Alzheimer's Disease Classification Through Imaging Genetic Data With IGnet.

Area of Science:

• Artificial Intelligence in Medicine
• Neuroimaging Analysis
• Genomic Data Science

Background:

• Deep learning is increasingly applied to Alzheimer's disease (AD) detection.
• Advancements in neuroimaging and sequencing generate large-scale data for AD research.
• Integrating diverse data types like imaging and genetics is crucial for robust AD classification.

Purpose of the Study:

• To develop and evaluate IGnet, a deep learning approach for automated AD classification.
• To integrate magnetic resonance imaging (MRI) and genetic sequencing data for enhanced AD detection.
• To leverage multi-disciplinary AI techniques for improved diagnostic accuracy in Alzheimer's disease.

Main Methods:

• Developed IGnet, a deep learning model combining computer vision (CV) and natural language processing (NLP).
• Utilized a 3D convolutional neural network (CNN) for processing 3D MRI data.
• Employed a Transformer encoder for analyzing genetic sequence data (single-nucleotide polymorphisms).

Main Results:

• Achieved 83.78% classification accuracy on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset.
• Obtained an Area Under the Receiver Operating Characteristic Curve (AUC-ROC) of 0.924.
• Demonstrated effective integration of baseline MRI scans and chromosome 19 genetic data.

Conclusions:

• The IGnet approach shows significant potential for automated Alzheimer's disease classification.
• Integrating imaging and genetic data via multi-disciplinary AI enhances diagnostic capabilities.
• This study highlights the value of AI in analyzing complex, multi-modal data for neurological disorders.