AbstractBackgroundBiomarker data related to AD are being collected at an ultra‐scale and are likely to unlock numerous opportunities for AD treatment. While most persons with AD dementia are confirmed to have a pathologic diagnosis of AD, additional vascular and non‐AD degenerative pathologies (e.g., Lewy bodies, TDP‐43, and hippocampal sclerosis) are found in most cases. Each of these pathologies has been previously shown to lower the threshold for cognitive impairment. Due to lack of good biomarkers for non‐AD degenerative pathologies, and overlapping cognitive phenotypes, there are few guidelines for diagnosis of multiple versus single etiology dementias.MethodWe used innovative harmonized structural MR image processing and artificial intelligence (AI) methods to model and subtype the vast and complex imaging data in relation to comorbid pathologies. We cost‐efficiently harnessed the available wealth of autopsy validated neuropathological confirmations from landmark neuroimaging studies (e.g., ADNI, NACC, ROS, MAP) and combine with efficient, scalable, and publicly available tools, powered by ‘cognitive systems’ including AI. Since prevalence of each pathology is not mutually exclusive of existence of other pathologies, we developed multi‐class deep learning models for simultaneous prediction of presence of these pathologies.ResultWhen combined with age, sex, MMSE, and Apoe 4 information, within clinically normal participants (independent validation), structural‐MRI based prediction accuracy was 86%/87%/93% for AD degenerative pathologies of amyloid, tau, and neurodegeneration, respectively, with expected 4%‐6% increase in performance in MCI and AD validation cohorts. Within the clinically diverse autopsy cohort, structural‐MRI was predictive of non‐AD pathological confirmations of comorbid TDP‐43 with an AUC of 0.83‐0.85 and comorbid neocortical diffused Lewy bodies with an AUC of 0.76‐0.79.ConclusionThere is an urgent need for biomarkers to identify patient’s neuropathology profile in vivo to inform public health planning, clinical trial enrichment, and development of effective early interventions. Our recent work aims to associate comorbid pathological confirmations across various autopsy studies with in vivo brain metrics, genomics, and plasma biomarkers for deeper understanding of pathologic basis of heterogenous trajectories of cognitive decline in aging. Focusing on routinely collected structural MRI data maximizes translational impact both in clinical practice as well as in clinical trials on AD.