MobAlz: An Optimized MobileNet Model for Early Alzheimer's Detection via Brain MRI Analysis

Alzheimer's disease is a unique form of dementia that impacts an individual's daily existence. In general, individuals who are 65 years of age or older are at a higher risk. The most detrimental aspect of this illness is its progressive deterioration over an extended period of time. Memory loss is minimal during the initial stages of this disease; however, it becomes severe after a few years. Therefore, it is imperative to detect Alzheimer's disease at an early stage in order to initiate the treatment process as promptly as feasible. Many researchers try to detect Alzheimer's disease with the help of Deep Learning (DL) from Magnetic Resonance Imaging (MRI). However, the proposed approach does not meet people's expectations. Therefore, using data augmentation, we are putting forth a more precise, fine-tuned transfer learning model for Alzheimer's disease identification. We are using MRI images of the brain to classify Alzheimer's disease. For feature extraction we are using the MobileNetV2 model. After the extraction, the features passed to the classification block. We have used an augmentation technique to remove the imbalance of the dataset. The proposed approach achieved an accuracy of 96.2% with a precision of 0.96.