Gini Based Learning for the Classification of Alzheimer’s Disease and Features Identification with Automatic RGB Segmentation Algorithm

Abstract

Magnetic Resonance Image segmentation is the process of partitioning brain data, which is regarded as a highly challenging task for medical applications, particularly in Alzheimer’s Disease (AD). In this study, we have developed a new automatic segmentation algorithm which can be seen as a novel decision making technique that can help diagnose decision rules studying magnetic resonance images of the brain. The proposed work consist of a total of five stages: (i) the preprocessing stage that involves the use of dilation and erosion methods via gray-scale MRI for brain extraction (ii) the application of multi-level thresholding using Otsu’s method with a threshold value of ( $${\mu _{i}}> 15$$ pixels) to determine the RGB color segment values (iii) the calculation of area detection (RGB segment scores) by applying our newly proposed automatic RGB Color Segment Score Algorithm to the predetermined RGB color segments (iv) creating the AD_dataset using the pixels of the lesion areas calculated via MR imaging (v) the post-processing stage that involves the application of Classification and Regression Tree (CART) algorithm to the AD_dataset. This study aims at contributing to the literature with the decision rules derived from the application of CART algorithm to the calculated RGB segment scores using our newly proposed automatic RGB Color Segment Score Algorithm in terms of the successful classification of AD.