Related Experiment Video
Updated: Jun 5, 2026
14:27
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
Predicting Autopsy-Confirmed Neuropathology across Clinical, Neuroimaging, and CSF Biomarkers using Machine Learning
Christopher Patterson1, Tamoghna Chattopadhyay1, Sophia I Thomopoulos1
1Imaging Genetics Center, Mark and Mary Stevens Neuroimaging and Informatics Institute, Keck School of Medicine, University of Southern California, Marina del Rey, CA, USA.
Biorxiv : the Preprint Server for Biology
|June 4, 2026
Summary
Machine learning models can predict Alzheimer's disease neuropathology using clinical data, CSF biomarkers, and neuroimaging. Different data types excel at identifying specific pathologies, aiding diagnosis and treatment.
Area of Science:
- Neurology
- Biomedical Engineering
- Computational Biology
Background:
- Accurate in vivo prediction of neuropathology is crucial for diagnosing and treating Alzheimer's disease and related dementias (ADRDs).
- Individuals with ADRDs often present with mixed pathologies, necessitating comprehensive diagnostic approaches.
- Inferring neuropathologic changes from clinical data, biofluid assays, and neuroimaging is an area of active research.
Purpose of the Study:
- To evaluate automated machine learning models for predicting 26 autopsy-confirmed neuropathological outcomes in vivo.
- To assess the predictive performance of clinical data, cerebrospinal fluid (CSF) biomarkers, and various neuroimaging modalities.
- To quantify the added value of neuroimaging and CSF biomarkers compared to clinical features alone.
Main Methods:
- Utilized automated machine learning models trained on data from 7,894 individuals curated by the AD Sequencing Project Phenotype Harmonization Consortium.
- Employed ensemble learning with stratified cross-validation to train predictive models.
- Assessed model performance using Spearman's rank correlation and Matthews correlation coefficient, considering co-occurring pathologies.
Main Results:
- Braak stage was among the most consistently predicted neuropathological outcomes.
- CSF biomarkers effectively predicted beta-amyloid and tau pathology.
- Diffusion MRI metrics showed superiority in predicting vascular brain injury, white matter injury, and Lewy body disease.
- Structural MRI measures outperformed clinical assessments for neurodegeneration and hippocampal sclerosis.
- White matter hyperintensities (WMH) complemented cognitive measures in predicting TDP-43 pathology.
Conclusions:
- Machine learning models can effectively infer various neuropathologies from multimodal in vivo data.
- Different data modalities (CSF, diffusion MRI, structural MRI, clinical data) offer complementary strengths in predicting specific pathologies.
- These findings establish a valuable baseline for comparing different data modalities in neuropathology inference.