout

def build_segm_scatterplot_dropdowns():

    options = [{'label': scalar, 'value': scalar} for scalar in scalar_list + list(df_numerical.columns)]
    options_trendlines = [{'label': scalar, 'value': scalar} for scalar in ['None', 'OLS', 'Lowess']]

    layout = html.Div([
                html.Div([
                    html.Div([
                        html.H6('Scalar Y:', className='table-options-title', style={'padding':'0px 10px 0px 0px'}),
                        dcc.Dropdown(id='dropdown-scalars-left',
                                     options=options,
                                     multi=False,
                                     value='FA',
                                     style={'width':'90px'}),
                    ], className='row', style={'margin':'0px 0px 0px 10px'}),
                    html.Div([
                        html.H6('Scalar X:', className='table-options-title', style={'padding':'0px 10px 0px 0px'}),
                        dcc.Dropdown(id='dropdown-scalars-right',
                                     options=options,
                                     multi=False,
                                     value='MD',
                                     style={'width':'90px'}),
                    ], className='row', style={'margin':'0px 0px 0px 30px'}),
                    html.Div([
                        html.H6('Trendline:', className='table-options-title', style={'padding':'0px 10px 0px 0px'}),
                        dcc.Dropdown(id='dropdown-scalars-trendline',
                                     options=options_trendlines,
                                     multi=False,
                                     value='None',
                                     style={'width':'120px'}),

                    ], className='row', style={'margin':'0px 0px 0px 30px'}),
                ], className='row', style=dict(display='flex', justifyContent='left', verticalAlign='center')),
            ]
        )
    return layout 

def build_parcel_boxplot_dropdowns():

    options_scalars = [{'label': scalar, 'value': scalar} for scalar in scalar_list]
    options_parcel_method = [{'label': parc, 'value': parc} for parc in dict_parcellations_statistics[segmentation_method].keys()]

    layout = html.Div([
                html.Div([
                    html.H6('Parc. Method:', className='table-options-title', style={'padding':'0px 10px 0px 10px'}),
                    dcc.Dropdown(id='dropdown-parcel-boxplot-left',
                                 options=options_parcel_method,
                                 multi=False,
                                 value=list(dict_parcellation_methods.keys())[0],
                                 style={'width':'150px'}),
                    html.H6('Scalar:', className='table-options-title', style={'padding':'0px 10px 0px 30px'}),
                    dcc.Dropdown(id='dropdown-parcel-scalars-right',
                                 options=options_scalars,
                                 multi=False,
                                 value='FA',
                                 style={'width':'120px'})
                ], className='row', style=dict(display='flex', justifyContent='left', verticalAlign='center')),
            ]
        )
    return layout    

def build_bubbleplot_dropdowns():

    options_pvalue = [{'label': scalar, 'value': scalar} for scalar in ['Scalar', 'p-value']]
    options_scalars = [{'label': scalar, 'value': scalar} for scalar in scalar_list+['Thickness']]
    options_size = [{'label': scalar, 'value': scalar} for scalar in ['True', 'False']]

    layout = html.Div([
                html.Div([

                    html.Div([

                        html.H6('Mode:', className='table-options-title', style={'padding':'0px 10px 0px 0px'}),
                        dcc.Dropdown(id='dropdown-bubbleplot-mode',
                                     options=options_pvalue,
                                     multi=False,
                                     value='Scalar',
                                     style={'width':'150px'}),

                    ], className='row', style={'margin':'0px 0px 0px 10px'}),

                    html.Div([

                        html.H6('Scalar:', className='table-options-title', style={'padding':'0px 10px 0px 0px'}),
                        dcc.Dropdown(id='dropdown-bubbleplot-left',
                                     options=options_scalars,
                                     multi=False,
                                     value='Thickness',
                                     style={'width':'150px'}),

                    ], className='row', style={'margin':'0px 0px 0px 30px'}),

                    html.Div([

                        html.H6('Size:', className='table-options-title', style={'padding':'0px 10px 0px 0px'}),
                        dcc.Dropdown(id='dropdown-bubbleplot-right',
                                     options=options_size,
                                     multi=False,
                                     value='True',
                                     style={'width':'120px'})

                    ], className='row', style={'margin':'0px 0px 0px 30px'}, id='div-bubble-dropdown-right'),

                    html.Div([

                        html.H6('Threshold:', className='table-options-title', style={'padding':'0px 10px 0px 0px'}),
                        dcc.Dropdown(id='dropdown-bubbleplot-threshold',
                                     options=[{'label': num/100, 'value': num/100} for num in list(np.arange(0, 10, 1)) + list(np.arange(15, 95, 5))],
                                     multi=False,
                                     value=0.05,
                                     style={'width':'120px'})

                    ], className='row', style=dict(display='none', margin='0px 0px 0px 30px'), id='div-bubble-dropdown-threshold'),
                
                ], className='row', style=dict(display='flex', justifyContent='left', verticalAlign='center', marginBottom='50px')),
            ]
        )

    return layout    

def build_fissure_image_dropdown(subject_id, segmentation_methods_available=dict_segmentation_methods.keys()):

    options = [{'label': segmentation_method, 'value': segmentation_method} for segmentation_method in list(segmentation_methods_available)+['None']]
    options_scalars = [{'label': scalar, 'value': scalar} for scalar in ['wFA']+scalar_list]

    layout = html.Div([
                html.Div([
                    
                    html.Div(
                        children=[
                            html.H6('Segm. Method:', className='table-options-title', style={'padding':'0px 20px 0px 10px'}),
                            
                            dcc.Dropdown(id='dropdown-subj-collapse-segm-methods',
                                         options=options,
                                         multi=False,
                                         value=segmentation_methods_available[0],
                                         style={'width':'120px'}),
                        ], className='row'),
                    
                    html.Div(
                        dbc.Button('Remove segmentation', 
                                outline=True,
                                color='danger',
                                id=dict(type='btn-remove-subject', index=subject_id),
                                style=dict(padding='0 15px', margin='0px 0px 0px 20px', fontSize='1.2rem')
                            ),
                        id=f"tooltip-div-wrapper-{subject_id}"),
                    
                ], className='row', style=dict(display='flex', justifyContent='left', verticalAlign='center', marginLeft='0')),

                html.Div(
                    children=[
                        html.H6('Scalar:', className='table-options-title', style={'padding':'0px 10px 0px 10px'}),
                        
                        dcc.Dropdown(id='dropdown-subj-collapse-scalars',
                                     options=options_scalars,
                                     multi=False,
                                     value='wFA',
                                     style={'width':'150px'}),
                    ], className='row', style=dict(display='flex', justifyContent='left', verticalAlign='center', marginTop='2px')),

                dbc.Tooltip("This will remove only the ROQS segmentation, indicated on the dropdown to the left",
                            target=f"tooltip-div-wrapper-{subject_id}",
                            id='subject-collapse-tooltip',
                            style=dict(fontSize='12pt'),
                            placement='bottom',
                            ),

            ]
        )


    return layout   

def build_individual_subject_segm_table_dropdown(segmentation_methods_available=dict_segmentation_methods.keys()):

    options = [{'label': segmentation_method, 'value': segmentation_method} for segmentation_method in segmentation_methods_available]
    options_stddev = [{'label': scalar, 'value': scalar} for scalar in ['Show', 'Hide']]

    layout = html.Div([
                html.Div([
                    html.H6('Segm. method:', className='table-options-title', style={'padding':'0px 10px 0px 10px'}),
                    dcc.Dropdown(id='dropdown-subj-collapse-table-segm-methods',
                                 options=options,
                                 multi=False,
                                 value=segmentation_methods_available[0],
                                 style={'width':'150px'}),
                    html.H6('Show Std.Dev.:', className='table-options-title', style={'padding':'0px 10px 0px 30px'}),
                    dcc.Dropdown(id='dropdown-subj-collapse-table-segm-std-dev',
                                 options=options_stddev,
                                 multi=False,
                                 value='Hide',
                                 style={'width':'120px'})
                ], className='row', style=dict(display='flex', justifyContent='left', verticalAlign='center')),
            ]
        )

    return layout

def build_individual_subject_parcel_table_dropdown(segmentation_methods_available=dict_segmentation_methods.keys()):

