A Hybrid Approach to Alzheimer’s Diagnosis: Predictive Modeling Using Diagnostic Data and MRI Images Integrated in a Web Interface

Alzheimer's is a neurological disease that worsens with time and has no known cure. There is no turning back once the diagnosis is made; cognitive deterioration simply becomes worse. This research aims to reduce the time taken to visit the doctor for a diagnosis, which is only done once the patient or their family is suspicious because of any symptoms shown. This time is reduced by sifting through data when a patient goes to their normal checkups, this way, any abnormality that points at a chance of Alzheimer’s can be caught and flagged by the ML (machine learning) model. Two ML models were trained to predict the percentage of chance of a patient having Alzheimer’s. One model was trained by feeding it datasets of biomarkers and the other by using MRI brain scans of patients with and without Alzheimer’s. These models were then integrated into a web interface for ease of use, where a combined prediction using probabilities calculated by both the models would be displayed.