A Comparitive Evaluation Of Custom CNN, Sequential CNN & Dense-Net Models to forecast Dementia

Dementia is a very common neurological condition that affects the elderly. Mild cognitive impairment sufferers run the risk of contracting this illness. It steadily worsens from mild to severe, making it challenging for the person to complete any task without assistance. To accurately assess the patient’s state and begin receiving intense medical care, a dementia diagnosis must be made as soon as possible. Manual diagnosis takes a lot of work, takes a long time, and is prone to error. In this study, three CNN models Dense-net, Sequential, and Custom Convolutional Neural Network are evaluated for the diagnosis of dementia. All these models’ performances were compared, and results were found. Sequential model obtained an accuracy of 83%, Dense-Net achieved an accuracy of 88% and the Customized Sequential model achieved an accuracy of 98%. Using Python Flask, a web application that offers the best graphical user interface is further developed. The graphical user interface can upload the MRI scan images and provide these images to the best trained model which predicts whether the person is suffering from dementia or not and can categorize the stage of the disease.