    options = [{'label': segmentation_method, 'value': segmentation_method} for segmentation_method in segmentation_methods_available]
    options_scalars = [{'label': scalar, 'value': scalar} for scalar in scalar_list]
    options_parcel_method = [{'label': parc, 'value': parc} for parc in dict_parcellations_statistics[segmentation_method].keys()]

    layout = html.Div([
                html.Div([
                    html.H6('Segm. method:', className='table-options-title', style={'padding':'0px 10px 0px 10px'}),
                    dcc.Dropdown(id='dropdown-subj-collapse-table-parcel-segm-methods',
                                 options=options,
                                 multi=False,
                                 value=segmentation_methods_available[0],
                                 style={'width':'150px'}),
                    html.H6('Parcel. method:', className='table-options-title', style={'padding':'0px 10px 0px 30px'}),
                    dcc.Dropdown(id='dropdown-subj-collapse-table-parcel-methods',
                                 options=options_parcel_method,
                                 multi=False,
                                 value=list(dict_parcellation_methods.keys())[0],
                                 style={'width':'150px'}),
                    html.H6('Scalar:', className='table-options-title', style={'padding':'0px 10px 0px 30px'}),
                    dcc.Dropdown(id='dropdown-subj-collapse-table-parcel-scalars',
                                 options=options_scalars,
                                 multi=False,
                                 value='FA',
                                 style={'width':'120px'})
                ], className='row', style=dict(display='flex', justifyContent='left', verticalAlign='center')),
            ]
        )
    return layout 

# Collapses and others ---------------------------------------------------------------------

def build_subjects_list():
    '''
    button_group = dbc.ButtonGroup(
        [dbc.Button(i, color='light', size='lg', style=dict(width='100%')) for i in loaded_subjects],
        vertical=True,
        style=dict(width='100%')
    )
    '''

    button_group = []
    for i, subject_id in enumerate(loaded_subjects):

        if i%2 == 0:
            background = 'rgba(50,50,70,.5)'
        else:
            background = 'rgba(60,60,70,.5)'

        button_group.append(html.Button(subject_id, 
            style=dict(fontSize='1.8rem', width='100%', backgroundColor=background, marginBottom='2px', color='rgb(255,255,255)'), 
            id={'type': 'subject-list-btns', 'index': subject_id}))
        

    return html.Div(button_group, style=dict(width='100%'))

def build_quality_collapse():

    layout = html.Div([
                
                html.Div([
                    build_graph_title("Quality evaluation"),
                    html.Button('X', style=dict(fontSize='1.5rem', margin='10px', padding='0 13px', fontFamily= 'Open Sans', borderRadius='20px'),
                                id='btn-exit-quality')
                
                ], className='twelve columns', style=dict(display='flex', justifyContent='space-between')),

                html.Div([
                        html.Div([
                            html.H6("Threshold:", style=dict(fontSize='1.8rem')), 
                            dcc.Dropdown(id='dropdown-quality-threshold',
                                         options= [{'label': num/100, 'value': num/100} for num in np.arange(95, 5, -5)],
                                         multi=False,
                                         value='0.7',
                                         style={'width':'100px', 'marginLeft':'5px'}),
                            html.H6("Scalar:", style=dict(fontSize='1.8rem', marginLeft='2rem')), 
                            dcc.Dropdown(id='dropdown-quality-scalar',
                                         options= [{'label': i, 'value': i} for i in ['wFA']+scalar_list],
                                         multi=False,
                                         value='wFA',
                                         style={'width':'100px', 'marginLeft':'5px'}),
                            ], className='row', style=dict(margin='20px 0 0 12px')),
                ], className='twelve columns', style=dict(display="flex", justifyContent="space-between")),
                
                html.Div([
                    dbc.Button("Restore Removed", color='info', outline=True, size='lg', id='restore_btn', style=dict(marginRight='1rem')),
                    dbc.Button("Remove Selected", color='danger', outline=True, size='lg', id='remove_btn', style=dict(marginRight='2rem')),
                ], className='twelve columns', style=dict(display='flex', justifyContent='flex-end')),

                html.Div([
                    dbc.Button("Unselect all", color='primary', outline=True, size='lg', id='unselect_all_btn', style=dict(marginRight='1rem')),
                    dbc.Button("Select all", color='primary', outline=True, size='lg', id='select_all_btn', style=dict(marginRight='2rem')),
                ], className='twelve columns', style=dict(display='flex', verticalAlign='center', justifyContent='flex-end', padding='10px 0px 10px 0px', marginBottom='-1rem')),

                html.Div(children=build_quality_images(), 
                    className="twelve columns", 
                    id='photo-container',
                    style=dict(margin="0rem 0rem 2rem 0rem")),
            
            ], style={'margin':'20px', 'height':"100vh", 'backgroundColor':'#FAFAFA', 'border-radius':'20px', 'border':'1px solid rgba(0,0,0,.125)'})

    return layout

def build_quality_images(threshold=0.7, scalar='wFA'):
    
    def get_quality_tab_children(segmentation_method, scalar='wFA'):
        
        children = []

        # Get error probs
        df = dict_error_prob[segmentation_method]
        df = df.drop(dict_removed_subjects[segmentation_method])
        
        # Order by error probs
        index_error_probs = df.query('error_prob > '+str(threshold)).index.tolist()
        index_error_probs.sort()

        # Get ouliers
        index_outliers = dict_scalar_outliers[segmentation_method]
        index_outliers = [x for x in index_outliers if x not in dict_removed_subjects[segmentation_method]]
        index_outliers.sort()

        index_no_quality = list(set(index_error_probs + index_outliers))

        # Retrieve images and segmentation
        for subject_id in index_no_quality:
            children.append(dcc.Loading(
                            html.Div([
                                html.Div([
                                    html.H6("Subject: {}".format(subject_id), 
                                        style=dict(fontSize='2rem')),
                                    dcc.Checklist(
                                        id={'type': 'remove-cbx', 'index': 'cbx-{}-{}'.format(segmentation_method, subject_id)},
                                        options=[{'label': 'Remove', 'value': 'Remove'}], 
                                        value=[], 
                                        style=dict(fontSize='1.8rem'), inputStyle=dict(marginRight="10px")),  

                                    ], className='twelve columns', style=dict(width='100%', display='flex', verticalAlign='center', justifyContent='space-between')),
                                
                                build_quality_badges(subject_id, index_error_probs, index_outliers),

                                html.Div([
                                    dcc.Graph(figure=build_fissure_image(subject_id, segmentation_method, scalar))
                                    ], className='twlve columns'),
                            
                            ], className = 'twelve columns')))
        return children


    def get_quality_tab(segmentation_method):
        tab = dbc.Tab(label=segmentation_method, children=html.Div(get_quality_tab_children(segmentation_method, scalar), style=dict(height='80vh', overflowY="auto", padding='20px 20px 20px 20px')))
        return tab

    tabs = []

    for segmentation_method in dict_segmentation_methods.keys():
        tabs.append(get_quality_tab(segmentation_method))

    return dbc.Tabs(tabs, style=dict(height='40px', verticalAlign='center', padding='0px 10px 0px 10px'))

def build_fissure_image(subject_id, segmentation_method, scalar = 'FA'):

    scalar_maps = dict_scalar_maps[segmentation_method][subject_id]
    scalar_maps_list = ['wFA','FA','MD','RD','AD']
    scalar_map = scalar_maps[scalar_maps_list.index(scalar)]

    fig = px.imshow(scalar_map, color_continuous_scale='gray', aspect='auto')

    if segmentation_method != 'None':

        segmentation = dict_segmentation_masks[segmentation_method][subject_id]

        contours = measure.find_contours(segmentation, 0.1)
        contour = sorted(contours, key=lambda x: len(x))[-1]

        fig.add_trace(go.Scatter(x=contour[:, 1], y=contour[:, 0]))
    
    fig.update_layout(height=250, width=450, paper_bgcolor='rgba(0,0,0,0)', legend_orientation="h", coloraxis_showscale=True)
    fig.update_layout(margin = dict(l=0, r=0,t=0,b=30))

    return fig 

def build_3d_visualization(subject_id):

    folderpath = path_dict[subject_id] + 'inCCsight/'
    filepath = folderpath + 'segm_watershed3d.npy'

    if os.path.exists(filepath):
       
