Ke YAN
Alibaba DAMO Academy, Beijing, China
19th April 2022, 4:00pm - 5:00pm (GST)
Title: | DeepLesion: Large-scale Universal Lesion Analysis in Medical Imaging based on Deep Learning |
Abstract: | Lesion analysis in medical images is a key part of radiologists' daily work. All abnormalities in an image need to be found, measured, described, and compared with previous scans. It is a time-consuming task requiring highly specialized knowledge, meanwhile prone to errors due to subtlety of lesions, heavy workload, lack of experience, and inter-observer variance. We aim to develop a series of machine learning algorithms to help doctors reduce time cost and improve accuracy. Specifically, we propose universal lesion detection, classification, measurement, and matching algorithms that can analyze lesions in a variety of organs in computed tomography (CT) images. They are based on our collected large-scale lesion dataset, DeepLesion, which has been released in 2018 and free to download. Our works involve latest deep learning algorithms such as object detection, attention-based instance segmentation, key point detection, self-supervised learning, and classification based on label relations. |
Bio: | Ke Yan is a Senior Medical Algorithm Expert in the Alibaba DAMO Academy, Beijing, China. He graduated from the Department of Electronic Engineering, Tsinghua University and obtained his PhD degree. Then, he worked as a postdoctoral fellow in the Imaging Biomarkers and Computer-Aided Diagnosis Laboratory, National Institute of Health, US. His mentors were Dr. Ronald Summers and Dr. Le Lu. His research mainly focuses on medical image analysis using machine learning, especially on lesion detection, classification, retrieval, and matching in CT images using deep learning. He published the DeepLesion dataset, a large-scale and universal CT lesion dataset. He also won the RSNA Trainee Research Prize in 2018 and Tsinghua University Excellent Doctoral Dissertation Award in 2016. He has published papers and abstracts on IEEE Transactions on Medical Imaging, Radiology: Artificial Intelligence, npj Digital Medicine, CVPR, MICCAI, RSNA, etc., and got >1900 citations. Picture is attached. |