Alzheimer Classification using Deep Learning and Augmentation

Abstract Alzheimer's disease (AD) is a debilitating neurological condition marked by memory loss and an ongoing decrease in cognitive function. Timely management and enhanced patient outcomes are contingent upon an early and precise diagnosis. Deep learning has become an effective technique for AD classification. With the use of picture augmentation techniques and convolutional neural networks (CNNs), it suggests a unique deep learning framework for AD diagnosis. The methodology provides intuitive visualizations of individual vulnerability by creating high resolution (HR) diseases probability diagrams from local brain regions in magnetic resonance imaging (MRI) data. Previous research based on CNNs, and pre-trained models show the promise of deep learning for AD classification, obtaining over 90% accuracy as well as F1 score for all AD phases. Using Convolutional Neural Networks (CNNs) and picture augmentation techniques on an augmented dataset, this research presents a unique deep learning framework for diagnosing AD and distinguishing four stages of the disease: moderate, non-demented, mild, and very mild. This framework is effective in accurately diagnosing AD phases, as proven by its remarkable 94.6% accuracy. In addition, the framework produces high-resolution disease probability maps, which is a major improvement over current techniques and offers clear visualizations of individual risk.