Abstract
AbstractBackgroundAlzheimer's (AD) and Parkinson's (PD) diseases remain the two most widespread neurodegenerative disorders of our time. Though PD is far less prevalent than AD, its complex symptomatology can be useful in helping understand early progression of AD. Beyond cardinal motor symptoms, the majority of PD patients develop mild cognitive impairment (MCI), which is a common prodromal stage of AD. We employed spatially adaptive machine learning models to reveal shared morphometric signatures between AD and PD‐MCIMethodWe computed shape features in 7 subcortical regions based on T1‐weighted (T1w) MRI data from 239 PD subjects with normal cognition (PD‐NC) controls/120 de novo PD‐MCI patients from the PPMI cohort, and 225 controls (ADNI‐CTL) /398 ADNI‐MCI patients from the ADNI cohort. We first trained two separate sparse spatially regularized (TV‐L1) Logit models, one to distinguish ADNI‐MCI from ADNI‐CTL and the other to separate PD‐MCI and PD‐NC. We then coupled the training of the two Logit models, penalizing the model difference weighted by local mutual information (MI) of the linear weights’ patterns. We used ROC AUC scores to assess model accuracy. As the MI map itself is updated during training, it reveals the degree of similarity between the degenerative patterns. To test its statistical significance, we bootstrapped a distribution of MI weights when optimizing Logit models for randomly labelled ADNI and PPMI subjects. The nonparametric p‐values were corrected using searchlight False Discovery Rate.ResultOut‐of‐sample ROC AUC for PD‐MCI prediction was 0.654 for the naïve model, and 0.704 for the joint model. Models and MI are displayed in Fig. 1. Only the left putamen showed a significantly similar pattern after FDR correction (Fig. 2).ConclusionWe utilized a spatially adaptive machine learning model to discover shared imaging biomarkers of cognitive decline in AD and PD. Predictive accuracy improves when the two models are trained jointly. Such shared information may provide clues for pathological understanding of cognitive decline across common neurodegenerative diseases.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.