Enhanced Alzheimer's Disease Abnormality Classification in Medical Imaging Using YOLOv8

This paper introduces a thorough investigation of the application of YOLOv8 to classify Alzheimer's Disease (AD) related abnormalities from medical images. The concern is to detect major AD markers, such as amyloid plaques and neurofibrillary tangles, using a YOLOv8 architecture tailored to a specific application. Accuracy, precision, recall, F1-score, and inference speed techniques are applied to assess the model's performance. The study uses an extensive corpus of Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) scan images. The results confirm the capability of YOLOv8 to classify AD-related abnormalities at high accuracy and high inference speed for automated diagnostic support. This paper addresses a detailed investigation of model architecture, training methods, and performance for different imaging modalities. The paper also addresses data augmentation methods, the effect of class imbalance and detection visualization. This paper presents useful contributions to applying YOLOv8 in early AD detection and tailored healthcare.