John Kornak, Professor
Department of Epidemiology and Biostatistics, University of California, San Francisco
Bayesian image analysis enhances image quality and critical clinical features by integrating prior knowledge with probabilistic models. Bayesian Image Analysis in Transformed Space (BITS) leverages transformed domains for efficient, parallelized analysis, enabling a broader range of prior structures and expanding potential applications. Bayesian Image Analysis in Fourier Space (BIFS) exploits frequency-space properties for structured priors that can capture important image properties beyond conventional Bayesian image analysis, while Data-Driven BIFS (DD-BIFS) learns priors empirically for greater adaptability. Bayesian Image Analysis in Wavelet Space (BIWS) captures multi-scale structures and textures, effectively modeling non-stationary image features. By improving computational efficiency and expanding the scope of Bayesian image analysis, BITS advances medical imaging quality and analysis, demonstrated in arterial spin labeling (ASL) and functional MRI (fMRI).