/VAP_Former

[MICCAI-2023]Visual-Attribute Prompt Learning for Progressive Mild Cognitive Impairment Prediction

Primary LanguagePython

VAP-Former

This is the official implementation of Visual-Attribute Prompt Learning for Progressive Mild Cognitive Impairment Prediction at MICCAI-2023

Table of Contents

Requirements

conda create -n VAPFormer python=3.7
conda activate VAPFormer
pip install -r requirements.txt

Dataset Preparation

The data can be accessed at http://adni.loni.usc.edu/data-samples/access-data/ You can use SPM12,CAT12 to preprocess data

Pre-trained Model

Task Link
AD v.s. NC OneDrive
sMCI v.s. pMCI OneDrive

How to train

Modify the code in train_modelclinical.py

with the following two lines of code

ld_helper = LoaderHelper(task=Task.NC_v_AD) # which defineds the task

model_uuid = train_camull(ld_helper, epochs=50) # function to train the model

The weights are stored in the weights file

How to transfer

Modify the code in trainclinical.py as follows:

ld_helper = LoaderHelper(task=Task.sMCI_v_pMCI)

model = load_model() # Modify the pretrained weight

model_uuid = train_camull(ld_helper, model=model, epochs=50)

evaluate_model(DEVICE, model_uuid, ld_helper)

Acknowledgement

This project is built upon the foundations of several open-source codebases, including camull-net and VPT.