Magnetic Resonance Imaging based Feature Extraction and Selection Methods for Alzheimer Disease Prediction

This paper proposes a methodology to predict Alzheimer’s disease patient using their brain MRI scans. Alzheimer’s disease is an irrecoverable one. It is a prolonged degenerative disorder and listed as one of the most frequent dementia threats in individuals over 65 years of age. The suggested solution will be tested on the Alzheimer’s disease Neuroimaging Initiative (ADNI) standard MRI datasets. We obtained MRI scans from two Alzheimer stages that are moderately demented and non-demented. Live Neuron Estimation, Gray-Level Co-occurrence Matrix (GLCM), and Random Forest Mapping are the techniques used to extract features. In the MRI images, Live Neurons known as white pixels. The features like homogeneity, contrast, and correlation determined using the Gray Level Co-Occurrence Matrix (GLCM) and Random Forest mapping helps us to identify the shape and size of other essential parts of the brain like temporal Lobe, occipital Lobe, frontal Lobe, insular. Features that contribute to the prediction identified using the correlation matrix. Distinct machine learning models were employed to predict the presence of disease. The accuracy is 96.4% by Random Forest Classifier, having an area of 82.1% under ROC-AUC. Furthermore, it has the best result obtained over PR Curve. We used a cross-validation score to fine-tune our Random Forest Classifier and configured 100 trees, predicting the best outcome of 95.