Here is a link to our notebook, and here is a link to our presentation. An @illinois.edu email
is required to access the notebook along with the Google Drive data. The notebook and additional code can also be found in the notebook folder in
this repo.
Code for the paper Detecting Heart Disease from Multi-View Ultrasound Images via Supervised Attention Multiple Instance Learning (MLHC'23)
The AS diagnosis task in this study uses the view_and_diagnosis_labeled_set from TMED2
- Step1. Download TMED2, please visit https://TMED.cs.tufts.edu and follow the instruction.
- Step2. Put the labeled and unlabeled image for the same study under same folder
Follow the instructions here: https://conda.io/projects/conda/en/latest/user-guide/install/index.html
pytorch 1.11.0
export ROOT_DIR="paths to this repo"
(e.g., '/ab/cd/SAMIL')
export DATA_INFO_DIR="path to the data_info/ folder"
(e.g., '/ab/cd/SAMIL/data_info')
export DATA_DIR="path to the actual data
(e.g., '/ab/cd/SAMIL/Echo_data/view_and_diagnosis_labeled_set/')
export CHECKPOINT_DIR="path to where pretrained view classifier and MOCO pretrained weights
(e.g., '/ab/cd/SAMIL/model_checkpoints/')
We releases the pretrained view classifiers and MOCO pretrained checkpoints here
Running SAMIL with study-level pretraining, go to runs/SAMIL
bash launch_experiment.sh run_here
While the focus of our paper is reproducibility, ultimately exact comparison to the results in our paper will be conflated by subtle differences such as the version of Pytorch etc (see https://pytorch.org/docs/stable/notes/randomness.html for more detail).
@article{huang2023detecting, title={Detecting Heart Disease from Multi-View Ultrasound Images via Supervised Attention Multiple Instance Learning}, author={Huang, Zhe and Wessler, Benjamin S and Hughes, Michael C}, journal={arXiv preprint arXiv:2306.00003}, year={2023} }