Short Course 2: Data Science and AI Models in Precision Cancer Research

May 15, 2025, 1:00pm-5:00pm PDT

Stanford Campus, Always M121Q

Instructor: Lei Xing, Stanford University; Kun-Sing Yu, Harvard University; Ruogu Fang, University of Florida, and Michael Gensheimer, Stanford University.

Lei Xing, PhD, the Jacob Haimson & Sarah S. Donaldson Professor of Medical Physics and Electrical Engineering (by courtesy) at Stanford University. He is also the Director of the Medical Physics Division of Radiation Oncology Department. His research is focused on AI in medicine, medical imaging, and biomedical physics. He has made unique and significant contributions to each of the above areas. Dr. Xing is an author on more than 450 peer reviewed publications, a co-inventor on many issued and pending patents, and a co- investigator or PI on numerous NIH, NSF, DOD, RSNA, ACS and many corporate grants. He is a fellow of AAPM (American Association of Physicists in Medicine), ASTRO (American Society for Radiation Oncology), and AIMBE (American Institute for Medical and Biological Engineering). He has received numerous awards, such as Google Faculty Scholar Award and E. H. Quimby Lifetime Achievement Awards from AAPM.

"Kun-Hsing "Kun" Yu, MD, PhD, is an Assistant Professor of Biomedical Informatics, Harvard Medical School. He developed the first fully automated artificial intelligence (AI) algorithm to extract thousands of features from whole-slide histopathology images, discovered the molecular mechanisms underpinning the microscopic phenotypes of tumor cells, and successfully identified previously unknown cellular morphologies associated with patient prognosis. His lab integrates cancer patients' multi-omics (genomics, epigenomics, transcriptomics, and proteomics) profiles with quantitative histopathology patterns to predict their clinical phenotypes. More than 30 research laboratories worldwide have independently validated the AI methods developed in the Yu Lab. Dr. Yu's work in pathology AI has received several accolades, including the American Medical Informatics Association New Investigator Award, the Department of Defense Career Development Award, the Google Research Scholar Award, and the National Institutes of Health (NIH) Maximizing Investigators' Research Award. He is a Fellow of the American Medical Informatics Association.

Ruogu Fang, PhD, tenured Associate Professor and Pruitt Family Endowed Faculty Fellow in the J. Crayton Pruitt Family Department of Biomedical Engineering at the University of Florida. Her research encompasses two principal themes: AI-empowered precision brain health and brain/bio-inspired AI. Her work involves addressing compelling questions, such as using machine learning techniques to quantify brain dynamics, facilitating early Alzheimer's disease diagnosis through novel imagery, predicting personalized treatment outcomes, designing precision interventions, and leveraging principles from neuroscience to develop the next-generation of AI. Fang's current research is also rooted in the confluence of AI and multimodal medical image analysis. She is the PI of an NIH NIA RF1 (R01-equivalent), an NSF Research Initiation Initiative (CRII) Award, an NSF CISE IIS Award, a Ralph Lowe Junior Faculty Enhancement Award from Oak Ridge Associated Universities (ORAU). She has also received numerous recognitions. She was selected as the Rising Stars (Engineering) by the Academy of Science, Engineering, and Medicine of Florida (ASEMFL), the inaugural recipient of the Robin Sidhu Memorial Young Scientist Award from the Society of Brain Mapping and Therapeutics, an Best Paper Award from the IEEE International Conference on Image Processing , the University of Florida AI Course Award, an UF Herbert Wertheim College of Engineering Faculty Award for Excellence in Innovation, an UF BME Faculty Research Excellence Award and Faculty Teaching Excellence Award, among others. Fang's research has been featured by Forbes Magazine, The Washington Post, ABC, RSNA, and published in Lancet Digital Health, JAMA, PNAS and npj Digital Medicine. She is an Associate Editor of the Journal Medical Image Analysis, a Topic Editor of Frontiers in Human Neuroscience, and a Guest Editor of CMIG. She is a reviewer for The Lancet, Nature Machine Intelligence, Science Advances, etc. Her research has been supported by NSF, NIH, Oak Ridge Laboratory, DHS, DoD, NVIDIA, and the University of Florida. She is President of Women in MICCAI (WiM) and Associate Editor of the Medical Image Analysis journal. At the heart of her work is the Smart Medical Informatics Learning and Evaluation (SMILE) lab, where she is tirelessly dedicated to creating groundbreaking brain and neuroscience-inspired medical AI and deep learning models. The primary objective of these models is to comprehend, diagnose, and treat brain disorders, all while navigating the complexities of extensive and intricate datasets.

