Early detection of Alzheimer's disease using deep learning methods

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Alzheimer's disease (AD) is a primary cause of dementia, necessitating early detection for effective treatment.
• Multimodal data, including clinical, biomarker, and neuroimaging information, are crucial for improving diagnostic accuracy.
• Current diagnostic methods can be limited, highlighting the need for advanced analytical tools.

Purpose of the Study:

• To develop and evaluate hybrid deep learning frameworks for early Alzheimer's disease detection.
• To enhance predictive accuracy by integrating diverse data types such as structured clinical data and magnetic resonance images (MRIs).
• To explore the potential of AI in improving Alzheimer's disease diagnosis and enabling timely interventions.

Main Methods:

• A novel hybrid AI framework was developed, combining models for structured data (LSTM, FNN) and MRI data (ResNet50, MobileNetV2).
• Long short-term memory (LSTM) networks captured temporal dependencies, while feedforward neural networks (FNNs) analyzed static patterns in structured data.
• Convolutional neural networks (ResNet50, MobileNetV2) were utilized for spatial feature extraction from MRI scans.
• The models were validated on the National Alzheimer's Coordinating Centre (NACC) and Alzheimer's Disease Neuroimaging Initiative (ADNI) datasets.

Main Results:

• The MRI-based model achieved a high accuracy of 96.19% on the ADNI dataset.
• The hybrid AI model demonstrated superior performance, attaining 99.82% accuracy on the NACC dataset.
• The study confirmed the effectiveness of LSTM models for early AD diagnosis using NACC data.

Conclusions:

• Hybrid AI models show significant promise for early and accurate detection of Alzheimer's disease.
• The findings suggest that AI-driven analysis of multimodal data can lead to improved diagnostic outcomes and facilitate earlier patient interventions.
• The research also proposes a method for the rigorous validation of transfer learning models in medical brain diagnostics.