Alzheimer's subtypes A supervised, unsupervised, multimodal, multilayered embedded recursive (SUMMER) AI study.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Genetics

Background:

• Alzheimer's disease (AD) is a complex, heterogeneous condition with no consensus on subtype classification.
• Identifying distinct AD subtypes is crucial for developing targeted interventions and improving patient outcomes.

Purpose of the Study:

• To leverage complementary AI approaches (supervised and unsupervised learning) within a recursive pipeline (SUMMER) to identify robust Alzheimer's disease subtypes.
• To integrate multimodal datasets including MRI, phenotypes, and genetic data for comprehensive subtype analysis.

Main Methods:

• Utilized the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, including MRI, SNP genetic data, and clinical information from 1133 European American participants.
• Applied unsupervised learning to MRI gray matter data (44 ROIs) to identify distinct AD clusters by sex.
• Employed XGBoost supervised learning with genetic data (SNPs) to validate the identified AD subtype clusters, stratified by sex.

Main Results:

• Identified distinct AD subtype clusters for men and women based on MRI data.
• Supervised learning models confirmed the homogeneity of these clusters with high accuracy (men: 0.85, women: 0.81).
• Genetic models identified AD subtypes in 34% of men and 47% of women; lower body weight was associated with the women's subtype.

Conclusions:

• AI-driven multimodal analysis successfully identified distinct Alzheimer's disease subtypes.
• Discriminative biomarkers (ROIs and SNPs) were extracted, paving the way for personalized diagnostics.
• This approach enhances the potential for tailored treatments and improved patient outcomes in complex diseases like AD.