        _, segmentation3d, wFA_v, _ = np.load(filepath, allow_pickle=True)

        verts, faces, normals, values = measure.marching_cubes_lewiner(wFA_v, 0)

        tri_FA = ff.create_trisurf(x=verts[:,0], y=verts[:,2], z=verts[:,1]*-1+70,
                                simplices=faces,
                                colormap=[(1,0,0), (1,0,0)],
                                aspectratio=dict(x=1, y=1, z=.66),
                                plot_edges = False,
                                show_colorbar = False)

        tri_FA['data'][0].update(opacity=0.2)

        verts, faces, normals, values = measure.marching_cubes_lewiner(segmentation3d, 0)

        tri_CC = ff.create_trisurf(x=verts[:,0], y=verts[:,2], z=verts[:,1]*-1+70,
                                simplices=faces,
                                colormap=[(0,0,1), (0,0,1)],
                                aspectratio=dict(x=1, y=1, z=.66),
                                plot_edges = False,
                                show_colorbar = False)

        tri_CC['data'][0].update(opacity=0.1)

        fig = go.Figure(tri_FA)
        fig.add_trace(tri_CC.data[0])
        fig.update_layout(title="3D Visualization")
        fig.update_layout(plot_bgcolor='rgba(0,0,0,0)', paper_bgcolor='rgba(0,0,0,0)')

        return fig

    else:
        return []

def build_subject_collapse(segmentation_method='ROQS', scalar_map='wFA', subject_id = list(path_dict.keys())[0]):

    segmentation_methods_available = []
    for segm_method in dict_segmentation_methods.keys():
        if subject_id in dict_scalar_statistics[segm_method].index:
            segmentation_methods_available.append(segm_method)

    layout = dbc.Card([

                html.Div([
                    html.Div([
                        build_graph_title("Subject " + subject_id),
                        ], className='row'),
                    html.Button('X', style=dict(fontSize='1.5rem', margin='10px', padding='0 13px', fontFamily= 'Open Sans', borderRadius='20px'),
                                id=dict(type='btn-exit-subject', index=subject_id))
                
                ], className='twelve columns', style=dict(display='flex', justifyContent='space-between')),

                html.Div([                
    
                    # Segmentation image
                    html.Div([
                        build_graph_title("Scalar maps"),
                        #dcc.Graph(figure=build_3d_visualization(subject_id))
                        dcc.Graph(figure=build_fissure_image(subject_id, segmentation_methods_available[0], scalar = scalar_map), id='subject_collapse_fissure_img'),
                        build_fissure_image_dropdown(subject_id, segmentation_methods_available),

                    ], className = 'four columns', style=dict(display='grid', justifyContent='center')),

                    html.Div(className='one column'),

                    html.Div([
                        build_graph_title("Segmentation data"),
                        build_individual_subject_segm_table(subject_id, segmentation_methods_available[0]),
                        build_individual_subject_segm_table_dropdown(segmentation_methods_available),

                        build_graph_title("Parcellation data"),
                        build_individual_subject_parcel_table(subject_id, segmentation_methods_available[0], parcellation_method='Witelson', scalar='FA'),
                        build_individual_subject_parcel_table_dropdown(segmentation_methods_available),

                    ], className='six columns'),

                ], className='row', style={'justifyContent':'center'}),

            ], style={'margin':'20px', 'backgroundColor':'#FAFAFA', 'border-radius':'20px', 'padding':'0px 0px 50px 0px'})
    return layout

def build_quality_badges(subject_id, index_error_probs, index_outliers):

    children = []

    if subject_id in index_error_probs:
        children.append(dbc.Badge("Abnormal Shape", color="secondary", pill=True, style=dict(marginRight='10px', fontSize='12pt', fontWeight='600')))

    if subject_id in index_outliers:
        children.append(dbc.Badge("FA Outlier", color="secondary", pill=True, style=dict(marginRight='10px', fontSize='12pt', fontWeight='600')))

    return html.Div(children=children, className='twelve columns', style=dict(marginLeft='27px'))

# DataTable functions ---------------------------------------------------------------------

def extend_colorscale(colormap, factor = 5):
    
    from plotly.colors import n_colors
    
    new_colormap = []
    for i in range(len(colormap)-1):
        new_colormap += n_colors(eval(colormap[i][3:]), eval(colormap[i+1][3:]), factor)
    
    for i, color in enumerate(new_colormap):
        new_colormap[i] = 'rgb' + str(color)
        
    return new_colormap
   
def color_table(df, use_values_limits = True, mean = None, stdev = None):

    if 'id' in df:
        numeric_columns = df.select_dtypes('number').drop(['index'], axis=1)
    else:
        numeric_columns = df.select_dtypes('number')

    colormap = px.colors.diverging.RdBu[::-1][2:-2]
    colormap = extend_colorscale(colormap, 4)
    
    styles = []

    for col in numeric_columns:

        values = df[col]

        if use_values_limits is True:
            mean = np.mean(values)
            min_value = np.min(values)
            max_value = np.max(values)
        else:
            min_value = mean - stdev
            max_value = mean + stdev
            
        limits = np.linspace(min_value, max_value, len(colormap))
            
        for value in values:
            idx = (np.abs(limits - value)).argmin()
            
            styles.append({
                'if': {
                    'filter_query': '{{{col}}} = {value}'.format(
                            col = col, value = value
                        ),
                    'column_id': col
                    },
                    'backgroundColor': colormap[idx]
                })

            styles.append({
                'if': {
                    'filter_query': '{{{col}}} = {value}'.format(col = col, value = value),
                    'state': 'selected',
                    'column_id': col,
                    },
                    'backgroundColor': colormap[idx]
                })

            styles.append({
                'if': {
                    'filter_query': '{{{col}}} = {value}'.format(col = col, value = value),
                    'state': 'selected',
                    'column_id': 'index',
                    },
                    'backgroundColor': 'rgb(228, 228, 255)'
                })

    return styles

def stripped_rows():

    style_data_conditional = []

    style_data_conditional.append({'if': {'row_index': 'odd'},
                                      'backgroundColor': 'rgb(248, 248, 248)'})
    style_data_conditional.append({'if': {'row_index': 'odd', 'state': 'selected'},
                                   'backgroundColor': 'rgb(228, 228, 228)'})
    style_data_conditional.append({'if': {'row_index': 'even'},
                                  'backgroundColor': 'rgb(255, 255, 255)'})
    style_data_conditional.append({'if': {'row_index': 'even', 'state': 'selected'},
                                   'backgroundColor': 'rgb(235, 235, 235)'})
    style_data_conditional.append({'if': {'state': 'selected'},
                                   "border": "3px solid blue"})
    return style_data_conditional

def build_segm_table(mode = 'Method', segmentation_method = 'ROQS', show_stdev = False, color = False, pvalue=False):

    list_scalars = ['FA','RD','AD','MD']

    if mode == 'subjects':
        df = dict_scalar_statistics[segmentation_method]
        df = df.drop(dict_removed_subjects[segmentation_method], errors='ignore')
        df = df.reset_index().rename(columns={"index": "Subject"})
        names = ['Subject'] + list_scalars
    
    else:

        if mode == 'Method':
            df = pd.DataFrame()
            for segmentation_method in dict_segmentation_methods.keys():
                df_aux = dict_scalar_statistics[segmentation_method]
                df_aux = df_aux.drop(dict_removed_subjects[segmentation_method], errors='ignore')
                df_aux = df_aux.reset_index()
                df_aux['Method'] = segmentation_method
                df = pd.concat([df, df_aux], axis=0) 
            category_list = list(dict_segmentation_methods.keys())
        else:
            df = dict_scalar_statistics[segmentation_method]
            df = df.drop(dict_removed_subjects[segmentation_method], errors='ignore')
            df = pd.concat([df_categories[mode], df], axis=1)
            category_list = list(set(df_categories[mode]))
            
        if pvalue is True:
            
            dict_pvalues = {}
            for cat1, cat2 in itertools.combinations(category_list, 2):
                df1 = df[df[mode] == cat1]
                df2 = df[df[mode] == cat2]
                dict_pvalues['{} x {}'.format(cat1, cat2)] = get_column_pvalues(df1, df2)
            
            df = pd.DataFrame().from_dict(dict_pvalues, orient='index', columns=scalar_statistics_names).reset_index()
            df = df.rename(columns = {'index':mode})
            names = [mode] + list_scalars
            
        else:
            df = df.groupby(mode).mean().reset_index()
            names = [mode] + list_scalars

