Abstract
Background: Persons with Parkinson’s disease (PD) differentially progress to cognitive impairment and dementia. With a 3-year longitudinal sample of initially non-demented PD patients measured on multiple dementia risk factors, we demonstrate that machine learning classifier algorithms can be combined with explainable artificial intelligence methods to identify and interpret leading predictors that discriminate those who later converted to dementia from those who did not.
Method: Participants were 48 well-characterized PD patients (Mbaseline age = 71.6; SD = 4.8; 44% female). We tested 38 multi-modal predictors from 10 domains (e.g., motor, cognitive) in a computationally competitive context to identify those that best discriminated two unobserved baseline groups, PD No Dementia (PDND), and PD Incipient Dementia (PDID). We used Random Forest (RF) classifier models for the discrimination goal and Tree SHapley Additive exPlanation (Tree SHAP) values for deep interpretation.
Results: An excellent RF model discriminated baseline PDID from PDND (AUC = 0.84; normalized Matthews Correlation Coefficient = 0.76). Tree SHAP showed that ten leading predictors of PDID accounted for 62.5% of the model, as well as their relative importance, direction, and magnitude (risk threshold). These predictors represented the motor (e.g., poorer gait), cognitive (e.g., slower Trail A), molecular (up-regulated metabolite panel), demographic (age), imaging (ventricular volume), and lifestyle (activities of daily living) domains.
Conclusion: Our data-driven protocol integrated RF classifier models and Tree SHAP applications to selectively identify and interpret early dementia risk factors in a well-characterized sample of initially non-demented persons with PD. Results indicate that leading dementia predictors derive from multiple complementary risk domains.
Plain language summary
Not all persons diagnosed with Parkinson’s disease, a clinical movement disorder, also develop cognitive impairment or dementia. This study looked at the question of which specific dementia risk factors are likely to affect who does or does not develop dementia. We used an Alberta-based Parkinson’s disease study which collected relevant information from the same people repeatedly over a long period of time. At the initial assessment, all of the participants in this study had Parkinson’s disease, but none had cognitive impairment or dementia. However, at the end of three-year study period, some of the initial group had developed dementia and some had not. We used some new mathematical techniques to sort through a large set of risk factors to discover those that best predicted which persons with Parkinson’s disease developed dementia three years later. An important finding is that we were able to detect a set of specific factors that increased dementia risk several years prior to dementia diagnosis. A promising clinical implication is that some of these early predictors may be modifiable. This means that it may be possible to develop strategies for reducing their impact on the level of dementia risk for some persons living with Parkinson’s disease. We plan to continue our research in this area by applying this new approach to broader studies, including the Canadian Consortium on Neurodegeneration in Aging.