Leveraging MATLAB for Machine Learning-Based Identification of Alzheimer’s Biomarkers Using PET Scan Data Analysis

Irfan Biswas and Md Helal Uddin Biswas

Background and Objective: Alzheimer’s disease (AD) is characterized by the accumulation of beta-amyloid and Tau proteins in the brain, making early detection crucial for effective intervention. This study aims to develop a Matrix Laboratory (MATLAB)-based machine learning approach to enhance the identification of these biomarkers using positron emission tomography (PET) scan data, reducing the risk of misdiagnosis. Materials and Methods: Brain imaging data, specifically PET scans highlighting beta-amyloid and Tau protein accumulations, were processed using MATLAB. The images were transformed into 1D RGB data tables for analysis. Machine learning models, including Fine K-Nearest Neighbors (KNN), Support Vector Machines (SVM) and neural networks, were trained and validated using MATLAB’s Classification Learner App to classify different severity levels of protein accumulation. The study involved a total of 5,402 PET scan images, split into training and testing datasets. Results: The MATLAB-based analysis demonstrated a high level of accuracy in identifying protein accumulation levels, achieving a test accuracy of 97.6% with the Fine KNN model and 98.07% with neural networks on new, unseen data. The RGB analysis technique effectively differentiated between healthy, medium-unhealthy and unhealthy biomarker levels, providing a reliable tool for evaluating AD progression. The automated approach significantly reduced the potential for human error in interpreting PET scan images. Conclusion: The study showcases the potential of using MATLAB for automated analysis of PET scans in diagnosing Alzheimer’s disease. This approach provides an efficient and accurate method for early detection, offering valuable insights for medical professionals and contributing to advancements in AD diagnostics.

View Fulltext Back