Big Data Analytics and Machine Intelligence in Biomedical and Health Informatics

3Pre-Trained CNN Models in Early Alzheimer’s Prediction Using Post-Processed MRI

Kalyani Gunda* and Pradeepini Gera

Department of Computer Science & Engineering, Koneru Lakshmaiah Education Foundation, K L University, Guntur, India

Abstract

Although having advanced medical equipment and technology, detecting Alzheimer’s is still a challenging issue. Innovative diagnostic and therapeutic devices improve individual wellbeing. As Alzheimer’s predominantly prey the elders, nearly 5.8 million aged are carrying Alzheimer’s asserted by the Alzheimer’s Association. Numerous researches functioning to predict dementia at an advanced exploration are idealistic. The proposed article works on both Alzheimer’s MRI details and images that are volitional. The initial objective was to test MR scan with dementia or not by non-image MRI evidence using Random Forest Classifier which obtained 87% accuracy without false prediction and also by predicting Alzheimer’s progression using advanced CNN models. At work, Gentle Dementia is more focused to train the early detection by omitting converted MR sessions. Various Transfer Learning Deep neural Networks like Residual Network (ResNet50), GoogleNet, VGG19 (Visual Geometric Group), MobileNet, and AlexNet are compared to classify Alzheimer’s. Model comparison evaluated to explicate model efficacy.

Keywords: Alzheimer’s disease, OASIS Dataset, Alzheimer’s 4-Class-Image-Dataset, Machine Learning and Transfer Learning, Random Forest Classifier, MobileNetV2, ...