PhD Students Li and Yang Selected for AAPM Early-Career Investigator Symposium

PhD students Xiang Li and Zhenyu Yang have each been selected to present their work in the prestigious John R. Cameron Early-Career Investigator Symposium at the American Association of Physicists in Medicine (AAPM) Annual Meeting and Exhibition, held July 10-14, 2022, in Washington, DC.

The Symposium is an annual competition for young investigators in honor of University of Wisconsin professor emeritus and founding AAPM member John Cameron, PhD. Abstract selection for the Symposium is a significant honor. Xiang and Zhenyu’s abstracts were two of just ten selected for the Symposium from a pool of nearly 400 abstract submissions – representing the top 2.5% of abstracts. This is a remarkable accomplishment for both students, and represents both the quality and cutting-edge nature of research at Duke.

Xiang Li

Xiang Li is a PhD student in the Lafata Lab. She is working towards her PhD in electrical and computer engineering at the Pratt School of Engineering. Xiang’s abstract is titled "Computational mapping of lymphocytic topology on digital pathology images with single cell resolution immunohistochemistry validation." In this work, Xiang and others developed a deep learning pipeline to detect lymphocytes and quantify their topological interactions on digital whole slide pathology images. Unlike other deep learning pipelines that rely on reader annotations of ground truth of lymphocytes on hematoxylin and eosin (H&E) images – which are time consuming and prone to inter-observer variability – this approach automatically generates large-scale training labels based on fused immunohistochemistry (IHC). To enable this, they have developed a novel image acquisition and registration technique, whereby deep learning can be trained and validated with IHC at single cell resolution. As a result, this deep learning pipeline can mimic IHC to predict lymphocytic topology on H&E images, which can be further used to quantify immune microenvironment on pathology images.

Xiang is mentored by Assistant Professor of Radiology, Radiation Oncology, and Electrical and Computer Engineering Kyle Lafata, PhD. Their research focuses on the theory and application of multi-scale computational biomarkers. Dr. Lafata said, "New approaches to computationally interrogate the immune microenvironment on digitized tissue samples will help bridge a knowledge gap between abstract imaging features and basic biology. Xiang’s work has diverse biomedical applications, from modeling immune response of cancer treatments, to detecting antibody mediated rejection of transplanted organs."

Zhenyu Yang

Zhenyu Yang is a PhD candidate in medical physics (radiation therapy track) at the Duke Medical Physics Graduate Program. Zhenyu’s abstract, "A Neural ODE Model for Visualizing Deep Neural Network Behaviors in Multi-Parametric MRI Based Glioma Segmentation," explores the development of a novel neural ordinary differential equation (ODE) model to visualize the deep neural network behavior during multi-parametric MRI based glioma segmentation. The model successfully visualizes the dynamics of both MR images after interactions with the deep neural network; and the segmentation formation process.

Zhenyu is mentored by Assistant Professor of Radiation Oncology Chunhao Wang, PhD, who said, "The integration of the recent neural differential equation theories in this work establishes a new approach of deep learning explainability."