    df = df.round(6).sort_index()

    if show_stdev is False:
        columns=[{"name": i, "id": i} for i in names]
        data=df[names]
    else:
        columns=[{"name": i, "id": i} for i in df.columns[:-1]]
        data=df

    layout = dash_table.DataTable(
        id = 'segm_table',
        columns = columns,
        data = data.to_dict('records'),
        
        page_action = 'none',
        fixed_rows = {'headers': True},
        style_table={
            'maxHeight': '300px', 
            'overflowY': 'auto'},
        style_header = {
            'fontWeight': 'bold',
        },
        style_cell = {
            'font_family': 'Open Sans',
            'font_size': '18px',
            'text_align': 'center'
        },
        style_as_list_view = True,
        export_format='xlsx',
        export_headers='display',
        style_data_conditional = stripped_rows(),
    )
    return layout

def build_parcel_table(mode = 'Method', segmentation_method = 'ROQS', parcellation_method = 'Witelson', scalar = 'FA', color = False, pvalue = False):
    
    list_regions = ['P1', 'P2', 'P3', 'P4', 'P5']

    if mode == 'subjects':
        df = pd.DataFrame()
        for region in list_regions:
            df = pd.concat([df, dict_parcellations_statistics[segmentation_method][parcellation_method][region][scalar]], axis=1)
        df.columns = list_regions
        df = df.drop(dict_removed_subjects[segmentation_method], errors='ignore')
        df = df.reset_index()
        df = df.rename(columns={"index": "Subject"})
        names = ['Subject'] + list_regions

    else:
    
        if mode == 'Method':
            
            df = pd.DataFrame()
            for segmentation_method in dict_segmentation_methods.keys():
                df_aux = pd.DataFrame()
                for region in list_regions:
                    df_aux = pd.concat([df_aux, dict_parcellations_statistics[segmentation_method][parcellation_method][region][scalar]], axis=1)
                df_aux = df_aux.drop(dict_removed_subjects[segmentation_method], errors='ignore')
                df_aux['Method'] = segmentation_method
                df = pd.concat([df, df_aux], axis=0)
            
            category_list = list(set(dict_segmentation_methods.keys()))
            

        else:

            category_list = list(set(df_categories[mode]))

            df = pd.DataFrame()
            for category in category_list:
                df_aux = pd.DataFrame()
                for region in list_regions:
                    df_aux = pd.concat([df_aux, dict_parcellations_statistics[segmentation_method][parcellation_method][region][scalar][df_categories[mode] == category]], axis=1)
                df_aux[mode] = category
                df = pd.concat([df, df_aux], axis=0)
            df = df.drop(dict_removed_subjects[segmentation_method], errors='ignore')
            

        if pvalue is True:
            
            df.columns = list_regions + [mode]

            dict_pvalues = {}
            for cat1, cat2 in itertools.combinations(category_list, 2):
                df1 = df[df[mode] == cat1]
                df2 = df[df[mode] == cat2]
                dict_pvalues['{} x {}'.format(cat1, cat2)] = get_column_pvalues(df1, df2)
            
            df = pd.DataFrame().from_dict(dict_pvalues, orient='index', columns=list_regions).reset_index()
            df = df.rename(columns = {'index':mode})
            names = [mode] + list_regions
            
        else:
            df = df.groupby(mode).mean()
            df.columns = list_regions
            names = [mode] + list_regions     


    layout = dash_table.DataTable(
        id = 'parcel_table',
        columns = [{"name": name, "id": name} for name in names],
        data = df.round(6).reset_index().to_dict('records'),

        page_action = 'none',
        fixed_rows = {'headers': True},
        style_table={
            'maxHeight': '300px', 
            'overflowY': 'auto'},
        style_header = {
            'fontWeight': 'bold',
        },
        style_cell = {
            'font_family': 'Open Sans',
            'font_size': '18px',
            'text_align': 'center'
        },
        style_as_list_view = True,
        export_format='xlsx',
        export_headers='display',
        style_data_conditional=stripped_rows()
        
    )

    return layout

def export_parcel_table(segmentation_method = 'ROQS', parcellation_method = 'Witelson', include_segm=True, include_parc=True, include_cat=True, groupby_cat=None, include_stdev=True):

    # Get data, organize with multiindex columns
    if include_parc is True:
        df = pd.concat(dict_parcellations_statistics[segmentation_method][parcellation_method].values(), 
               axis = 1, 
               keys = dict_parcellations_statistics[segmentation_method][parcellation_method].keys())
    
    # Include info for whole CC
    if include_segm is True:
        df_segm = pd.concat(dict(CC = dict_scalar_statistics[segmentation_method].drop(columns=['Method'])), axis=1)
        
        if include_parc is True:
            df = pd.concat([df_segm, df], axis=1)
        else:
            df = df_segm

    # Remove info from removed subjects
    if groupby_cat is None:
        df.at[dict_removed_subjects[segmentation_method], df.columns] = ''
    else:
        df = df.drop(dict_removed_subjects[segmentation_method])

                
    # Add categories info
    if include_cat is True or groupby_cat is not None:
        df_aux = pd.concat(dict(Categories=df_categories), axis=1)
        df = pd.concat([df_aux, df], axis=1)

    # Group by category
    if groupby_cat is not None:
        df = df.groupby(('Categories', groupby_cat)).mean()    

    # Drop or not Std Dev columns
    if include_stdev is False:
        for col_tuple in df.columns:
            if 'StdDev' in col_tuple[-1]:
                df = df.drop(columns=col_tuple)

    # Define filename
    if include_segm is True and include_parc is False:
        filename = "inCCsight_data_{}.xlsx".format(segmentation_method)
    elif include_segm is False and include_parc is True:
        filename = "inCCsight_data_{}.xlsx".format(parcellation_method)
    elif include_segm is True and include_parc is True:
        filename = "inCCsight_data_{}_{}.xlsx".format(segmentation_method, parcellation_method)
    else:
        filename = "inCCsight_data.xlsx"

    return send_data_frame(df.to_excel, filename)

def build_export_parcel_table_modal():

    options_segm_config = [{'label': item, 'value': item} for item in dict_segmentation_methods.keys()]
    options_parc_config = [{'label': item, 'value': value} for item, value in zip(list(dict_parcellation_methods.keys()), ['Witelson', 'Hofer', 'Chao', 'Cover', 'Freesurfer'])]
    options_data_config = [{'label': item, 'value': item} for item in ['Categories', 'Segmentation', 'Parcellation']]
    options_stdv_config = [{'label': item, 'value': item} for item in ['Show Std. Dev.']]
    
    options_groupby_check = [{'label': item, 'value': item} for item in ['Group by:']]
    options_groupby_config = [{'label': item, 'value': item} for item in df_categories.columns]

    segm_config = dbc.FormGroup(
        row=True,
        children = [
            dbc.Label("Segmentation Method:", width=4, style=dict(justifyContent='flex-end')),
            dbc.Col(
                dcc.Dropdown(
                    id="export_segm_method",
                    options=options_segm_config,
                    value='ROQS'),
                
                width=8)], 
            )

    parc_config = dbc.FormGroup(
        row=True,
        children = [
            dbc.Label("Parcellation Method:", width=4, style=dict(justifyContent='flex-end')),
            dbc.Col(
                dcc.Dropdown(
                    id="export_parc_method",
                    options=options_parc_config,
                    value='Witelson'),
                
                width=8)], 
            )

    data_config = dbc.FormGroup(
        row=True,
        children = [
            dbc.Label("Include data from:", width=4, style=dict(justifyContent='flex-end')),
            dbc.Col(
                dcc.Checklist(
                    id="export_data_config",
                    options=options_data_config,
                    value=['Categories', 'Segmentation', 'Parcellation'],
                    inputStyle=dict(marginRight='5px')),
                
                width=8)], 
            )

    extra_config = dbc.FormGroup(
        row=True,
        children = [
            dbc.Col(
                width=4,
                children = [
                    dcc.Checklist(
                        id="export_show_stdev",
                        options=options_stdv_config,
                        value=['Show Std. Dev.'],
                        inputStyle=dict(marginRight='5px'))
                    ],
                ),

