| Contents |
This talk will focus on handling missing data problems in multimodal fusion for neuroimaging data. We primarily consider two typical cases with missing values: missing due to dropouts in the longitudinal study and failing specific modalities due to acquisition or quality control failure. For multimodal fusion for the longitudinal neuroimaging data, we consider canonical correlation analysis for two longitudinal variables that are possibly sampled at different time resolutions with irregular grids, missing at specific time points, and dropping out. We modeled trajectories of the multivariate variables using random effects and found the most correlated sets of linear combinations in the latent space. Our numerical simulations showed that the longitudinal canonical correlation analysis (LCCA) effectively recovers underlying correlation patterns between two high-dimensional longitudinal data sets. The second part of this talk will discuss general missing problems in multimodal fusion. Current literature often includes only complete cases or imputes once based on some external imaging restoration models. However, the current practice uses only some available information or potentially generates imputation-derived multimodal fusion features. However, we developed a full-information maximum likelihood-based multimodal fusion method to handle general missing problems in multimodal fusion. We applied the proposed methods to data from the Alzheimer’s Disease Neuroimaging Initiative and identified the longitudinal profiles of morphological brain changes and amyloid cumulation. And here’s my short bio: Dr. Lee is an Associate Professor of Clinical biostatistics in Psychiatry at Columbia University and New York State Psychiatric Institute. Her research primarily focuses on developing statistical methods in neuroimaging data, including multimodal fusion, time-frequency data analysis, latent variable analysis, and mediation analysis. Dr. Lee received her Ph.D. degree in Statistics and Operations Research at the University of North Carolina at Chapel Hill and completed her postdoctoral training in Biostatistics and Biomedical engineering under the supervision of Drs. Dzung L. Pham at NIH and Brian S. Caffo at Johns Hopkins University. She joined Mental Health Data Science at NYSPI and the Department of Biostatistics and Psychiatry at Columbia University in 2013 as a faculty. |