![\"\"](\"https:\/\/cdn.vanderbilt.edu\/vu-URL\/wp-content\/uploads\/sites\/193\/2019\/07\/22141428\/Hao-LI-180x300.jpg\")
<\/p>\nVISE welcomes Hao Li, PhD, to our Fall Seminar Series.<\/p>\n
<\/p>\n
Date:\u00a0<\/strong>Thursday, November 6, 2025 Title: Abstract:<\/strong><\/p>\n Recent advances in deep learning have significantly advanced medical image segmentation, evolving from data-driven to data-centric paradigms. This transition highlights the growing importance of data quality in achieving reliable and efficient segmentation. Data-driven strategies focus on enhancing neural network (NN) architectures and loss functions using existing datasets, whereas data-centric methods emphasize improving data quality, consistency, and diversity to strengthen model performance and generalizability. This talk covers both perspectives. On the data-driven side, state-of-the-art neural networks are developed for robust segmentation with supervised learning across diverse medical applications, enabling the effective extraction of clinically relevant biomarkers. On the data-centric side, domain shifts are addressed through unsupervised domain adaptation and test-time adaptation to improve data harmonization across institutions, scanners, and patient populations. Incorporating expert knowledge further enhances interpretability and reliability under heterogeneous supervision, while uncertainty-guided semi-supervised learning tackles the common challenge of limited data availability to achieve more accurate and generalizable segmentation performance. Evaluations across diverse imaging modalities, including multi-center MRI, CT, ultrasound, and endoscopic datasets, demonstrate that deep learning\u2013based segmentation methods achieve strong performance, particularly when efforts focus on improving data fidelity and consistency. This transition from data-driven to data-centric medical image segmentation establishes a sustainable direction for developing generalizable, trustworthy, and clinically deployable algorithms by aligning methodological innovation with the realities of real-world medical data.<\/p>\n Bio:<\/strong> \u00a0<\/strong><\/p>\n","protected":false},"excerpt":{"rendered":" VISE welcomes Hao Li, PhD, to our Fall Seminar Series. Hao Li, PhD Research Assistant Professor Department of Computer Science 91ºÚÁÏÍø Date:\u00a0Thursday, November 6, 2025 Location:\u00a0SVC 5326 Time:\u00a011:45 am for lunch,12:00 start Title: From Data-driven to Data-centric Medical Image Segmentation Abstract: Recent advances in deep learning have significantly advanced medical image segmentation, evolving from…<\/p>\n","protected":false},"author":670,"featured_media":11337,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"spay_email":"","jetpack_publicize_message":"","jetpack_is_tweetstorm":false,"jetpack_publicize_feature_enabled":true,"_links_to":"","_links_to_target":""},"categories":[12],"tags":[],"acf":[],"jetpack_featured_media_url":"https:\/\/cdn.vanderbilt.edu\/vu-URL\/wp-content\/uploads\/sites\/193\/2019\/07\/22141428\/Hao-LI-scaled.jpg","jetpack_publicize_connections":[],"jetpack_sharing_enabled":false,"jetpack_shortlink":"https:\/\/wp.me\/p98pzF-31u","_links":{"self":[{"href":"https:\/\/www.vanderbilt.edu\/vise\/wp-json\/wp\/v2\/posts\/11624"}],"collection":[{"href":"https:\/\/www.vanderbilt.edu\/vise\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.vanderbilt.edu\/vise\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.vanderbilt.edu\/vise\/wp-json\/wp\/v2\/users\/670"}],"replies":[{"embeddable":true,"href":"https:\/\/www.vanderbilt.edu\/vise\/wp-json\/wp\/v2\/comments?post=11624"}],"version-history":[{"count":1,"href":"https:\/\/www.vanderbilt.edu\/vise\/wp-json\/wp\/v2\/posts\/11624\/revisions"}],"predecessor-version":[{"id":11625,"href":"https:\/\/www.vanderbilt.edu\/vise\/wp-json\/wp\/v2\/posts\/11624\/revisions\/11625"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.vanderbilt.edu\/vise\/wp-json\/wp\/v2\/media\/11337"}],"wp:attachment":[{"href":"https:\/\/www.vanderbilt.edu\/vise\/wp-json\/wp\/v2\/media?parent=11624"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.vanderbilt.edu\/vise\/wp-json\/wp\/v2\/categories?post=11624"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.vanderbilt.edu\/vise\/wp-json\/wp\/v2\/tags?post=11624"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
\nLocation:\u00a0<\/strong>SVC 5326
\nTime:\u00a0<\/strong>11:45 am for lunch,12:00 start<\/p>\n
\n<\/strong>From Data-driven to Data-centric Medical Image Segmentation<\/p>\n
\nDr. Hao Li\u00a0is a Research Assistant Professor in the Department of Computer Science at 91ºÚÁÏÍø. His research focuses broadly on algorithmic development for medical image analysis across both medical image computing and computer-aided intervention. He develops methods in multi-modal learning, unsupervised domain adaptation, and uncertainty-guided semi-supervised learning. with\u00a0applications to segmentation, registration, and detection across various imaging modalities such as MRI, CT, ultrasound and endoscopy in both 2D and 3D domains. His research aims to build robust and efficient solutions for real-world clinical and surgical applications, advancing disease assessment, diagnostic decision support, preoperative analysis, and real-time image-guided procedures.<\/p>\n