            dbc.Col(
                width=8,
                children = [
                    dcc.Checklist(
                        id="export_groupby_check",
                        options=options_groupby_check,
                        inputStyle=dict(marginRight='5px')),
                    dcc.Dropdown(
                        id="export_groupby_config",
                        options=options_groupby_config,
                        value='Folder'),
                    ],
                )
            ])


    form = dbc.Form([segm_config, parc_config, data_config, extra_config])

    modal = dbc.Modal([
                
                dbc.ModalHeader("Download table data"),
                dbc.ModalBody(form, style=dict(fontSize='14pt')),
                dbc.ModalFooter(
                        children=[
                            dbc.Label(id='disable_download_message', style=dict(marginRight='10px', opacity='0.5')),
                            dbc.Button("Download", id="download_data_btn", size='lg', style=dict(marginRight='10px')),
                            Download(id="parcel_download_all"),
                            dbc.Button("Close", id="close_modal_btn", size='lg')
                            ], 
                        style=dict(justifyContent='flex-end'))
                ], 

                size='lg',
                centered=True,
                id="export_data_modal"),

    return modal

def build_individual_subject_segm_table(subject_id, segmentation_method = 'ROQS', show_stdev=False):

    df = dict_scalar_statistics[segmentation_method].reset_index() 
    df = df[df['index'] == subject_id].rename(columns={'index':'Subject'})
    
    if show_stdev is True:
        names = ['Subject'] + scalar_statistics_names
    else:
        names = ['Subject', 'FA', 'RD', 'AD', 'MD']

    layout = dash_table.DataTable(
        id = 'individual_subject_segm_table',
        columns = [{"name": name, "id": name} for name in names],
        data = df.round(6).reset_index().to_dict('records'),

        page_action = 'none',
        fixed_rows = {'headers': True},
        style_table={
            'maxHeight': '300px', 
            'overflowY': 'auto'},
        style_header = {
            'fontWeight': 'bold',
        },
        style_cell = {
            'font_family': 'Open Sans',
            'font_size': '18px',
            'text_align': 'center'
        },
        style_as_list_view = True,
        export_format='xlsx',
        export_headers='display',
        style_data_conditional=stripped_rows()
        
    )

    return html.Div(layout, style=dict(margin='-30px 0px 2px 0px'), id='individual_subject_segm_table_container')

def build_individual_subject_parcel_table(subject_id, segmentation_method = 'ROQS', parcellation_method='Witelson', scalar='FA'):

    list_regions = ['P1', 'P2', 'P3', 'P4', 'P5']

    df = pd.DataFrame()
    for region in list_regions:
        df = pd.concat([df, dict_parcellations_statistics[segmentation_method][parcellation_method][region][scalar]], axis=1)
    df.columns = list_regions
    
    df = df.reset_index()
    df = df[df['index'] == subject_id].rename(columns={'index':'Subject'})
    

    names = ['Subject'] + list_regions

    layout = dash_table.DataTable(
        id = 'individual_subject_parcel_table',
        columns = [{"name": name, "id": name} for name in names],
        data = df.round(6).reset_index().sort_index().to_dict('records'),

        page_action = 'none',
        fixed_rows = {'headers': True},
        style_table={
            'maxHeight': '300px', 
            'overflowY': 'auto'},
        style_header = {
            'fontWeight': 'bold',
        },
        style_cell = {
            'font_family': 'Open Sans',
            'font_size': '18px',
            'text_align': 'center'
        },
        style_as_list_view = True,
        export_format='xlsx',
        export_headers='display',
        style_data_conditional=stripped_rows()
        
    )

    return html.Div(layout, style=dict(margin='-30px 0px 2px 0px'), id='individual_subject_parcel_table_container')

# Extra functions

def get_number_of_folders():
    if len(list(group_dict.values())) > 0:
        return len(set(np.hstack(list(group_dict.values()))))
    else:
        return 1

def get_column_pvalues(df1, df2):
    df1 = df1.dropna()._get_numeric_data()
    df2 = df2.dropna()._get_numeric_data()
    
    pvalues = []
    if list(df1.columns) == list(df2.columns):    
        for r in df1.columns:
            for c in df2.columns:
                if r == c:
                    pvalues.append(stats.ttest_ind(df1[r], df2[c], nan_policy='omit')[1])
        return pvalues
    else:
        raise ValueError("DataFrame column values don't match")

def empty_figure_with_text(text):

    return {
        "layout": {
            "height": 300,
            "paper_bgcolor": 'rgba(0,0,0,0)',
            "xaxis": {
                "visible": False
            },
            "yaxis": {
                "visible": False
            },
            "annotations": [
                {
                    "text": text,
                    "xref": "paper",
                    "yref": "paper",
                    "showarrow": False,
                    "font": {
                        "size": 28
                    }
                }
            ]
        }
    }

# ---------------------------------- LAYOUT -----------------------------------------------
app.layout = html.Div(
    children=[
        html.Div(
            id="top-row",
            children=[
            # Top row left
                html.Div(
                    className="row",
                    id="top-row-header",
                    children=[
                        html.Div(
                            id="header-container",
                            children=[
                                build_banner(),
                                html.Div(
                                    html.H5(
                                        children=[
                                            html.H6("This is a data exploration and visualization tool for diffusion tensor images of the corpus callosum."),
                                            html.H6("Further information can be found in the "),
                                            html.A(href='https://github.com/thaiscaldeira/inCCsight)', children='GitHub page.', style=dict(marginTop='8px', marginLeft='7px')),
                                            ],
                                        id="instruct",
                                        className='row',
                                    ),
                                    className='twelve columns'),
                                
                                html.Div(className='twelve columns', style=dict(display='flex', justifyContent='center', verticalAlign='center'),
                                    children=[
                                        html.H5("Category:", style=dict(color='white', padding='0 10px 0 54px')),
                                        dcc.Dropdown(id='dropdown_category',
                                            options=[{'label': method, 'value': method} for method in ['Method']+list(set(df_categories.columns))],
                                            multi=False,
                                            value='Method',
                                            style={'width':'150px'})
                                    ],
                                    id="div_category",
                                ),  
                                
                                html.Div(className='twelve columns', style=dict(display='flex', justifyContent='center', verticalAlign='center'),
                                    children=[
                                       html.H5("Segm. Method:", style=dict(color='white', padding='0 10px 0 0')),
                                       dcc.Dropdown(id='dropdown_segm_methods',
                                             options=[{'label': method, 'value': method} for method in dict_segmentation_methods.keys()],
                                             multi=False,
                                             value=list(dict_segmentation_methods.keys())[0],
                                             style={'width':'150px'})
                                    ],
                                    id="div_select_segm_method",
                                ),

                                html.Div(className='twelve columns', id='quality-button-container', 
                                    children=[
                                        dbc.Button(
                                            "Check quality",
                                            size='lg',
                                            color="light",
                                            id='quality-button',
                                            style=dict(marginRight='20px')
                                        ),
                                        
                                        dbc.Button(
                                            'Download tabled data', 
                                            size='lg',
                                            color="light",
                                            id='parcel_download_all_btn',
                                        ),

                                ]),
                            
                            ],
                        )
                    ],
                ),

            # Top row right
                html.Div(
                    className="row",
                    id="top-row-graphs",
                    children=[

                        html.Div([

                            #Subjects list
                            html.Div([
                                build_graph_title("Subjects"),
                                html.Div(
                                    build_subjects_list(), 
                                    style=dict(height='550px', margin='0 0 50px 0', overflowY='auto', minWidth='300px', maxWidth='450px'),
                                    id="subject-table-container")], 
                                className = "column", 
                                id="subject-list-container",
                            ),

                            html.Div(className="four columns"),

                        ], className="one-half column", id="left-panel"),


                        html.Div([

                            html.Div(
                                id='summary-card-container',
                                className='column',
                                children=[
                                dbc.Card(
                                    id = "summary-card",
                                    children = [
                                        html.H1(str(len(set(list(path_dict.keys())))), className='numcard'),
                                        html.H5("Subjects", className='titlecard'),
                                        html.H1(str(get_number_of_folders()), className='numcard', style={'margin-top': '10px'}),
                                        html.H5("Folders", className='titlecard'),
                                    ],
                                ),
                            ]),

                        ], className="one-half column", id="right-panel"),
                    ],
                ),
            ],
        ),
        
