Infos Jean-Benoit Delbrouck
Email : jbdel at stanford dot edu
Email : Jean-Benoit.delbrouck at umons dot ac dot be
Adress : Address: 1701 Page Mill Rd, Palo Alto, CA 94304, United States
Blurb My research lies at the intersection of vision and language in medical AI. I'm interested in using paired medical images and radiology reports to improve the performances of trained models on a wide range of benchmarks. More specifically, I focus on NLP techniques to extract rich semantic features from the reports to combine them with visual representations. I have had the chance to publish my top-tier AI conferences such as ACL, EMNLP and ICLR.
News - We are organizing the BioNLP workshop at ACL 2023, check it out!
https://vilmedic.app/misc/bionlp23/sharedtask/
https://vilmedic.app/misc/bionlp23/leaderboard/
-We generate chest x-rays using stable diffusion, check it out
2022 recap Early 2022, we got our medical AI library ViLMedic accepted to ACL. ViLMedic offers dozens of pretrained models at the intersection of vision and language in the medical field. ViLMedic was leveraged to evaluate Domino, a systematic error discovery method, accepted as Spotlight at ICLR 2022. A new paper in Radiology Report Generation was accepted at EMNLP 2022 where we leverage graph-based semantic radiologists annotations to evaluate the factual correctness of AI-generated radiology reports. We released a new Radiology report summarization dataset of 70k samples with dozens of new modality-anatomy pairs, accepted at ML4H. Last but not least, we fine-tuned stable diffusion on chest x-rays. Check it out, its amazing :)
Publications 2023 2022 2021 2020 2019 2018 2017
Teaching Lecturer Mentoring
Groups: Individuals: Jurys