IEEE–OC EMBS Chapter
Meeting Date: July 29, 2019
Time: 6:30 PM Networking & Food; 7:00 PM Presentation
Speaker: Peter D. Chang, MD of Co-Director, Center for Artificial Intelligence in Diagnostic Medicine UCI Health | Dept. of Radiologic Sci.
RSVP: requested, through website
Event Details: IEEE vTools
Summary: Despite the growing popularity of deep learning neural networks for various medical imaging applications, the vast majority of algorithms to date represent early proof-of-concept designs that will require a degree of evolution before achieving practical clinical utility. In this talk, we explore several different ways in which the current generation of deep learning applications can be advanced including:
(1) Reformulating the question: in medicine, there are often times more than one way to ask the same question—how do we reformulate a task in a way that both maximizes clinical utility and also best leverages the strength of various deep learning algorithms?
(2) Customizing deep learning algorithms: what are the unique technical challenges posed by medical imaging data and how we design custom deep learning architectures to account for them?
(3) Clinical implementation: what are some practical experiences learned from implementing deep learning tools in the clinical setting, and what are some regulatory hurdles that will need to be considered?
Bio: Dr.Chang’s unique perspective as both a physician and software engineer allows him to lead a team of engineers, data scientists, and clinicians to combine state-of-the-art techniques in machine learning with an intuitive understanding of the image interpretation process as a human radiologist. His work has yielded many collaborative efforts with principal investigators in multiple clinical specialties as well as many imaging departments across the United States and leadership roles in national and international organizations such as the American Society of Neuroradiology, Radiological Society of North America, and the International Society for Magnetic Resonance in Medicine. His software algorithms have been granted provisional patents and won multiple awards, including a first-place finish at the international 2016 MICCAI grand challenge for automated analysis of brain tumors. He has pursued a dedicated year of research in deep learning at the University of California San Francisco.
His primary insights arise from the ability to customize deep learning algorithms to the unique properties and goals of medical imaging data. He has extensive experience with the fundamental tools of deep learning including Python and Tensorflow, and has successfully written customized software for distributed file systems and GPU servers to process large-scale datasets including one million mammograms, 50,000 chest ICU films and 20,000 head CTs. In addition, he has broad expertise with the entire medical image processing pipeline including PACS networking and DICOM manipulation to efficiently generate large anonymized databases. He has integrated these databases into a customized web-based platform for efficient viewing and fast, large-scale annotation of medical images. Recently, his work has included the development of a novel hybrid 3D/2D architecture for detection of hemorrhage on non-contrast head CT, recording human-like performance at over 97% accuracy in a cohort of over 10,000 CTs ; this work was received the Cornelius G. Dyke Award for best scientific paper by an assistant professor (or below) in AJNR during 2017-18.