ROTATIONAL MOMENT SHAPE FEATURE EXTRACTION AND DECISION TREE BASED DISCRIMINATION OF MILD COGNITIVE IMPAIRMENT CONDITIONS USING MR IMAGE PROCESSING

Abstract

Mild Cognitive Impairment (MCI) is the preclinical, asymptomatic stage for Alzheimer’s condition, which affects a large amount of the aging population around the world. Detection of MCI condition can ensure timely intervention needed for handling the disease severity. Morphological alterations of the Lateral Ventricle (LV) are considered a significant biomarker for diagnosing MCI conditions. This work aims at analyzing the shape alterations of LV from brain Magnetic Resonance (MR) images using Rotational moment shape features and differentiating MCI conditions using Decision Tree (DT) based classification. Trans-axial brain MR images are obtained from a publicly available OASIS database. Segmentation of LV is performed using the Reaction Diffusion level set, and the results are validated against Ground Truth. Rotational moment shape features are extracted from the segmented LV images. DT is implemented for the differentiation of control and MCI subjects. Results show that Rotational moment shape features are able to capture the alterations of LV in control and MCI subjects (p<0.05). The classification model achieves a high detection accuracy of 96.73% and an F-measure of 96.82%. Hence, the proposed method can be used as an automated diagnostic tool to predict and monitor the cognitive decline in MCI subjects and can aid in disease management.