Classification of Breast Cancer Histopathology Images using a Modified Supervised Contrastive Learning Method

Implemented a modified supervised contrastive learning method based on the labels and similarity between images. Also, apply histopathology augmentations, which reached state-of-the-art results in detecting malignant tumors on BreahHis dataset.

Requirement

This repository code is compaitible with Python 3.6 and 3.8, Pytorch 1.2.0, and Torchvision 0.4.0.

Commands

Modified-Supervised-Contrastive Representation Learning Phase1 (Assuming in directory 'src')

python -m self_supervised_phase1/run.py --data_fold <'train_data_fold_path'> --LR <learning_rate - 0.00001> --epoch <150> --description <'experiment_name'>

Calculate Similarity Phase (Assuming in directory 'src')

python -m similarity/run.py --data_fold <'train_data_fold_path'> --LR <learning_rate - 0.00001> --epoch <150> --description <'experiment_name'>

Modified-Supervised-Contrastive Representation Learning Phase2 (Assuming in directory 'src')

python -m self_supervised_phase2/run.py --data_fold <'train_data_fold_path'> --LR <learning_rate - 0.00001> --epoch <150> --description <'experiment_name'>

Fintuning using pretrained Efficient-net b2 on BreakHis (Assuming in directory 'src')

python -m supervised.experiments.breakhis --train_data_fold <'train_data_fold_path'> --test_data_fold <'test_data_fold_path'> --magnification <'40x'/'100x'/'200x'/'400x'> --model_path <'pretrained model path'> --LR <learning_rate - 0.00002> --epoch <150> --description <'experiment_name'>

Evaluation

python - m supervised.evaluation.evaluation --dataset_name "Breakhis"

If we want to test the model of Bach dataset, first finetune the previous pretrained weights, on BACH dataset in the following command, and the evalute the new model.

Fintuning using pretrained Efficient-net b2 on BACH (Assuming in directory 'src')

python -m supervised.experiments.bach --train_data_fold <'train_data_fold_path'> --test_data_fold <'test_data_fold_path'> --magnification <'40x'/'100x'/'200x'/'400x'> --model_path <'pretrained model path'> --LR <learning_rate - 0.00002> --epoch <150> --description <'experiment_name'>

Evaluation

python - m supervised.evaluation.evaluation --dataset_name "BACH"