A deep learning framework identifies dimensional representations of Alzheimer's Disease from brain structure

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Neurology

Background:

• Brain disease heterogeneity poses challenges for precise diagnosis and prognosis.
• Identifying distinct disease subtypes is crucial for developing targeted treatments.

Purpose of the Study:

• To introduce and validate Smile-GAN (SeMI-supervised cLustEring-Generative Adversarial Network), a semi-supervised deep-clustering method.
• To identify neuroimaging signatures of disease subtypes by examining neuroanatomical heterogeneity.

Main Methods:

• Utilized T1-weighted MRI data from two studies (2,832 participants, 8,146 scans) including cognitively normal individuals and those with cognitive impairment/dementia.
• Applied Smile-GAN, a semi-supervised deep-clustering approach, to regional brain volumes.
• Analyzed longitudinal data to identify distinct disease progression pathways.

Main Results:

• Smile-GAN identified four distinct patterns or axes of neurodegeneration.
• Two unique progression pathways were revealed through longitudinal analysis.
• Pattern expression predicted future neurodegeneration rates and pathways, complementing existing biomarkers like amyloid/tau.

Conclusions:

• Smile-GAN effectively identifies neuroanatomical heterogeneity and disease subtypes.
• The identified patterns and pathways offer potential for precision diagnostics in brain diseases.
• Deep-learning derived biomarkers show promise for targeted clinical trial recruitment and patient management.