Alzheimer’s prediction via CNN-SVM on chatbot platform with MRI

Abstract

Artificial intelligence (AI), consisting of models and algorithms capable of concluding data to produce future predictions, has revolutionary potential in various aspects of human life. One application is an Alzheimer’s disease (AD) prediction chat robot (chatbot). Only now has a method provided very accurate findings and recommendations regarding the early detection of AD using magnetic resonance imaging (MRI). Therefore, this research aims to measure AD prediction performance in four stage classes, namely very mild demented, mild demented, moderate demented, and non-demented, using brain MRI images trained in the convolutional neural network (CNN)- support vector machine (SVM) model. The research involved nine combination schemes of dataset proportions and preprocessing in the CNNSVM model. Evaluation shows that scheme 1 produces the highest accuracy, precision, recall, and F1-score, namely 98%, 99%, 98%, and 98%. The chatbot, trained using CNN, achieved 99.34% accuracy in question responses, and was then combined with AD prediction models for improved accuracy. The test results show that the chatbot functionality runs well for each transition, with a functionality score reaching 99.64 points out of 100.00. This success shows excellent potential for early detection of AD. This research brings new hope in preventing AD through AI, with potential positive impacts on human health and quality of life.