【EyeMoSt & EyeMoSt+】

This repository provides the code for our paper 【MICCAI 2023 Early Accept】"Reliable Multimodality Eye Disease Screening via Mixture of Student's t Distributions" and 【Medical Image Analysis submission 2024】"Confidence-aware multi-modality learning for eye disease screening"
Current official implementation of EyeMoSt
All codes are released in the version of EyeMoSt+.

Requirment

Pytorch 1.3.0
Python 3
sklearn
numpy
scipy
...

Datasets

Code Usage

1. Prepare dataset

Download the datasets and change the dataset path:
OLIVES dataset path
GAMMA dataset basepath and datapath

2. Pretrained models

Download pretrained models and put them in ./pretrain/

2.1 CNN-based

Fundus (2D): Res2Net
OCT (3D): Med3d

2.2 Transformer-based

Fundus (2D): Swin-Transformer
OCT (3D): UNETR

3. Train

3.1 Train Baseline

Run the script main_train2.shmain_train2.sh python baseline_train3_trans.py to train the baselines (change model_name& mode), models will be saved in folder results

3.2 Train Our Model

Run the script main_train2.sh main_train2.sh python train3_trans.py to train our model (change model_name), models will be saved in folder results

4. Test

4.1 Test Baseline

Run the script main_train2.sh main_train2.sh python baseline_train3_trans.py to test our model (change model_name& mode)

4.2 Test Our Model

Run the script main_train2.sh main_train2.sh python train3_trans.py to test our model (change model_name& mode)

Citation

If you find EyeMoSt helps your research, please cite our paper:

@InProceedings{EyeMoSt_Zou_2023,
author="Zou, Ke
and Lin, Tian
and Yuan, Xuedong
and Chen, Haoyu
and Shen, Xiaojing
and Wang, Meng
and Fu, Huazhu",
title="Reliable Multimodality Eye Disease Screening via Mixture of Student's t Distributions",
booktitle="Medical Image Computing and Computer Assisted Intervention -- MICCAI 2023",
year="2023",
publisher="Springer Nature Switzerland",
address="Cham",
pages="596--606",
}