Diagnosis ADHD with Convolutional-LSTM Model

Motivation

• Diagnosing mental disorders is a considerably complex task for behavioral health professionals
• Many factors complicate the process:
  • People exhibit individual behaviors with the symptoms of their disorder(s)
  • No objective biological markers associated with mental disorders
  • Mental disorders often overlap with one to many others
  • Similarity of symptoms among different diseases can lead to inaccurate diagnosis 
• Potential Solution lies with functional Magnetic Resonance Imaging (fMRI) technology

functional Magnetic Resonance Imaging

• Measures brain activity by detecting changes associated with blood flow
  • Known as blood-oxygen-level dependent (BOLD) method
• This technique relies on the fact that blood flow and neuronal activation are coupled.
• When an area of the brain is in use, blood flow to that region also increases
• Does so to provide energy to the neurons, which do not have internal reserves of energy.
• fMRI machine captures blood flow and “lights up” brain areas in images
• Indicates that part of the brain is responsible for handling a certain activity

Objective

• Construct a hybrid model that captures and analyzes both the spatial and temporal aspects of an fMRI dataset
• This model will consist of a Convolutional Neural Network and Recurrent Neural Network
• Convolutional Neural Network:
  • Retrieves spatial features of the data
  • Extracts the details of active areas in the brain
• Recurrent Neural Network
  • Retrieves the temporal features of the data
  • Model the flow of the blood that is associated to certain disorders (or activities)
• Goal: Construct a 3D Convolutional Neural Network + LSTM-based RNN to diagnose the ADHD disorder with the provided fMRI data sample

Currently In Progress

• The model architecture is fully implemented and undergoing a 5-fold cross-validation analysis
• Implemented using Keras with a Tensorflow Backend
• Dataset can be found here: http://preprocessed-connectomes-project.org/adhd200/index.html