Ph.D. in Mathematical Sciences Dissertation Defense
Kruti Pandya
A Bayesian latent scale nonlinear mixed effect model
for misaligned longitudinal data.
MS Biostatistics, SUNY-Albany, 2006
MSc Mathematics, Gujarat University, 2002
BSc Mathematics, St. Xavier’s College, Ahmedabad 2000
Defense date: Friday, May 6, 2022
Time: 9:00AM
Zoom link: https://pdx.zoom.us/j/3215819653
Physical Location: FMH 418, 1855 SW Broadway, 97201
Committee chair: Bruno Jedynak, PhD
Committee members: Jong Sung Kim, PhD
Mau Nam Nguyen, PhD
Wayne Wakeland, PhD
Murat Bilgel, PhD
Abstract:
In order to understand the entire course of slow progressing diseases, it is essential to characterize long term disease dynamics from a healthy stage to the disease stage. Cohort studies typically recruit subjects at different stages of the disease and then follow them longitudinally for a relatively short period of time. In this dissertation, we propose a novel Bayesian nonlinear mixed effect model with latent time scale to characterize long term disease dynamics using the observed short term longitudinal data from cohort studies without relying on clinical diagnosis. This model can account for the multimodal misaligned data that is collected as longitudinal noisy samples. We train the proposed model using RStan, which is based on Hamiltonian Monte Carlo to estimate the model parameters. In order to test the novel model for parameter recovery, we first conducted a simulation data experiment and later applied the model to two real world datasets; Alzheimer’s Disease Neuroimaging Initiative (ADNI) for dementia and the optical coherence tomography (OCT) data for multiple sclerosis.
The results from the simulation experiment indicated that all ground truth fixed effect parameters are within the estimated 90 % credible interval. While, more than 80 % of ground truth random effects are within their estimated 90 % credible interval. For the ADNI dataset, we were able to predict time from Alzheimer’s disease with RMSE of 1.85 years and found that the latent scale was significantly correlated with discrete disease staging scale, Clinical Dementia Rating Sum of Boxes (CDRSB) with correlation coefficient of 0.66. Finally, for the OCT data, the estimated slopes of the biomarkers were consistent with the hypothesis in the published literature and the latent scale was significantly correlated with the commonly used measure, expanded disability status scale (EDSS) with correlation coefficient of 0.45. In addition to that, we have also demonstrated the use of this model for prediction of future biomarker templates given the subjects baseline visit measurements. The above results support the broader clinical utility of the model for staging subjects and predicting biomarker templates from limited short term longitudinal data.