A NEURAL NETWORK BASED APPROACH FOR PREDICTING A PATIENT'S CONVERSION TO ALZHEIMER'S DISEASE BASED ON BRAIN SCAN DATA
Open Access
Author:
Srinivasan, Mahadevan
Graduate Program:
Electrical Engineering
Degree:
Master of Science
Document Type:
Master Thesis
Date of Defense:
May 13, 2011
Committee Members:
David Jonathan Miller, Thesis Advisor/Co-Advisor David Jonathan Miller, Thesis Advisor/Co-Advisor George Kesidis, Thesis Advisor/Co-Advisor
Keywords:
Alzheimer s disease MCI Conversion Neural Network MCI to AD
Abstract:
We studied several neural network architectures for predicting whether a patient will convert to Alzheimer's disease after being initially diagnosed with Mild Cognitive Impairment. The first architecture to be studied was what we call a Localized Neuron Architecture which tries to learn the non-linear relationship between the brain atrophy and the age of the patient. Next, we studied how good the performance is when using a standard multilayer perceptron architecture. We used the brain scan data of the first visit only since the prediction is prognostic. Furthermore, we observed how including the age of the patient when the base line scan was taken would impact the performance. On a previous study based on this data, a support vector machine (SVM) was used to predict conversion. Here, we are using a neural network in place of an SVM. Also, we are trying to predict when the conversion will happen. The challenge is to train a neural network of the correct size and correct structure such that the error in predicting conversion and the standard deviation in the prediction of time at which conversion takes place are minimal.
