VALSE Webinar 22-05期总第271期基于领域知识的机器学习在医学影像分析中的应用

2022 年 3 月 10 日 VALSE

报告时间	2022年03月16日 (星期三) 晚上20:00 (北京时间)
主题	基于领域知识的机器学习在医学影像分析中的应用
主持人	于乐全 (香港大学) 窦琪 (香港中文大学)
直播地址	https://live.bilibili.com/22300737

报告嘉宾：王连生 (厦门大学)

报告题目：一种放射影像预训练方法

报告嘉宾：崔智铭 (香港大学)

报告题目：Shape-aware tooth segmentation in digital dentistry

报告嘉宾：秦璟 (香港理工大学)

报告题目：Rethink Deep Learning Models for Medical Image Analysis from an Epistemological Perspective

Panel嘉宾：

王连生 (厦门大学)、崔智铭 (香港大学)、秦璟 (香港理工大学)、周郁音 (UCSC)、于乐全 (香港大学)

Panel议题：

1. 在医学图像处理中，考虑领域知识的主要目的是什么？会对医学图像处理产生什么样的影响？

2. 在医学图像处理中，所用到的领域知识包括哪些方面？这些领域知识是容易获得吗？会有病人隐私等问题考虑吗？

3. 由于医学场景下领域知识往往很零散，使得其与深度学习的结合有一定困难。目前领域知识与深度学习这两者结合的难点和痛点在哪里？

4. 领域知识在医学图像处理中的重要性体现在哪些方面？现有的方法可以很好的学习到想要的领域知识吗？未来还可以在哪些方面做改进？未来的发展趋势怎样吗？

5. 现有的实际AI医疗应用能够满足工业界的实际需求吗？这些应用是否是基于医学场景领域知识来研发设计？

*欢迎大家在下方留言提出主题相关问题，主持人和panel嘉宾会从中选择若干热度高的问题加入panel议题！

报告嘉宾：王连生 (厦门大学)

报告时间：2022年03月16日 (星期三)晚上20:00 (北京时间)

报告题目：一种放射影像预训练方法

报告人简介：

王连生，毕业于香港中文大学，现为厦门大学计算机系副教授，医学院消化病学系双聘教授，博士生导师，数字福建健康医疗大数据研究所副所长。长期从事人工智能医疗研究，主持和参与20余项科研项目，包括国家自然科学基金仪器专项、国家重点研发项目、国家自然科学基金面上和青年项目等，获得腾讯犀牛鸟科研奖、CSPE Young Investigator、福建省科技进步奖和自然科学奖，带领团队先后9次在国际医学影像比赛中获得冠军。近年来以一作或者通讯作者完成的相关成果发表于Nature Machine Intelligence、人工智能顶会议如CVPR、AAAI、IJCAI和医学图像处理权威期刊如IEEE TMI、MedIA等。

个人主页：

https://xmu-lswang.github.io

报告摘要：

近些年来，基于深度学习的方法在放射影像分析中取得了很好的效果，而预训练是取得这些成果的基础。预训练方法期望获得迁移性更好的图像表征，以往的预训练通过在源域上进行大规模全监督或者自监督学习来达到目的。但是，全监督的预训练需要图像的标签，制作标签是一个费时费力的工程，而自监督学习目前与全监督还有一定的差距。为了解决这个重要问题，提出一种交叉监督的方法，该方法使用放射影像自带的报告来作为监督信息进行预训练，也可以同时处理患者的多张放射影像，从而获得更加全面的信息。

参考文献：

[1] Zhou, Hong-Yu, Xiaoyu Chen, Yinghao Zhang, Ruibang Luo, Liansheng Wang, and Yizhou Yu. "Generalized radiograph representation learning via cross-supervision between images and free-text radiology reports." Nature Machine Intelligence (2022): 1-9.

报告嘉宾：崔智铭 (香港大学)

报告时间：2022年03月16日 (星期三)晚上20:20 (北京时间)

报告题目：Shape-aware tooth segmentation in digital dentistry

报告人简介：

Zhiming Cui is now a Ph.D. candidate in the computer vision and graphics lab at the University of Hong Kong, supervised by Professor Wenping Wang. Before that, he got his B.Eng and M.Eng degrees from Northeastern University in 2014 and 2017, respectively. During his Ph.D. study, he has visited UII and ShanghaiTech University, advised by Professor Dinggang Shen. His interests include medical image analysis (segmentation, domain adaptation/generalization, and AI in orthodontics) and 3D vision. His work has been published on several top journals and conferences, including Nature Communications, CVPR, ECCV, MICCAI, ISBI, TMI, MedIA, JBHI, and CAGD, etc.

个人主页：

https://erdanc.github.io/

报告摘要：

With the development of deep learning in medical imaging, how to learn the anatomical structure knowledge to guide specific tasks (i.e., segmentation and detection) is an essential yet challenging problem. This talk will present some of our recent works about shape-aware tooth segmentation in digital dentistry, including 3D CBCT and intra-oral scanning data. We will also discuss how to effectively define and learn anatomical knowledge in medical imaging.

参考文献：

[1] Cui, Zhiming, Changjian Li, and Wenping Wang. "ToothNet: automatic tooth instance segmentation and identification from cone beam CT images." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019.

[2] Cui, Zhiming, et al. "TSegNet: an efficient and accurate tooth segmentation network on 3D dental model." Medical Image Analysis 69 (2021): 101949.

[3] Cui, Zhiming, et al. "Hierarchical Morphology-Guided Tooth Instance Segmentation from CBCT Images." International Conference on Information Processing in Medical Imaging. Springer, Cham, 2021.

