Project Spotlight: PI Predictor

Shandong Wu, PhD, is a tenure-track Assistant Professor in the Department of Radiology at the University of Pittsburgh with joint appointments in the Departments of Biomedical Informatics and Bioengineering. Dr. Wu leads the Intelligent Computing for Clinical Imaging (ICCI) lab and serves as the Technical Director for AI Innovations in Radiology at Pitt/UPMC. His research group is engaged in a range of interdisciplinary projects with clinical and translational applications, including computational analysis of biomedical images, machine learning, radiomics and radiogenomics, and AI in clinical workflows. In the PI Predictor project, Dr. Wu has formed a team of pathologists and oncologists to develop an augmented breast cancer recurrence risk prediction model by using advanced computational modeling and integrating standard-of-care clinical data.

Please share a little about your background and your research experiences.

My background is in Computer Science (Computer Vision) with additional clinical training in radiology research. My main research areas include computational biomedical imaging analysis, artificial intelligence in clinical/translational applications, big (health) data coupled with machine/deep learning, imaging-based clinical studies, and radiomics/radiogenomics/radioproteomics. My research is supported by NIH/NCI, RSNA, UPMC Enterprises (via the Pittsburgh Health Data Alliance), The Pittsburgh Foundation, and Amazon. As a PI, I’ve received more than $4.5 million dollars in research funds over the past five years.

What led you to the PHDA?

The initial idea of our PHDA-funded project went back about four years ago when I started to build some collaborative work with UPMC Magee-Womens Hospital’s pathologists and medical oncologists. At that time, a main theme of my research was to develop quantitative imaging biomarkers to improve clinical diagnosis and prognosis prediction of breast cancer patients. While regularly attending the weekly breast cancer conference at Magee, I identified a research idea to develop an augmented breast cancer recurrence risk prediction model by using advanced computational modeling methods and integrating standard-of-care multi-modality clinical data. I formed a team with my clinical collaborators, Dr. David Dabbs and Dr. Rohit Bhargava (both are breast cancer pathologists) and Dr. Rachel Jankowitz (a medical oncologist specializing in breast cancer) to start a pilot study, and we have achieved encouraging results. This pilot project and collaboration led us to the PHDA for funding to further develop and evaluate our computational models for breast cancer recurrence risk prediction and for estimating potential benefit from chemotherapy.

Walk us through your project.

We started a pilot project on roughly 100 breast cancer patients to show the feasibility of our ideas, and then, in the PHDA-funded project, we formulated more specific research aims and goals to improve the models and to perform an evaluation on a larger set of de-identified data collected at UPMC. To this end, we have developed automated deep learning algorithms for imaging segmentation. We also have evaluated a large set of imaging features and clinical variables for machine learning modeling. We are currently running integrated analyses on the data set we have collected and are performing different experiments for an optimal scheme to build the prediction models. We have generated several important papers and abstracts from this study.

 How do you and your project partners’ strengths complement each other?

This is a multi-disciplinary project and requires expertise from both computational science and clinical science. I lead a large lab (Intelligent Computing for Clinical Imaging (ICCI) lab) consisting of 15 trainees with expertise in computer science, artificial intelligence, bioengineering, radiology, biology, and statistics. More importantly, I have strong clinical partners: pathologists Dr. David Dabbs and Dr. Rohit Bhargava and medical oncologist Dr. Rachel Jankowitz to work with on this project. It has been clear that our complementary expertise and close collaboration are key to implementing this project. The input from my clinical partners has guided some of the important processes and development of computational methods. Simultaneously, the strong computational expertise in my lab also enabled the implementation of advanced data analysis and machine learning modeling.

How is the PHDA uniquely positioned to assist your team and grow your project to commercialization?

The PHDA has really been in a unique position to help our research team. At the early stage of the project, we particularly benefitted from the assistance from the PHDA teams in helping to clarify and focus on the pain-points, design a business model, and explore the potential market. These are the areas where we as researchers need more help in terms of the plans for commercialization, and we did receive professional and comprehensive assistance from Center for Commercial Applications of Healthcare Data (CCA) teams. I personally really appreciate the help from Kathrin Gassei, Senior Program Manager at the CCA and Pitt’s sciVelo, and Julie Cramer from Pitt’s Innovation Institute, among other helpful folks, during the phase of the project developing for funding and also in the process of executing the project.

What are your project’s next steps?

After finishing the proposed goals, we are interested in pursuing a follow-up funding request with UPMC Enterprises for the clinical validation of the developed models. A prospective study at UPMC/Magee will be in plan for this purpose along with the consideration for an independent validation study in external collaborative medical centers.