Alzheimer’s disease (AD) and mild cognitive impairment (MCI) are two dementia-related brain illnesses that are prevalent among elders in the twenty-first century. MCI is treated as the initial stage of AD, and once the illness reaches the AD stage, there is no escape from certain death. The accuracy and efficacy of current multiclass computer-based approaches to diagnose AD and MCI are constrained by traditional machine learning (ML) classifiers due to their shallow architecture. This makes it challenging to make a prompt and accurate diagnosis of AD and MCI. This research proposes a framework employing electroencephalography (EEG) to diagnose MCI, AD, and healthy subjects (HSs) to boost multiclass performance and efficacy. EEG is a portable, non-invasive, and affordable means to identify brain problems as compared to expensive and time-consuming techniques like computed tomography (CT) scans, positron emission tomography (PET), magnetic resonance imaging (MRI), and the mini-mental state examination (MMSE). To circumvent these issues, the Cognitive Decline Recognition Network (CDR-Net) architecture has been developed to identify MCI, AD, and healthy individuals using EEG data. The proposed architecture allows for the acquisition of EEG data, data preprocessing (down-sampling, noise cleaning, segmentation, and digital picture construction), feature extraction and classification using CDR-Net, as well as performance assessment and cross-validation stages. Our suggested CDR-Net architecture produced better multiclass accuracy, sensitivity, and specificity of 99.25%, 99.13%, and 99.32%, respectively. By using 10 folds and leave-one-out cross validations, stability, consistency, and data overfitting and underfitting concerns are addressed. This framework will serve as a foundation for future systems designed to detect multiple brain disorders.
Alvi et al. (Mon,) studied this question.