MedFuseNet

Official Pytorch implementation of MedFuseNet: Fusing Local and Global Deep Feature Representations with Hybrid Attention Mechanisms for Medical Image Segmentation

Architecture:

Usage:

Recommended environment:

Python 3.8

Pytorch 1.11.0

torchvision 0.12.0

Data preparation:

Synapse Multi-organ dataset: Sign up in the official Synapse website (https://www.synapse.org/#!Synapse:syn3193805/wiki/89480) and download the dataset.

ACDC dataset: Download the preprocessed ACDC dataset from Google Drive (https://drive.google.com/file/d/13qYHNIWTIBzwyFgScORL2RFd002vrPF2/view).

Training:

For ACDC training run python train.py --dataset ACDC
For Synapse Multi-organ training run python train.py --dataset Synapse

Test:

For ACDC test run python test.py --dataset ACDC
For Synapse Multi-organ test run python test.py --dataset Synapse