Support vector machine learning and diffusion-derived structural networks predict amyloid quantity and cognition in...

Area of Science:

• Neurology
• Genetics
• Medical Imaging

Background:

• Down syndrome (trisomy 21) is linked to a near-certain risk of Alzheimer's disease neuropathology.
• Alzheimer's disease is characterized by amyloid plaques and neurofibrillary tangles.
• Non-invasive methods are needed to predict disease progression.

Purpose of the Study:

• To evaluate diffusion-weighted imaging and connectomic modeling for predicting Alzheimer's disease markers.
• To assess the prediction of brain amyloid plaque burden, baseline cognition, and cognitive change.
• To utilize machine learning for non-invasive Alzheimer's disease prognosis in Down syndrome.

Main Methods:

• Ninety-five participants with Down syndrome underwent Pittsburgh Compound B (PiB) PET-MR scans and memory assessments.
• Support vector regression was employed for predictive modeling.
• Graph theory metrics (node degree, strength, connection density) of the structural connectome were analyzed.

Main Results:

• Graph theory metrics effectively predicted global amyloid deposition.
• Baseline structural network connection density predicted current cognitive performance.
• Reduced white matter connectivity correlated with amyloid positivity and faster cognitive decline.

Conclusions:

• Structural connectome analysis is effective for predicting Alzheimer's disease neuropathology in Down syndrome.
• Machine learning methods offer a non-invasive approach to assess Alzheimer's disease prognosis.
• White matter integrity plays a crucial role in Alzheimer's disease progression.