        # Dashboard row
        html.Div(
            className='row',
            children=[    
        
                dbc.Collapse(
                    dbc.Card(
                        dbc.CardBody(
                            children=build_quality_collapse(),
                            style=dict(padding='0')
                            ),
                        style=dict(border='1px solid rgba(0,0,0,0)'),
                        ),
                    id="quality-collapse",
                    className='quality columns',
                ),
            
                html.Div(
                    id='dashboard',
                    style=dict(height='100vh', overflowY='auto', overflowX='hidden'),
                    className='twelve columns',
                    children=[

                        dbc.Collapse(
                            dbc.Card(dbc.CardBody()), 
                            id="subject-collapse",
                        ),

                        html.Div(
                            className = 'row',
                            id = 'tables-row',
                            children = [

                                html.Div(
                                    id = 'segm_table_super_container',
                                    className = 'six columns',
                                    children = [
                                        build_graph_title("Segmentation data"),
                                        html.Div(
                                            id = 'segm_table_container',
                                            children = [build_segm_table(show_stdev = False, color = True)]),
                                        html.Div(
                                            id = 'segm_table_options',
                                            className = 'table-options',
                                            children = [

                                                html.H6('Mode:', className='table-options-title'),

                                                dcc.Dropdown(
                                                    id='segm_table_dropdown_mode',
                                                    className = 'options-dropdown',
                                                    options=[{'label': 'Overall', 'value': 'overall'},
                                                             {'label': 'Subjects', 'value': 'subjects'},
                                                             {'label': 'p-values', 'value': 'pvalue'}],
                                                    multi=False,
                                                    value='overall'),

                                                html.H6('Std.Dev.:', className='table-options-title'),

                                                dcc.Dropdown(
                                                    id = 'segm_table_dropdown_stdev',
                                                    className = 'options-dropdown',
                                                    options = [
                                                        {'label': 'Show', 'value': True},
                                                        {'label': 'Hide', 'value': False},
                                                    ],
                                                    value = False,
                                                ),

                                            ],
                                        ),
                                    ],
                                ),
                                
                                html.Div(
                                    id = 'parcel_table_super_container',
                                    className = 'six columns',
                                    children = [
                                        build_graph_title("Parcellation data"),
                                        
                                        html.Div(
                                            id = 'parcel_table_container',
                                            children = [build_parcel_table(
                                                            segmentation_method = 'ROQS', 
                                                            parcellation_method = 'Witelson', 
                                                            scalar = 'FA')]),
                                        
                                        html.Div(
                                            id = 'parcel_table_options',
                                            className = 'table-options',
                                            children = [
                                                
                                                html.H6('Mode:', className='table-options-title'),

                                                dcc.Dropdown(
                                                    id='parcel_table_dropdown_mode',
                                                    className = 'options-dropdown',
                                                    options=[{'label': 'Overall', 'value': 'overall'},
                                                             {'label': 'Subjects', 'value': 'subjects'},
                                                             {'label': 'p-values', 'value': 'pvalue'}],
                                                    multi=False,
                                                    value='overall',
                                                ),
                                                
                                                html.H6('Parcel. method:', className='table-options-title'),

                                                dcc.Dropdown(
                                                    id = 'parcel_table_dropdown_method',
                                                    className = 'options-dropdown',
                                                    options = [
                                                        {'label': 'Witelson', 'value': 'Witelson'},
                                                        {'label': 'Hofer & Frahm', 'value': 'Hofer'},
                                                        {'label': 'Chao et al.', 'value': 'Chao'},
                                                        {'label': 'Cover et al.', 'value': 'Cover'},
                                                        {'label': 'Freesurfer', 'value': 'Freesurfer'},
                                                    ],
                                                    value = 'Witelson',
                                                ),

                                                html.H6('Scalar:', className='table-options-title'),

                                                dcc.Dropdown(
                                                    id = 'parcel_table_dropdown_scalar',
                                                    className = 'options-dropdown',
                                                    options = [
                                                        {'label': 'FA', 'value': 'FA'},
                                                        {'label': 'MD', 'value': 'MD'},
                                                        {'label': 'RD', 'value': 'RD'},
                                                        {'label': 'AD', 'value': 'AD'},
                                                    ],
                                                    value = 'FA'
                                                ),                

                                            ],
                                        ),
                                    ],
                                )
                                

                            ]
                        ),

                        html.Div(
                            className="row",
                            id="bottom-row",
                            style=dict(marginRight='20px'),
                            children=[

                                html.Div(
                                    id="bottom-row-left-column",
                                    className="seven columns",
                                    children=[
                                    
                                        html.Div(
                                            id="boxplot-container1",
                                            style=dict(marginBottom='-30px'),
                                            children=[
                                                build_graph_title("Segmentation BoxPlots"),
                                                dcc.Loading(dcc.Graph(id="segm_boxplots", figure=build_group_segm_boxplot())),
                                            ],
                                        ),

                                        html.Div(
                                            id="boxplot-container2",
                                            #style=dict(maginTop='0px'),
                                            children=[
                                                build_graph_title("Parcellation BoxPlots"),
                                                dcc.Loading(dcc.Graph(id="parc_boxplots", figure=build_parcel_boxplot())),
                                                build_parcel_boxplot_dropdowns(),
                                            ],
                                        ),
                                    ],
                                ),
                                
                                # Scatterplot
                                html.Div(
                                    id="scatterplot-container",
                                    className="five columns",
                                    children=[
                                        build_graph_title("Scatter Plot"),
                                        dcc.Loading(dcc.Graph(id="scatter_plot", figure=build_segm_scatterplot())),
                                        build_segm_scatterplot_dropdowns(),
                                    ],
                                ),
                            ],
                        ),

                        html.Div(
                            className="row",
                            id="second-bottom-row",
                            style=dict(marginRight='20px'),
                            children=[
                            
                                # Scattermatrix
                                html.Div(
                                    id="scattermatrix-container",
                                    className="seven columns",
                                    children=[
                                        build_graph_title("Scatter Matrix"),
                                        dcc.Loading(dcc.Graph(id="scatter_matrix", figure=build_segm_scattermatrix())),
                                    ],
                                ),

                                html.Div(
                                    className="five columns",
                                    children = [

                                        # Midline plots
                                        html.Div(
                                            id="midline-container",
                                            children=[
                                                build_graph_title("Midline Plots"),
                                                dcc.Loading(dcc.Graph(id="midline_graph", figure=build_midline_plot())),
                                                build_midlineplot_dropdown(),
                                            ],
                                        ),

                                        # Bubble plot
                                        html.Div(
                                            id="bubbleplot-container",
                                            children=[
                                                build_graph_title("Bubble Plots"),
                                                dcc.Loading(dcc.Graph(id="bubble_plots", figure=build_bubble_grouped())),
                                                build_bubbleplot_dropdowns(),
                                            ],
                                        ),

                                    ],
                                ),
                            ],
                        ),

                        html.Div(
                            className="row",
                            id="third-bottom-row",
                            style=dict(marginRight='20px'),
                            children=[


                                html.Div(build_export_parcel_table_modal(), id='example_div'),
                                html.Div(id='bridge-div'),
                            ],
                        ),        
                    ]
                )
            ]
        )
    ],
)

# --------------------------------- CALLBACKS ---------------------------------------------

# Banner --------------------------------------------------------------

# Enable/Disable segm method dropdown
@app.callback(
    Output("dropdown_segm_methods","disabled"),
    [Input("dropdown_category","value")])
def update_dropdown_disabled(mode):
    if mode == 'Method':
        return True
    else:
        return False

# Tables --------------------------------------------------------------

# Change segm table mode
@app.callback(
    Output("segm_table_container", "children"),
    [Input("segm_table_dropdown_mode", "value"),
     Input("dropdown_segm_methods", "value"),
     Input("dropdown_category","value"),
     Input("segm_table_dropdown_stdev", "value"),
     Input('photo-container', 'children')])
def change_segm_table_mode(table_mode, segmentation_method, mode, show_stdev, removed):
    pvalue = False
    if table_mode == 'subjects':
        mode = 'subjects'
    elif table_mode == 'pvalue':
        pvalue = True
    return [build_segm_table(mode = mode, segmentation_method = segmentation_method, show_stdev = show_stdev, color = False, pvalue=pvalue)]

