Kelvin Wong is the Director of Neuroimaging at TT and WF Chao Center for BRAIN for neurological disorders. His laboratory specialized in using deep learning in medical imaging derived features in addition to clinical features from electronic medical records for outcome prediction. A decade ago, he applied machine learning method to solve the diagnostic dilemma in glioblastoma patients undergoing chemoradiation therapy to differentiate the treatment induced pseudoprogression from tumor progression using multiparametric imaging features. Few years ago, we pushed the AI boundary and use it for interpretable biomarker discovery on glioblastoma gene expression data. We discovered two novel prognostic genes, Aquaporin-1 (AQP1) and Mesenchyme Homeobox 2 (MEOX2) that were not well known to be associated with glioblastoma or low grade glioma survival and out results are subsequently confirmed by followup studies by other groups. Recently, we focus on using facial video and audio to identifying subtle stroke in the emergency room and use AI denoising to detect stroke early on in CT.