Michael Gensheimer, MD, Clinical Associate Professor of Radiation Oncology, Stanford University. He is a radiation oncologist whose clinical practice focuses on treatment of head and neck cancer. Much of his research involves analysis of large-scale electronic medical record datasets to predict patient outcomes and pick the most effective treatments. His predictive models have been deployed at Stanford and tested in randomized studies. His open source nnet-survival software that allows use of neural networks for survival modeling has been used by researchers internationally.

Abstract

Data science and AI models have gained popularity in medical research. This short course provides an introduction and demonstration of these advanced models in cancer research. We will start with an introduction of foundation models and how they will be related to individual medicine. An imaging foundation models for neurodegenerative disease will be presented to illustrate the development and application of foundation models in medicine. We will than explain the use of large language models in precision oncology and oncology practice through examples. Finally, we will present the cutting-edge research in a multimodality AI for next generation of biomedicine provides power tools but also provides a lot of opportunities in research and applications in clinical oncology.

Short Course 3: Bayesian Designs of Clinical Trials Using Historical Data: From Theory to Practice

May 18, 2025, 8:30am - 5:00pm, May 18, 2025

Stanford Campus, LKSC 120

Instructor: Ming-Hui Chen, Min Lin, and Yanyan Zhu, University of Connecticut

Dr. Ming-Hui Chen is a Board of Trustees Distinguished Professor and Head of Department of Statistics at University of Connecticut (UConn). He was elected to Fellow of American Association for the Advancement of Science (AAAS) in 2024, Fellow of International Society for Bayesian Analysis in 2016, Fellow of Institute of Mathematical Statistics in 2007, and Fellow of American Statistical Association in 2005. He received the University of Connecticut AAUP Research Excellence Award in 2013, the UConn College of Liberal Arts and Sciences (CLAS) Excellence in Research Award in the Physical Sciences Division in 2013, the University of Connecticut Alumni Association's University Award for Faculty Excellence in Research and Creativity (Sciences) in 2014, the ICSA Distinguished Achievement Award in 2020, and the Distinguished Science Alumni Award from Purdue University in 2023. He has published 460+ peer-reviewed journal articles and five books including two advanced graduate-level books on Bayesian survival analysis and Monte Carlo methods in Bayesian computation. He has supervised 42 PhD students. He served as, President of ICSA (2013), Chair of the Eastern Asia Chapter of International Society for Bayesian Analysis (2018), President of New England Statistical Society (2018-2020), and the 2022 JSM Program Chair. Currently, he is Co Editor-in-Chief of Statistics and Its Interface, inaugurated Co Editor-in-Chief of New England Journal of Statistics in Data Science, and an Associate Editor for several other statistical journals.

Min Lin is a fourth-year PhD student in Statistics at the University of Connecticut and a Research Fellow at Servier Pharmaceuticals. His research at Servier focuses on developing propensity-score-integrated Bayesian approaches for leveraging external control data. More broadly, his research interests include Bayesian clinical trial design and the development of new methods for effective sample size calculation to quantify the benefit-risk trade-off of external data borrowing. Additionally, he collaborates with the Connecticut Agricultural Experiment Station, providing statistical insights into invasive aquatic plant management at Candlewood Lake, with a focus on the effects of grass carp and winter drawdowns.

Yanyan Zhu is a fourth-year PhD student in Statistics at the University of Connecticut. She is holding a Vertex Fellowship, leading research projects on joint models of longitudinal and survival data. She is an active member of the Statistics in Pediatric Drug Development – Extrapolation Sub-team. Her research interests also include Bayesian design of clinical trials.

Abstract

Bayesian sample size determination (SSD) has a long history. The early work on Bayesian SSD can be traced back to 1990’s. Recently, several new methods on Bayesian designs of clinical trials has been developed with a focus on controlling type I error and power. In this short course, an overview of the literature on Bayesian SSD will be provided. The general theory and various methods of Bayesian SSD will be presented. The short course will also highlight several important applications in designing clinical trials to demonstrate the superiority of Bayesian SSD.