# Change parcel table mode
@app.callback(
    Output("parcel_table_container", "children"),
    [Input("dropdown_category","value"),
     Input("dropdown_segm_methods", "value"),
     Input("parcel_table_dropdown_mode", "value"),
     Input("parcel_table_dropdown_method", "value"),
     Input("parcel_table_dropdown_scalar", "value"),
     Input('photo-container', 'children')])
def change_parcel_table_mode(mode, segmentation_method, table_mode, parcellation_method, scalar, removed):
    
    pvalue = False
    if table_mode == 'subjects':
        mode = 'subjects'
    elif table_mode == 'pvalue':
        pvalue = True

    return [build_parcel_table(mode = mode, 
                segmentation_method = segmentation_method, 
                parcellation_method = parcellation_method,
                scalar = scalar,
                color = False,
                pvalue = pvalue)]

# Highlight table row borders upon clicking
@app.callback(
    Output("segm_table", "style_data_conditional"),
    [Input("segm_table", "selected_cells")],
    [State("segm_table", "data"),
     State('segm_table', 'style_data_conditional')])
def paint_segm_table(selected_cells, table_data, style_data_conditional):
 
    n_rows = len(table_data)

    for row in range(n_rows):
        rule = {"if": {"row_index": row}, "border": "3px solid blue"}
        if rule in style_data_conditional:
            style_data_conditional.remove(rule)

    if selected_cells is not None:

        for cell in selected_cells:
            row = cell['row'] 
            style_data_conditional.append({"if": {"row_index": row},
                                           "border": "3px solid blue"})
    
    return style_data_conditional

# Highlight table row borders upon clicking
@app.callback(
    Output("parcel_table", "style_data_conditional"),
    [Input("parcel_table", "selected_cells")],
    [State("parcel_table", "data"),
     State('parcel_table', 'style_data_conditional')])
def paint_parcel_table(selected_cells, table_data, style_data_conditional):
    n_rows = len(table_data)

    for row in range(n_rows):
        rule = {"if": {"row_index": row}, "border": "3px solid blue"}
        if rule in style_data_conditional:
            style_data_conditional.remove(rule)

    if selected_cells is not None:

        for cell in selected_cells:
            row = cell['row'] 
            style_data_conditional.append({"if": {"row_index": row},
                                           "border": "3px solid blue"})
    
    return style_data_conditional

# Open export data modal
@app.callback(
    Output("export_data_modal", "is_open"),
    [Input("parcel_download_all_btn", "n_clicks"),
     Input("close_modal_btn", "n_clicks")])
def toggle_export_modal(open_clicks, close_clicks):
    if open_clicks is not None:
        
        trigger = dash.callback_context.triggered[0]
        if trigger['prop_id'] == 'parcel_download_all_btn.n_clicks':
            return True
        else:
            return False

# Disable download button
@app.callback(
    [Output("download_data_btn", "disabled"),
     Output("disable_download_message", "children")],
    [Input("export_data_config", "value")])
def toggle_download_btn(value):
    if value == []:
        return True, "You must include at least one type of data."
    else:
        return False, ""

@app.callback(
    Output("parcel_download_all", "data"), 
    [Input("download_data_btn", "n_clicks")],
    [State("export_segm_method", "value"),
     State("export_parc_method", "value"),
     State("export_data_config", "value"),
     State("export_groupby_check", "value"),
     State("export_groupby_config", "value"),
     State("export_show_stdev", "value"),
     ])
def download_parceldata(n_clicks, segmentation_method, parcellation_method, data_config, groupby_check, groupby_config, stdev_config):
    if n_clicks is not None:

        include_cat = False
        include_segm = False
        include_parc = False
        include_stdev = False
        groupby_cat = None

        if data_config is not None:
            if 'Categories' in data_config:
                include_cat = True
            if 'Segmentation' in data_config:
                include_segm = True
            if 'Parcellation' in data_config:
                include_parc = True

        if groupby_check == ['Group by:']:
            groupby_cat = groupby_config

        if stdev_config == ['Show Std. Dev.']:
            include_stdev = True

        return export_parcel_table(segmentation_method, parcellation_method, include_segm, include_parc, include_cat, groupby_cat, include_stdev)


# Graph updates -------------------------------------------------------

# Update segm box-plots
@app.callback(
    Output("segm_boxplots", "figure"),
    [Input("dropdown_segm_methods","value"),
     Input("dropdown_category","value"),
     Input('photo-container', 'children')])
def update_segm_boxplots(segm_method, mode, removed):
    return build_group_segm_boxplot(mode=mode, segmentation_method=segm_method)

# Update segm box-plots
@app.callback(
    Output("parc_boxplots", "figure"),
    [Input("dropdown_segm_methods","value"),
     Input("dropdown_category","value"),
     Input("dropdown-parcel-boxplot-left","value"),
     Input("dropdown-parcel-scalars-right","value"),
     Input('photo-container', 'children')])
def update_segm_boxplots(segm_method, mode, parc_method, scalar, removed):
    return build_parcel_boxplot(scalar=scalar, mode=mode, segmentation_method=segm_method, parcellation_method=parc_method)

# Update midline plot
@app.callback(
    Output("midline_graph", "figure"),
    [Input("dropdown_segm_methods","value"),
     Input("dropdown_category","value"),
     Input("dropdown-midline-scalars","value"),
     Input('photo-container', 'children')])
def update_midlineplot(segm_method, mode, scalar, removed):
    return build_midline_plot(mode=mode, segmentation_method=segm_method, scalar=scalar)

# Update scatter matrix
@app.callback(
    Output("scatter_matrix", "figure"),
    [Input("dropdown_segm_methods", "value"),
     Input("dropdown_category","value"),
     Input('photo-container', 'children')])
def update_scattermatrix(segm_method, mode, removed):
    return build_segm_scattermatrix(mode=mode, segmentation_method=segm_method)

# Update scatter plot
@app.callback(
    Output("scatter_plot","figure"),
    [Input("dropdown_segm_methods","value"),
     Input("dropdown_category","value"),
     Input("dropdown-scalars-right","value"),
     Input("dropdown-scalars-left","value"),
     Input("dropdown-scalars-trendline","value"),
     Input('photo-container', 'children')])
def update_scatterplot(segm_method, mode, scalar_x, scalar_y, trendline, removed):
    if trendline == 'None':
        trendline = None
    else:
        trendline = trendline.lower()
    return build_segm_scatterplot(mode=mode, segmentation_method=segm_method, scalar_x=scalar_x, scalar_y=scalar_y, trendline=trendline)

# Update bubble plot
@app.callback(
    Output("bubble_plots","figure"),
    [Input("dropdown_segm_methods","value"),
     Input("dropdown_category","value"),
     Input("dropdown-bubbleplot-mode", "value"),
     Input("dropdown-bubbleplot-left", "value"),
     Input("dropdown-bubbleplot-right", "value"),
     Input("dropdown-bubbleplot-threshold", "value"),
     Input('photo-container', 'children')])
def update_bubbleplot(segm_method, mode, bubble_mode, scalar, size, threshold, removed):
    if size == 'True':
        size = True
    elif size == 'False':
        size = False

    if bubble_mode == 'Scalar':
        return build_bubble_grouped(mode=mode, segmentation_method=segm_method, scalar=scalar, size=size)
    else:
        return build_bubble_grouped_pvalue(mode=mode, segmentation_method=segm_method, scalar=scalar, threshold=threshold)

# Update bubble plot dropdown options
@app.callback(
    [Output('div-bubble-dropdown-right', 'style'),
     Output('div-bubble-dropdown-threshold', 'style')],
    [Input('dropdown-bubbleplot-mode', 'value')],
    [State('div-bubble-dropdown-right', 'style'),
     State('div-bubble-dropdown-threshold', 'style')])
def update_bubbleplot_dropdown(bubble_mode, style_right, style_threshold):
    if bubble_mode == 'Scalar':
        style_right['display'] = 'flex'
        style_threshold['display'] = 'none'
    else:
        style_right['display'] = 'none'
        style_threshold['display'] = 'flex'
    return [style_right, style_threshold]


