Diagnosis of Alzheimer's disease from 3D MR images with statistical features of hippocampus

Abstract

Alzheimer's disease (AD) is the most common cause of dementia with a progressive course, beginning with neuronal dysfunction, and evolving to irreversible loss of neurons. Early AD detection and treatment allow better clinical results, preserving the loss of cognitive functions for longer periods, with great social economical impact. [2].Subtle hippocampus atrophy can be observed even in early AD, becoming an important biomarker for the disease, especially with disease progression. Volumetric MRI studies already indicate small volume loss of those anatomical structures and, hippocampus appears to shrink. However, there is recent evidence that this shrinkage can be reversed and perhaps prevented in people with depression and bipolar disorder, with effective treatment if timely diagnosed. This paper is on detection of AD from AD, MCI and Normal based on statistical features of Hippocampus derived using MIPAV and further classified by ANN. Our primary goals are: (1) Locate VOI (Volume of Interest) for Hippocampus on 3D MRI. (2)Extract features from this VOI (3) Construct feature vector (4) Train ANN (5) Test ANN. We developed feed forward back propagation neural network in Matlab and trained. Testing of trained ANN revealed the overall accuracy of 80%. VOI analysis is performed using MIPAV. MR Images used are from (ADNI database). 30 AD and 30 Normal images were used for investigation. © 2011 Springer-Verlag.