Publication Date

Fall 2023

Degree Type


Degree Name

Master of Science (MS)


Computer Engineering


Magdalini Eirinaki; Harry Li; Alessandro Bellofiore


Enhancing medical imaging stroke diagnosis applications with artificial intelligence (AI) tools to determine lesion volume, location and clinical metadata is vital toward guiding patient treatment and procedure. A major hardship in developing stroke diagnosis AI tools is the scarcity of publicly available clinical 3D stroke datasets. Through working with Johns Hopkins University, University of Michigan’s ICPSR data repository and SJSU research, we gained access to potentially the largest 3D MRI stroke dataset with clinical metadata annotated by neuroradiologists known as ICPSR 38464. With the ICPSR 38464 dataset recently being available through institutional review board (IRB) approval or exemption, we were potentially one of the first few teams to design and develop computer vision (CV) and natural language processing (NLP) deep learning (DL) models based on it. We mainly referred to Johns Hopkins University’s superb results for ischemic stroke lesion segmentation based on their 3D DAGMNet model, designing and developing our own Attention Squeeze Excitation (SE) 3D UNet model for general stroke lesion segmentation to account for not only ischemic, but also hemorrhagic stroke lesion segmentation. We then leveraged our 3D extracted features to design and develop a 3D CNN To LSTM model for stroke MRI captioning. This effort mapping the CV to NLP side is critical in AI for stroke diagnosis systems because it helps neuroradiologists by not only automatically tracing the lesions for them, but also generating captions of medical history and prior medication, which is the foundation to an even more time consuming task known as clinical report generation. In our thesis, we are potentially one of the first teams to set this foundation and do it with ICPSR 38464 to help neuroradiologists diagnose stroke MRI patients, ultimately saving the doctors time and allowing for more patients to be helped for early stroke intervention.