3D Patch-based Student-Teacher Pyramid Matching in 3D Magnetic Resonance Imaging

PyTorch lightning implementation of A Patch-based Student-Teacher Pyramid Matching Approach to Anomaly Detection in 3D Magnetic Resonance Imaging (MIDL 2024).

Getting Started

Install packages with:

$ pip install -r requirements.txt

Dataset

The data can be accessed at the following link.

• IXI
• BRATS

The path to the data is ./data/SKULL/{BRATS, IXI}.

Train the teacher

Start training the teacher with the following line:

python run_stpm3D --run_mode pretrain

Train the student

Start training the student with the following line:

python run_stpm3D --run_mode train

Evaluation

The evaluation of the network can be started with the following line:

python run_stpm3D --run_mode eval

Code References

• Student-Teacher Feature Pyramid Matching for Anomaly Detection
• Reinventing 2D Convolutions for 3D Images

Citation

• ...