Reiman, Eric M.

Abstract INTRODUCTION The Consortium for Clarity in Alzheimer's disease related dementias (ADRD) Research Through Imaging (CLARiTI) is a study that aims to collect standardized imaging and plasma biomarkers on 2000 Clinical Core participants enrolled across all Alzheimer's Disease Research Centers (ADRC) sites. We sought to summarize the known heterogeneity across centers regarding scientific focus and initial enrollment plans for CLARiTI. METHODS We developed and distributed a survey capturing information on the 36 CLARiTI site's theme/expertise, recruitment plans, and the intersection of CLARiTI with other ADRC imaging efforts. RESULTS Anticipated CLARiTI enrollees spanned 11 different categories of suspected etiologies underlying impairment. A wide range of risk factors were endorsed across sites regarding the enrollment of unimpaired individuals. Variability also existed regarding site‐level strategies in enrollment into CLARiTI versus other imaging efforts. DISCUSSION We anticipate that the 2000 individuals that will enroll into CLARiTI will reflect the clinical heterogeneity already in place across the ADRC network. Highlights The ADRC Consortium for Clarity in ADRD Research Through Imaging (CLARiTI) will leverage and contribute to the existing Alzheimer's Disease Research Centers (ADRC) program by supporting standardized imaging and plasma collection across all centers. We summarize the variation in scientific focus and enrollment plans across ADRC sites participating in CLARiTI. The anticipated CLARiTI cohort will reflect the clinical heterogeneity that already exists across the ADRC network. CLARiTI will contribute to scientific goals related to the detection of multi‐etiological signatures relevant for Alzheimer's disease and related disorders (ADRDs).

AbstractBackgroundCross‐modality translation between MRI and PET presents significant challenges due to the distinct mechanisms underlying those modalities. Currently, the leading method for this translation is CycleGAN. Previous studies have proved a strong correlation between BBBMs and brain amyloid status measured by PET. However, the impact of blood‐based biomarkers (BBBMs) on PET image synthesis has not yet been thoroughly evaluated. In this study, we propose Plasma CycleGAN, a generative model based on CycleGAN to synthesize PET images from MRI, while also incorporating BBBMs and other clinical covariates. We show that Plasma CycleGAN outperforms the state‐of‐the‐art CycleGAN for MRI‐to‐PET translation. Our method is the first to integrate BBBMs in a conditional cross modality translation from MRI to PET.MethodOur model was built on CycleGAN architecture, while the backbone was improved by integrating BBBMs as additional conditions in the input domain. To be specific, we introduce a masked perturbation in the MRI image, whose voxel‐level magnitudes are learnable parameters in the neural network. The dataset we used to evaluate our model is downloaded from the ADNI database. The cohort contains 160 subjects with matching BBBM, T1‐weighted MRI images, amyloid PET, and clinical features, including age, sex, weights, etc. The synthesized PET images are compared with the ground truth PET images for performance evaluation using traditional metrics like structural similarity index (SSIM), peak‐signal‐to‐noise ratio (PSNR) and mean square error (MSE). In addition, the Pearson correlation coefficients (PCCs) of the masked SUVR ratios were calculated.ResultWe evaluated the baseline CycleGAN model using MRI input, as well as combined inputs of MRI+BBBM and other clinical features. Our model achieved an SSIM of 0.72±0.16 and PSNR of 22.8±2.3 in PET image synthesis task using MRI+ Aβ42/40, both yielding the best results among all input combinations. MRI+sex achieved the best result using MSE measurement. PCCs of all experimental results ranged from 0.44 to 0.5.ConclusionWe introduce Plasma CycleGAN, the first generative model to integrate BBBMs and clinical features in PET image synthesis from MRI images. In PET image synthesis task, our proposed integrated model outperforms the baseline CycleGAN model.

AbstractBackgroundLewy body (LB) diseases can present with overlapping prodromal, cognitive, motor, autonomic or neuropsychiatric symptoms. Intuitively, greater symptom severity should correlate with greater pathological burden, but this has not been consistently shown. LB pathology does not translate to clinical expression in Incidental LB disease. While several composite scores/toolkits utilize overlapping schemes for diagnosis, a composite data‐derived score to predict LB pathologic burden is lacking. We aimed to utilize an autopsy‐confirmed cohort, and validated, brief, and standardized clinical, functional, pathologic assessments, and machine learning (ML)/quantitative modeling techniques to identify clusters and importance rankings among assessments/measures to predict LB pathologic severity and density.MethodA total of 234 subjects with neuropathological finding of LB at autopsy in the Arizona Study of Aging and Neurodegenerative Disorders (AZSAND) who were classified as Cognitively Unimpaired or Mild Cognitive Impairment after their first study clinical evaluation were included. Of these, 145 subjects had complete scores for 32 of 46 measures (clinical, cognitive, behavioral and functional) that were utilized as predictors of LB pathology. Olfactory function was assessed using the University of Pennsylvania Smell Identification Test (UPSIT). LB severity was assessed by the Unified Staging System for Lewy Body Disorders (USSLB). Several ML algorithms and quantitative prediction models were explored and compared: artificial neural network (ANN), partial least square regression (PLSR), support vector regression (SVR), relevance vector regression (RVR) and ensemble forest regression (EFR) with leave‐one‐out (LOO) scheme. Game‐theory based Shapley methods assessed the impact, including rankings, consistency and magnitude, of model predictors.ResultRVR predicted aggregate LB density with large effect size (R2 = 0.691, p<0.00001). ANN predicted USSLB severity stage with large effect size (R2 = 0.74, p<2.5e‐43). All other ML algorithms/models provided substantial prediction. Across all models, UPSIT was the most influential predictor (>65%), followed by Controlled Oral Word Association Test (COWAT) and age.ConclusionThese preliminary and exploratory results support the utilization of ML techniques/models to assess LB pathologic burden with key measures collected in relatively small samples. UPSIT was consistently ranked highest impactful among clinical and functional measures/predictors. Integrated together in ML/data‐derived composites, UPSIT, COWAT and other clinical characteristics may be of antemortem utility to predict USSLB stages.