CMStatistics 2023: Start Registration
View Submission - CMStatistics
B0875
Title: Imaging mediation analysis for longitudinal outcomes Authors:  Cai Li - St. Jude Children's Research Hospital (United States) [presenting]
Abstract: The focus is on improving cognitive outcomes for pediatric cancer survivors who undergo aggressive cancer treatments that may affect the central nervous system. Specifically, a new mediation framework is proposed for longitudinal neurocognitive outcomes pertaining to a clinical trial for medulloblastoma, the most common malignant brain tumor in children, using high-dimensional imaging mediators to identify causal pathways and corresponding white matter microstructures. The proposed approach takes into account both the spatial and temporal dependencies and smoothness of the mediators and outcomes, enhancing the detection power of informative voxels and accurately characterizing longitudinal patterns, concurrently. The results offer insights into how to enhance long-term neurodevelopment and strategically spare brain regions that might be impacted by radiation therapy. This understanding will be crucial in planning future treatment protocols, ultimately benefiting brain cancer survivors. The validity and effectiveness of the method are affirmed through numerical studies.