Role of surfacelet transform in diagnosing Alzheimer’s disease

Abstract

The clinical diagnosis of Alzheimer’s disease (AD) is based on the tedious questionnaire and prolonged tests, which require involvement of the patient, his/her family, and the clinician. Ultimately, the success of subjective evaluation depends on the expertise of the clinician. To eradicate this subjectivity, the present work proposes a method to automate the diagnosis of AD. The proposed method represents complex brain structure of the subject (captured by MRI, a safe and non-invasive medical imaging modality) in terms of quantifiable features and then performs classification of normal and AD subjects. It does so in three phases: (1) MRI volume is normalized and the gray matter, which is the most affected tissue in AD is extracted from the regions of interest. (2) Surfacelet transform, a 3D multi-resolution transform that captures intricate directional details, is applied on the volume to extract features. The features are then reduced by selecting relevant and non-redundant features using a combination of Fisher discriminant ratio, and minimum redundancy and maximum relevance feature selection methods. (3) Based on these features, AD patients and normal subjects are classified using support vector machine classifier with 10 runs of 10 fold cross-validation. The performance of the method was computed using three measures (sensitivity, specificity, classification accuracy) on three datasets, constructed from the publically available database. It achieved 78.04%, 98.00% and 84.37% classification accuracies on the three datasets respectively. The effectiveness of the proposed method is validated by comparing it with other existing methods under identical experimental settings using the above three performance measures as well as receiver operating characteristic curves, ranking and statistical tests. Overall, it significantly outperformed the existing methods. It indicates that the proposed method has the potential to assist clinicians in the decision making for the diagnosis of AD.