ASD-DiagNet

This repository contains the implementation of ASD-DiagNet algorithm.

Research article

Taban Eslami, Fahad Saeed, Vahid Mirjalili, Alvis Fong and Angela Laird (2019) ASD-DiagNet: A hybrid learning approach for detection of Autism Spectrum Disorder using fMRI data, Frontiers in Neuroinformatics, 13 (2019): 70.

Enviroment Setup

Hardware requirements

• A server containing CUDA enabled GPU with compute capability 3.5 or above.

Software requirements

• Python version 3.5 or above
• Pytorch version 0.4.1
• CUDA version 8 or above
• Jupyter notebook

Parameter setting

Please provide the parameters in the first cell of Jupyter notebook as follows:

• Atlas name: ("cc200", "aal", or "dosenbach160")

  e.g. p_ROI = "cc200"

• Number of k for k-fold cross-validation:

  e.g. p_fold = 10

• Classification mode: ("whole" or "percenter")

  e.g. p_mode = "percenter"

• Name of the center: (in case of performing per-center classification)

  e.g. p_center = "Stanford"

• Classification method: ("ASD-DiagNet", "rf" or "SVM))

  e.g. p_Method = "ASD-DiagNet"

• Utilizing augmentation technique: (in case of using ASD-DiagNet, True or False)

  e.g. p_augmentation = False

In case of any questions please contact: taban.eslami@wmich.edu