<![CDATA[Alzheimer's disease (AD) is a progressive neurodegenerative disorder characterized by cognitive decline and memory loss, necessitating early and accurate diagnosis for effective intervention. This study evaluates the performance of the K-Nearest Neighbor (K-NN) algorithm on a pre-processed Alzheimer MRI dataset, focusing on the challenge of imbalanced classes. The dataset, sourced from Kaggle, comprises 6400 MRI images resized to 128x128 pixels and categorized into four classes: Non-Demented, Mild Demented, Moderate Demented, and Very Mild Demented. Pre-processing involved segmentation using the Canny edge detection method and feature extraction through Hu Moments. The dataset was split into training (80%) and testing (20%) sets, with features scaled to a mean of 0 and variance of 1. The K-NN algorithm was evaluated using cross-validation with five different k values, revealing moderate performance metrics: accuracy ranging from 45.86% to 50.47%, precision from 41.87% to 47.00%, recall from 45.86% to 50.47%, F1-score from 42.42% to 47.58%, and ROC AUC from 55.18% to 58.87%. The results highlight the significant impact of class imbalance on the algorithm's performance, particularly for the underrepresented Moderate Demented class. This study underscores the need for techniques to address class imbalance to enhance classification accuracy. Future research should explore advanced methods such as data augmentation, re-sampling, and ensemble learning, as well as the evaluation of other machine learning models. These findings contribute to the field of medical image analysis and have practical implications for improving diagnostic tools for Alzheimer's disease.]]>
Copyrights © 2024