# Subject collapse ----------------------------------------------------

# Open subject collapse
@app.callback(
    [Output("subject-collapse", "is_open"),
     Output("subject-collapse", "children")],
    [Input({'type': 'subject-list-btns', 'index': ALL}, 'n_clicks'),
     Input({'type': 'btn-exit-subject', 'index': ALL}, 'n_clicks')],
    [State({'type': 'subject-list-btns', 'index': ALL}, 'id')])
def open_subject_collapse(n_clicks, exit_clicks, ids):
    if n_clicks is not None:

        trigger = dash.callback_context.triggered[0] 

        if (trigger['value'] is None): 
            return dash.no_update

        btn_id = ast.literal_eval(trigger["prop_id"][:-9])

        if btn_id['type'] == 'btn-exit-subject':
            return False, []        

        subject_id = btn_id['index']
        
        global selected_subject_id
        selected_subject_id = subject_id
        
        return True, [build_subject_collapse(subject_id = subject_id)]
    else:
        return dash.no_update

@app.callback(
    Output('individual_subject_segm_table_container', 'children'),
    [Input('dropdown-subj-collapse-table-segm-methods', 'value'),
     Input('dropdown-subj-collapse-table-segm-std-dev', 'value')])
def update_segm_table_subj_collapse(segmentation_method, show_stdev):
    if show_stdev == 'Show':
        show_stdev = True
    else:
        show_stdev = False

    return build_individual_subject_segm_table(selected_subject_id, segmentation_method, show_stdev)

@app.callback(
    Output('individual_subject_parcel_table_container', 'children'),
    [Input('dropdown-subj-collapse-table-parcel-scalars', 'value'),
     Input('dropdown-subj-collapse-table-parcel-segm-methods', 'value'),
     Input('dropdown-subj-collapse-table-parcel-methods', 'value')])
def update_parc_table_subj_collapse(scalar, segmentation_method, parcellation_method):
    return build_individual_subject_parcel_table(selected_subject_id, segmentation_method, parcellation_method, scalar)

@app.callback(
    Output('subject_collapse_fissure_img', 'figure'),
    [Input('dropdown-subj-collapse-segm-methods', 'value'),
     Input('dropdown-subj-collapse-scalars', 'value')])
def update_fissure_image(segmentation_method, scalar):
    return build_fissure_image(subject_id=selected_subject_id, segmentation_method=segmentation_method, scalar=scalar)

# Update remove subject tooltip
@app.callback(
    Output('subject-collapse-tooltip', 'children'),
    [Input('dropdown-subj-collapse-segm-methods', 'value')])
def update_remove_subject_tooltip(segmentation_method):
    return "This will remove only the {} segmentation, indicated on the dropdown to the left".format(segmentation_method)

# Quality collapse ----------------------------------------------------

# Bridge div
@app.callback(
    Output('bridge-div', 'className'),
    [Input({'type':'btn-remove-subject', 'index': ALL}, 'n_clicks')],
    [State('dropdown-subj-collapse-segm-methods', 'value')])
def quality_control_bridge(remove_clicks, dropdown_segmentation_method):
    
    global dict_removed_subjects

    trigger = dash.callback_context.triggered[0]
    btn_id = ast.literal_eval(trigger["prop_id"][:-9])
    
    if remove_clicks != [None]:
        dict_removed_subjects[dropdown_segmentation_method].append(btn_id['index'])
        return 'trash_'+btn_id['index']
    else:
        return dash.no_update

# Remove selected subjects
@app.callback(
    Output('photo-container', 'children'),
    [Input('remove_btn', 'n_clicks'),
     Input('restore_btn', 'n_clicks'),
     Input('dropdown-quality-threshold', 'value'),
     Input('dropdown-quality-scalar', 'value'),
     Input('bridge-div', 'className')],
    [State({'type': 'remove-cbx', 'index': ALL}, 'id'),
     State({'type': 'remove-cbx', 'index': ALL}, 'value')])
def remove_quality_images(n_clicks, restore_clicks, threshold, scalar, bridge_flag, ids, values):

    global dict_removed_subjects

    trigger = dash.callback_context.triggered[0]
    removed_counter = 0

    if restore_clicks is not None and trigger['prop_id'] == 'restore_btn.n_clicks':

        for segmentation_method in dict_segmentation_methods.keys():
            removed_counter += len(dict_removed_subjects[segmentation_method])
            dict_removed_subjects[segmentation_method] = []

    elif n_clicks is not None and trigger['prop_id'] == 'remove_btn.n_clicks':

        for dict_id, value in zip(ids,values):
            if value == ['Remove']:
                segmentation_method, subject_key = dict_id['index'].rsplit('-')[-2:]
                dict_removed_subjects[segmentation_method].append(subject_key)
                dict_removed_subjects[segmentation_method] = list(set(dict_removed_subjects[segmentation_method]))
                removed_counter += 1
    
    elif trigger['prop_id'] == 'dropdown-quality-scalar.value' or trigger['prop_id']=='dropdown-quality-threshold.value':
        return build_quality_images(threshold, scalar)

    elif trigger['prop_id'] == 'bridge-div.className':
        return build_quality_images(threshold, scalar)        

    if removed_counter > 0:
        return build_quality_images(threshold, scalar)
    else:
        return dash.no_update

# (Un)Select all subjects (check all remove checkboxes)
@app.callback(
    Output({'type': 'remove-cbx', 'index': ALL}, 'value'),
    [Input('select_all_btn', 'n_clicks'),
     Input('unselect_all_btn', 'n_clicks')],
    [State({'type': 'remove-cbx', 'index': ALL}, 'value')])
def select_all_subjects(select_clicks, unselect_clicks, list_bts):
    
    trigger = dash.callback_context.triggered[0]

    if select_clicks is not None and trigger['prop_id'] == 'select_all_btn.n_clicks':
        return [['Remove']] * len(list_bts)

    elif unselect_clicks is not None and trigger['prop_id'] == 'unselect_all_btn.n_clicks':
        return [[]] * len(list_bts)

    else:
        return dash.no_update

# Open quality collapse
@app.callback(
    [Output("dashboard", "className"),
     Output("quality-collapse","is_open")],
    [Input("quality-button","n_clicks"),
     Input("btn-exit-quality","n_clicks")],
    [State("dashboard", "className")])
def open_quality_collapse(n_clicks, exit_clicks, className):
    if n_clicks is not None:
        trigger = dash.callback_context.triggered[0]
        if trigger["prop_id"].split(".")[0][-18:-2] == 'btn-exit-quality':
            return ["twelve columns", False]
        else:
            if className == 'notquality columns':
                return ["twelve columns", False]
            elif className == 'twelve columns':
                return ["notquality columns", True]
    else:
        return ["twelve columns", False]

'''
@app.callback(
    [Output('scatter_plot_2', 'figure'),
     Output('scatter_matrix_2', 'figure')],
    [Input('scatter_plot', 'selectedData'),
     Input('scatter_matrix', 'selectedData')],
    [State('scatter_plot', 'figure'),
     State('scatter_matrix', 'figure')])
def get_selected_data(selection1, selection2, fig1, fig2):
    
    if fig1 is None:
        raise PreventUpdate
    else:

        selectedpoints = []
        for selected_data in [selection1, selection2]:
            if selected_data and selected_data['points']:
                selectedpoints = selectedpoints + [p['customdata'] for p in selected_data['points']]
        
        if fig1 is not None and fig2 is not None:
            fig1 = go.Figure(fig1)
            fig1.update_traces(selectedpoints=selectedpoints, customdata=list(path_dict.keys()))
            
            fig2 = go.Figure(fig2)
            fig2.update_traces(selectedpoints=selectedpoints, customdata=list(path_dict.keys()))
        
        return fig1, fig2
'''
'''
@app.callback(
    Output('scatter_matrix', 'figure'),
    [Input('scatter_matrix', 'selectedData'),
     Input('scatter_plot', 'selectedData')],
    [State("dropdown-scalars-right","value"),
     State("dropdown-scalars-left","value")])
def update_selected_points(selection1, selection2, scalar_x, scalar_y):
    selected_points = []
    for selected_data in [selection1, selection2]:
        if selected_data is not None:
            for point_dict in selected_data['points']:
                print(point_dict)
                selected_points.append(point_dict['pointIndex'])
    if selected_points == []:
        selected_points = None
    return build_segm_scattermatrix(selected_points=selected_points)
'''


# SERVER CONFIG ---------------------------------------------------------------------------------

port = opts.port

def open_browser():
    url = "http://localhost:{0}".format(port)
    webbrowser.open(url)

if __name__ == "__main__":
    Timer(1.25, open_browser).start()
    app.run_server(debug=False, port=port, host='0.0.0.0')