Sunday, Aug 3: 2:00 PM - 3:50 PM
0710
Topic-Contributed Paper Session
Music City Center
Room: CC-207D
Ophthalmology and vision science are crucial in the medical field due to their fundamental role in daily life and overall well-being. They enable individuals to interact with their environment, perform essential tasks, and connect with others. Vision health is closely linked to cognitive development, quality of life, and the early detection of systemic diseases. By prioritizing vision care, we enhance a vital aspect of human health, promoting independence and improving outcomes across various life stages.
Research in ophthalmology and vision science seeks to uncover the intricate mechanisms underlying the structure and function of the eye in health, disease progression, and potential treatment effects. Such studies often generate comprehensive datasets, presenting significant challenges in statistical analysis. These challenges include: (i) managing diverse data types, such as imaging data (OCT and MRI), functional and structural measures, metabolomics/genomic data, and electronic health records; (ii) addressing complex data structures, including inter-eye correlations, repeated measurements over time and imbalanced data; and (iii) handling complications from irregularly spaced visits, missing data, and the integration of time-varying and static covariate effects. Overcoming these challenges requires sophisticated statistical modeling and the development of specialized methods tailored to the complexities of ophthalmic research.
This session aims to bring together biostatisticians who specialize in the analysis of complex ophthalmic data. The primary objectives are to discuss the development of innovative statistical and computational methods and novel computational tools in ophthalmic research. Additionally, we endeavor to determine translational impact in healthcare, accelerating effective treatments, enhancement of clinical decision support systems, and promotion of health equity. Statistical methods include but not limited to machine learning methods for ophthalmic imaging data, finite sample correction, time-to-event analysis, regression models for complex dependencies, etc.
Applied
Yes
Main Sponsor
Section on Statistics in Epidemiology
Co Sponsors
Committee on Applied Statisticians
ENAR
Presentations
Neovascular Age-related Macular Degeneration (wet AMD) is a chronic eye disorder. While it accounts for only about 10-15% of all AMD cases, it is responsible for the majority of severe vision impairment associated with AMD, making early detection and management crucial. In the context of precision medicine for wet AMD, we introduce a novel joint model of longitudinal Drusen biomarkers and AMD progression. The proposed model is designed to handle a large set of longitudinal biomarkers by incorporating a nested structure that accounts for subject-biomarker interactions. The model employs nonparametric functional trajectories using a 'nested Dirichlet process prior' to manage nested clustering where subject clusters are nested within biomarker clusters. The proposed method's nested structure supports a more comprehensive and scalable approach, enabling the model to capture the heterogeneous effects of biomarkers across different patient subgroups. For instance, variable selection is performed both globally (to identify key biomarkers) and locally (to examine their varying effects across subgroups). This allows the model to highlight situations where a biomarker may have opposite effects on different subgroups, offering valuable insights to medical experts on patient stratification. We evaluate the performance of the proposed approach through simulation studies and apply it to real-world data analysis. Our findings align with recent AMD literature while also identifying drusen biomarkers previously deemed insignificant, exhibiting both positive and negative effects on AMD progression, varying across different patient subgroups.
Glaucoma is a chronic condition that can cause significant psychosocial distress, yet screening for distress remains rare due to the high cost of collecting gold-standard labels. To address this challenge, we develop a novel semi-supervised learning approach that leverages a large historical dataset containing proxy indicators of distress to inform prediction in a smaller, prospectively collected cohort with gold-standard outcomes. Rather than treating the historical data as directly labeled, we use it to construct informative priors on model parameters in a Bayesian regression framework, while explicitly accounting for discrepancies between proxy and true outcomes. This prior-informed strategy enables us to borrow strength from the large but imperfect dataset without assuming outcome equivalence, a common limitation in traditional semi-supervised methods. We demonstrate our approach through simulation studies and real-world data from glaucoma patients at the Duke Eye Center, showing improved prediction accuracy and robust uncertainty quantification in low-label clinical settings.
Glaucoma is a major cause of blindness and vision impairment worldwide, and visual
field (VF) tests are essential for monitoring the conversion of glaucoma. While previous
studies have primarily focused on using VF data at a single time point for glaucoma
prediction, there has been limited exploration of longitudinal trajectories.
Additionally, many deep learning techniques treat the time-to-glaucoma prediction as
a binary classification problem (glaucoma Yes/No), resulting in the misclassification of
some censored subjects into the nonglaucoma category and decreased power. To
tackle these challenges, we propose and implement several deep-learning approaches
that naturally incorporate temporal and spatial information from longitudinal VF data
to predict time-to-glaucoma. When evaluated on the Ocular Hypertension Treatment
Study (OHTS) dataset, our proposed convolutional neural network (CNN)-long short-term
memory (LSTM) emerged as the top-performing model among all those examined.
Speaker
Ruiwen Zhou, Washington University in St. Louis
Transfer learning (TL) is widely employed to address data inequality in healthcare systems, including ophthalmology and vision health. Various TL approaches are integrated into deep learning frameworks to accommodate complex dependencies, such as longitudinal and image data; however, interpretability remains a significant challenge. This study aims to develop an explainable, regression-based TL method for high-dimensional longitudinal studies. We propose a TL algorithm based on general linear mixed-effects models to transfer useful information from multimodal source domains, enhancing the estimation of fixed-effect coefficients and facilitating variable selection in the target domain while effectively managing complex data dependencies, such as repeated measurements. A simulation study was conducted to validate the predictive performance of the proposed approach.
Estimating heterogeneous treatment effect (HTE) for survival outcomes has gained increasing attention, as it captures the variation in treatment efficacy across patients or subgroups in improving survival or delaying disease progression. However, most existing methods focus on post hoc subgroup identification rather than simultaneously estimating HTE and selecting causal subgroups. In this paper, we propose an interpretable HTE estimation framework that uses meta-learners with the conditional inference tree to estimate CATE for survival outcomes and identify predictive subgroups simultaneously. We evaluated the performance of our method through comprehensive simulation studies in various randomized clinical trial (RCT) settings. Furthermore, we demonstrate its application in a large RCT for age-related macular degeneration (AMD), a progressive polygenic eye disease, to estimate the HTE of an antioxidant and mineral supplement on time-to-AMD progression and to identify genetics-based subgroups with enhanced treatment effects. Our method offers a direct interpretation of the estimated HTE and provides evidence to guide precision medicine and healthcare.
Keywords
interpretable heterogeneous treatment effect
precision medicine
randomized clinical trials
subgroup identification
age-related eye disease studies (AREDS)
Co-Author(s)
Na Bo, Virginia Commonwealth University
Ying Ding, University of Pittsburgh
Speaker
Na Bo, Virginia Commonwealth University