A HYBRID DEEP LEARNING AND MACHINE LEARNING FRAMEWORK FOR EARLY ALZHEIMER'S DISEASE DETECTION FROM MRI

Abstract

This research presents a hybrid approach based on deep learning (DL) and machine learning (ML) techniques for early detection of Alzheimer's disease (AD). The initial stages of AD can be difficult to discern as it courses through its progression. it is critical that an accurate and early detection of the disease occurs to facilitate appropriate intervention and treatment.  An appropriate image analysis dataset is crucial, so the dataset utilized is from the Alzheimer's disease Neuroimaging Initiative (ADNI) which comprises magnetic resonance imaging (MRI) images for four classes: non-demented, very mild demented, mildly demented, and moderately demented. To improve the quality of the input images histogram equalization was employed to improve contrast and adjust for potential subtle structural details. Then features were extracted using transfer learning-based pre-trained Convolutional Neural Networks (CNNs) mainly MobileNetV3 and ResNet50 which are known for their speed, accuracy, and direct applicability to medical images. Features were extracted and were then used with a few traditional machine learning classifiers including Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), K-Nearest Neighbor (KNN), and XGBoost for classification. LR provided the best performance with an accuracy score of 99% for features extracted from ResNet50 and MobileNetV3. The proposed approach is less computationally expensive, and it reduces the complexity of the model. The result of this study is that a hybrid model that uses a transfer learning and a machine learning model provides a successful prediction of AD at an earlier stage of dementia.