Transfer Learning and Machine Learning With MRI Radiomics For Alzheimer's Disease Diagnosis

Abstract

Alzheimer's disease (AD) is a neurological disorder that causes considerable disability. Thus, it is essential to look at new methods for early disease detection that need the least amount of human intervention. Radiomics is a relatively new field of study that involves turning the features and information found in medical images into quantifiable data and then mining that data to provide better decision support. The goal of this work is to highlight how important radiomics provides in the various classifications of Alzheimer's stages by applying machine learning and transfer learning models using MRI data which can help in the early diagnosis of Alzheimer's. Support Vector Machines (SVM), Random Forest, and Gradient Boosting Machines (GBM) are combined with the VGG19 convolutional neural network (CNN) architecture and were applied for classifying radiomics data, testing transfer learning and machine learning classifiers by assessing their output based on accuracy, precision, recall, and F1score.Results show that the VGG19 model performs robustly in Alzheimer's MRI image classification, attaining high training accuracy (98.43%), precision (97.04%), recall (96.68%), and F1-score (96.85%). The finding refers to VGG-19 as suitable for use with a range of medical imaging modalities, because of its general-purpose architecture. Its adaptability to different types of images accounts for its flexibility when used in radiomic research involving several datasets.