Maximizing Accuracy in Alzheimer’s Disease Prediction: A Optuna Hyper Parameter Optimization Strategy Using MRI Images

The increasing incidence of Alzheimer’s disease (AD), the most common type of dementia and a major contributor to overall cognitive decline in the elderly and middle-aged, will pressure healthcare delivery systems. China currently has more people living with Alzheimer’s than any other nation because of its rapidly ageing population. As a result, developing reliable methods for detecting AD disease at an early stage and treating it effectively is of critical importance. Decades of research and development have also led to several automated technologies and methods for Alzheimer’s disease (AD) diagnosis. Diagnostic techniques that emphasise speed, accuracy, and early diagnosis can help reduce the negative impact that this disease has on the patient’s mental health. ML and DL have substantially improved the accuracy of medical imaging systems for Alzheimer’s disease diagnosis. However, multi-class classification is made more difficult by the fact that the brain has strongly associated anatomical characteristics. However, most deep learning models are unable to produce satisfactory output in practical settings, and they also often demand longer training times. To address this issue, we developed a Convolutional Neural Networks-based model and applied the Optuna parameter optimisation method. To put our proposed method to the test, we have amassed 8,980 MRI scans from the OASIS dataset. We found that the Optuna hyperparameter optimisation model outperformed both traditional CNN and Deep CNN models. The training time and trainable parameters of Optuna optimisation model is also smaller when compared to CNN and Deep CNN model.