A hybrid deep learning architectures and feature extraction techniques for alzheimer disease recognition

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Alzheimer's disease (AD) is a progressive dementia with increasing global prevalence.
• Early detection of AD is critical for effective symptom management and intervention.
• Current diagnostic methods require improvement for timely identification of AD and its early stages.

Purpose of the Study:

• To develop an accurate and efficient deep learning model for discriminating between Alzheimer's disease (AD), Mild Cognitive Impairment (MCI), and cognitively normal (CN) individuals.
• To leverage spatial information from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset for improved classification.
• To introduce a novel Fuzzy Scoring method to enhance the performance of a hybrid deep learning architecture.

Main Methods:

• Utilized a hybrid deep learning architecture combining ResNet and Convolutional Neural Network (CNN) with a novel Fuzzy Scoring technique.
• Implemented Adaptive Grey Wolf Optimization Algorithm (AGWOA) for feature selection and reduced time complexity.
• Applied z-score normalization for data pre-processing.

Main Results:

• The proposed Fuzzy Scoring based ResNet-CNN (FS-ResNet CNN) model achieved a high classification accuracy of 97.89% on the ADNI dataset.
• The model demonstrated computational efficiency, reduced sensitivity to noise, and memory savings.
• Outperformed existing state-of-the-art methods in classifying AD, MCI, and CN subjects.

Conclusions:

• The FS-ResNet CNN model offers a promising, accurate, and efficient approach for the early detection and classification of Alzheimer's disease.
• The integration of Fuzzy Scoring significantly optimizes deep learning network performance for neurodegenerative disease diagnosis.
• This hybrid deep learning strategy holds potential for improving clinical diagnosis and patient management in Alzheimer's disease research.