This short course starts with a brief review of early development of Bayesian SSD. Then, a comprehensive overview of the general framework of the Bayesian design will follow. Next, an in-depth exposition and discussion of the Bayesian design of a noninferiority trial for a medical device will be provided.

A key challenge in Bayesian SSD is determining the minimum required sample size to achieve a prespecified frequentist power, as Bayesian power calculations often depend on computationally intensive Monte Carlo sampling. We discuss methods to reduce computational burden and illustrate the methodology using a dedicated R package currently under development.

Pediatric extrapolation leverages existing data, often from adult or older pediatric populations to support drug development in children. This part will introduce a Bayesian framework that ensures the exact type I error control in the absence of borrowing. To determine the appropriate level of borrowing from the historical adult data, an analytical procedure will be presented to respect a pre-specified Type I error inflation threshold. The methodology will be demonstrated via an R Shiny app.

The intended audience for this course includes statisticians, data scientists, and clinical trial practitioners who hold at least a masters-level degree in statistics or a related field. The primary teaching objectives for this course are to (1) provide practitioners with a better understanding of basic concepts of Bayesian clinical trial design and sample size determination, (2) help practitioners understand the benefits and challenges of applying Bayesian methods in clinical trial designs, and (3) to demonstrate the implementation and evaluation of Bayesian designs through real applications and user-friendly interfaces such as R packages and R Shiny apps. By providing applied practitioners with a comprehensive understanding of important concepts related to Bayesian design of clinical trials, they will be much more equipped for appropriately using these new methods in real applications.

Short Course 4: On RMST in survival analysis

May 18, 2025, 8:30am - 5:00pm,PDT

Stanford Campus, LKSC 130

Instructor: Lu Tian, Stanford University

Lu Tian is Professor of the Department of Biomedical Data Science at Stanford University. He received his Doctor of Science degree in Biostatistics from Harvard University. Dr. Tian has rich experience in statistical methodological research, and study design for randomized clinical trials. He has published more than 300 research papers in both statistical and clinical journals. He is an elected Fellow of American Statistical Association. His current research interest is in developing statistical methods in precision medicine, meta-analysis, randomized clinical trial, and survival analysis.

Abstract

Since Professor Sir David Cox proposed the hazard ratio (HR) and its inference procedure in 1972, the HR has been routinely used for quantifying the treatment effect on time to event endpoint. The Cox regression model has also been commonly used for the association and prediction analysis. While an ideal summary measure for the treatment effect should be model-free to avoid the model misspecification at the analysis, the HR is not such a measure. Indeed, there are very few model-free measures for time to event endpoint. The most popular one is either based on the median failure time or the event rate at a specific time point. Both measures are “local” and cannot catch the short- or long-term survival profile.

In the past few years, issues of utilizing HR have been extensively discussed. The validity of the HR estimate depends on a strong model assumption; that is, the ratio of two hazard functions is constant over time. When the proportion hazards assumption is not met, the HR estimate is difficult, if not impossible to interpret. In fact, the parameter for which the empirical HR estimates is not a simple weighted average of the HR over time and it generally depends on the censoring distributions. This is highly undesirable. Moreover, even when the proportional hazards assumption is plausible, the physical/clinical interpretation of a single ratio such as the HR estimate between two groups can not be easily translated into the clinical utility of a new treatment in the absence of a reference hazard level.

In this short course, I will illustrate issues of the HR and present an alternative, that is, the t-year mean survival time (restricted mean survival time, RMST). It is the mean event-free time up to a specific time point, say, t-year. The RMST was introduced in the statistical literature in 1949, but had not got much attention until recently. Here is the list of the topics we will discuss in this short course.

