날짜 : 2022년 9월 22일 목요일 14시 30분
장소 : 제 1공학관 A542호

제목: AI in Medical Image Analysis: Opportunities and Challenges

강사: 우종혜 교수님 (Harvard Medical School)

Abstract:

Over the last several years, machine/deep learning has achieved unprecedented success and state-of-the-art performances in a variety of medical image analysis tasks, such as image segmentation, inpainting, synthesis, etc. In addition, recent deep learning approaches with medical imaging data performed on par with or better than expert clinicians in some cases, thereby demonstrating the potential to aid clinicians in the diagnosis or decision-making process. Yet, a swath of challenges—and opportunities—arise in developing new models and deploying the models for clinical purposes, such as limited interpretability and difficulty in generalization, due to domain shifts, between source and target domains. In the first half of this talk, I will present brief background and current trends in deep learning for medical image analysis alongside the evolution of efforts in this rapidly growing area within my lab that carry promise for overcoming some of the limitations for clinical applications. In the second half of this talk, I will specifically focus on efforts to combat domain shifts via unsupervised domain adaptation in clinically meaningful settings, when deploying the models in the target domains.