Mix-and-Match Multiple Instance Learning (MM-MIL)
This is the code repository for our paper: Weakly supervised identification of microscopic human breast cancer-related optical signatures from normal-appearing breast tissue.
Overview
We propose MM-MIL for the discovery of novel optical signatures when only coarse-grained and ambiguous annotations are available. We applied the proposed method to the investigation of human breast cancer-related optical signatures based on Simultaneous Label-free Autofluorescence Multiharmonic (SLAM) microscopy and unveiled non-obvious cancer-related optical signatures in peri-tumoral regions.
System Environment
- Ubuntu 18.04
- Python
- Pytorch
- Nvidia GPU + CUDA
- jsondiff
- tdpm
- tifffile
- Captum
Usage
(Will be updated)
Main Results
(Will be updated)
Citation
If you use this code and relecant data, please cite the corresponding paper where the original methods appeared. (Will be updated)