1. Issues of Hazard Ratio (HR)

2. Restricted mean survival time (RMST)

3. Power comparison between RMST-based test and HR-based tests

4. Empirical choice of a truncation time for RMST

5. Study design based on RMST

6. RMST for stratified analysis of survival data.

7. RMST to non-inferiority trials

8. RMST to quantify long-term survival benefit

9. RMST to assess duration of response

10. RMST in the presence of competing risks

11. RMST for recurrent event time data

12. Average hazard as an extension of RMST.

Short Course 5: Bayesian Statistics and Bayesian Models for Practical Dose Finding and Dose Optimization Oncology Clinical Trials

May 18, 2025, 8:30am - 5:00pm, PDT

Stanford Campus, Edwards R358

Instructor: Yuan Ji, University of Chicago, Ying Lu, Stanford University and Dehua Bi, Stanford University

Dr. Yuan Ji is Professor of Biostatistics at The University of Chicago. His research focuses on innovative Bayesian statistical methods for translational cancer research. Dr. Ji is author a large number of publications in peer-reviewed journals including across medical and statistical journals. He is the inventor of many innovative Bayesian adaptive designs such as the mTPI and i3+3 designs, which have been widely applied in dose-finding clinical trials worldwide. His work on cancer genomics has been reported by a large number of media outlets in 2015. He received Mitchell Prize in 2015 by the International Society for Bayesian Analysis. He is an elected fellow of the American Statistical Association.

Ying Lu, Ph.D., is Professor in the Department of Biomedical Data Science, and by courtesy in the Department of Radiology and Departement of Health Research and Policy, Stanford University. He is the Co-Director of the Stanford Center for Innovative Study Design and the Biostatistics Core of the Stanford Cancer Institute. Before his current position, he was the director of VA Cooperative Studies Program Palo Alto Coordinating Center (2009-2016) and a Professor of Biostatistics and Radiology at the University of California, San Francisco (1994-2009). His research areas are biostatistics methodology and applications in clinical trials, statistical evaluation of medical diagnostic tests, and medical decision making. He serves as the biostatistical associate Editor for JCO Precision Oncology and co-editor of the Cancer Research Section of the New England Journal of Statistics and Data Science. Dr. Lu is an elected fellow of the American Association for the Advancement of Science and the American Statistical Association. Dr. Lu initiated the Stat4Onc Annual Symposium with Dr. Ji and Dr. Kummar in 2017 and is the PI of the R13 NCI grant for this conference.

Dr. Dehua Bi is a biostatistician specializing in Bayesian parametric and nonparametric (BNP) methods for clinical trial design, survival analysis, and machine learning applications. He is currently a postdoctoral fellow at the Stanford Cancer Institute, jointly mentored by Dr. Ying Lu and Dr. Crystal Mackall. Dr. Bi earned his Ph.D. from the University of Chicago under the mentorship of Dr. Yuan Ji. His doctoral research focused on developing novel BNP methods, designing early-phase clinical trials, and creating flexible sample size estimation and adaptation approaches under the Bayesian framework.

Abstract

Project Optimus from FDA aims to shift the paradigm of oncology dose selection by emphasizing the importance of finding an optimal dose with desirable efficacy and safety. Traditionally, the optimal dose in oncology is equivalent to the maximum tolerated dose (MTD) since most oncology drugs have been cytotoxic, making the highest tolerable dose also the most efficacious dose. Due to the development of targeted and immune oncology therapeutics, MTD is no longer the default optimal dose. Statistical designs solely aimed at finding the MTD therefore cannot fulfill the requirement of Project Optimus. In this short course, I will provide a comprehensive review and introduction of practical statistical designs for dose optimization. Through the learning provided by this short course, attendees will be exposed to the main ideas and motivations for key innovative statistical designs and strategies. The short course is constructed not long to teach “hows” but “whys”, so that attendees will develop ability to assess the pros and cons of different designs for their individual needs in practice. The following topics will be covered with three sessions:

Session 1: Brief review of Bayesian statistics and modeling. This session will expose attendees to some basic concept and techniques of Bayesian statistics as many designs for early-phase oncology trials are Bayesian.

Session 2: Review of key dose-finding Designs, including but not restricted to 3+3, CRM, mTPI, mTPI-2 (keyboard), BOIN, and i3+3. This session will review a set of major dose-finding designs aiming to identify the MTD, which sets the stage for innovative dose-optimization strategies and designs.

Session 3: Introduce strategies and designs for oncology dose optimization. This session will introduce a few key designs and strategies for dose optimization like eff-tox dose-finding designs, expansion cohorts trials, backfill designs, seamless phase 1-2 designs, PK-empowered dose finding, and randomized dose comparison, etc.

Session 4: Q&A Throughout the short course, available software and tools will be illustrated for the course attendees.