[4] Cui Zhiming et al. “A fully automatic AI system for tooth and alveolar bone segmentation from cone-beam CT images.” Nature Communications. 2022 (accepted for published), 2022.

报告嘉宾：秦璟 (香港理工大学)

报告时间：2022年03月16日 (星期三)晚上20:40 (北京时间)

报告题目：Rethink Deep Learning Models for Medical Image Analysis from an Epistemological Perspective

报告人简介：

QIN, Jing (Harry)is currently an associate professor in Centre for Smart Health, School of Nursing, The Hong Kong Polytechnic University. His research focuses on creatively leveraging advanced virtual reality (VR)and artificial intelligence (AI)techniques in healthcare and medicine applications and his achievements in relevant areas has been well recognized by the academic community. He won the Hong Kong Medical and Health Device Industries Association Student Research Award for his PhD study on VR-based simulation systems for surgical training and planning. He won 5 best paper awards for his research on AI-driven medical image analysis and computer-assisted surgery. He served as a local organization chair for MICCAI 2019, technical program committee (TPC)members for many academic conferences, speakers for many invited talks, and referees for many prestigious journals in relevant fields.

个人主页：

https://harry-qinjing.github.io/index.html

报告摘要：

Our group has developed dozens of deep learning models for various medical image analysis tasks. Some of them won prestigious best paper awards while some others are highly cited and widely used in subsequent research works. However, we still face a lot of challenges that prohibit these models to be widely and effectively used in clinical practice. In this talk, we shall re-discuss these models from an epistemological perspective, aiming at understanding these interconnected challenges more clearly and paving the way in the intervening haze for developing clinically applicable models.

Panel嘉宾：Yuyin Zhou (University of California, Santa Cruz)

嘉宾简介：

Dr. Yuyin Zhou is currently an Assistant Professor of Computer Science and Engineering at UC Santa Cruz. She received her Ph.D. from the Computer Science Department at Johns Hopkins University in 2020 and was a postdoctoral researcher at Stanford University from 2020 to 2021. Yuyin’s research interests span the fields of medical image computing, computer vision, and machine learning, especially the intersection of them. She has published many papers at top-tier conferences and journals including CVPR, ICCV, AAAI, TPAMI, TMI, MedIA, etc. Yuyin Zhou has led the ICML 2021 workshop on Interpretable Machine Learning in Healthcare, the ICCV 2021 workshop on Computer Vision for Automated Medical Diagnosis, and co-organized ML4H 2021, the 9th CVPR MCV workshop. She served as a senior program committee for IJCAI 2021 and AAAI 2022, an area chair for MICCAI 2022, CHIL 2022.

个人主页：

https://yuyinzhou.github.io/

报告主持人：于乐全 (香港大学)

主持人简介：

Dr. Lequan Yu is an Assistant Professor at the Department of Statistics and Actuarial Science, the University of Hong Kong. Before joining HKU, he was a postdoctoral fellow at Stanford University. He obtained his Ph.D. degree from The Chinese University of Hong Kong and bachelor’s degree from Zhejiang University, both in Computer Science. He also experienced research internships in Nvidia and Siemens Healthineers. His research interests are developing advanced machine learning methods for biomedical data analysis, with a primary focus on medical images. He has won the CUHK Young Scholars Thesis Award 2019 and has also won the Best Paper Awards of Medical Image Analysis-MICCAI in 2017 and International Workshop on Machine Learning in Medical Imaging in MICCAI 2017. He served as area chair of MICCAI 2022, senior PC member of IJCAI, AAAI, and the reviewer for top-tier journals and conferences, such as Nature Machine Intelligence, IEEE-PAMI, IEEE-TMI, Medical Image Analysis, etc.

个人主页：

https://yulequan.github.io/

Panel主持人：窦琪 (香港中文大学)

主持人简介：

窦琪，香港中文大学计算机科学与工程系助理教授，研究方向为机器学习及其在医学影像分析和手术机器人领域的应用。曾获IEEE ICRA 医疗机器人最佳论文奖、Medical Image Analysis - MICCAI 最佳论文奖、MICCAI Young Scientist Award Finalist、香港科学会青年科学家奖、ISBI LUNA 2016肺结节检测挑战赛冠军、蝉联MICCAI 手术流程识别挑战赛2016&2019冠军、蝉联MICCAI多模态手术机器人动作识别挑战赛2020&2021冠军等。担任多个领域内顶级会议Program Co-Chair和Area Chair, 以及期刊副主编。多次在NeurIPS, CVPR, ICCV, ICML上组织医学影像相关Workshops。

个人主页：

http://www.cse.cuhk.edu.hk/~qdou/

特别鸣谢本次Webinar主要组织者：

主办AC：窦琪 (香港中文大学)、于乐全 (香港大学)

活动参与方式

1、VALSE每周举行的Webinar活动依托B站直播平台进行，欢迎在B站搜索VALSE_Webinar关注我们！

直播地址：

https://live.bilibili.com/22300737；

历史视频观看地址：

https://space.bilibili.com/562085182/

2、VALSE Webinar活动通常每周三晚上20:00进行，但偶尔会因为讲者时区问题略有调整，为方便您参加活动，请关注VALSE微信公众号：valse_wechat 或加入VALSE QQ R群，群号：137634472）；

*注：申请加入VALSE QQ群时需验证姓名、单位和身份，缺一不可。入群后，请实名，姓名身份单位。身份：学校及科研单位人员T；企业研发I；博士D；硕士M。