Curvelet- and Contourlet-Based CNN for the Early Prediction of Alzheimer's Disease

Alzheimer's Disease (AD) is the most common form of dementia. It gradually progresses from mild to severe, interfering with the patient's ability to live without assistance. Currently, the diagnosis of AD heavily relies on clinical practices. However, these methods suffer from subjectivity and alternity. This paper introduces a curvelet transform based-convolutional neural network (CNN) (DeepCurvMRI) model and contourlet transform based-CNN (DeepContMRI) for early-stage AD disease detection from Magnetic Resonance Imaging (MRI) images. When applied on open MRI dataset, DeepCurvMRI achieved an accuracy of 97.5% and an AUC score of 99.2 %, whereas DeepContMRI attained an accuracy of 95.78% and an AUC score of 98.4 %, which both are higher than conventional deep learning methods. This performance demonstrates that incorporating image transform techniques as feature extraction methods in deep learning models improves model accuracy and efficiency, towards better early prediction of AD.