An Explainable Lesion-guided Multi-task Learning For Alzheimer;s Disease Diagnosis

Flowchart of our DeepTAAD

Classification result of our method

The roc curve of our model are shown as follow:

Requirements

• python 3.7
• pytorch 1.6.0
• torchvision 0.7.0
• pickle
• nibabel
• setproctitle
• medcam
• medpy

Training

Run the training script on the ADNI dataset. Distributed training is available for training the proposed DeepTAAD, where --nproc_per_node decides the numer of gpus and --master_port implys the port number. you can use the following command to train the model with task assistance loss CUDA_VISIBLE_DEVICES=0 python3 -m torch.distributed.launch --nproc_per_node=1 --master_port 20003 train.py

Testing

If you want to test the model which has been trained on the ADNI dataset, run the testing script as following. python test_ADNI.py