Virginia Tech®home


Ongoing Projects:

Artificial Intelligence in Radiation Oncology, Problems and Solutions Book Project: Chartering the future of Radiation Oncology Services with use Artifical Intelligence and its implications. This book is envisioned to serve as a comprehensive resource for scientists, physicians, and engineers interested in developing the next generation Radiation Oncology technologies by taking advantage of artificial intelligence tools. This book is being co-edited by Dr. Seong K. Mun of Virginia Tech and Dr. Sonja Dieterich of UC Davis with an expected publish of September 22 under World Scientific Publishing Group.

Center of Excellence, Artificial Intelligence for Medical Imaging (AIMI):

The Arlington Innovation Center: HR is developing a Center of Excellence in Artificial Intelligence for Medical Imaging (AIMI) in partnership with the Department of Radiology of the Central Virginia Veterans Healthcare System, Richmond, Virginia. The details of the project are available at

Project SoterRO for Radiation Oncology: 

Modern radiation therapy of cancer involves a series of powerful technologies, integrated clinical team, and massive amount of documentation procedures. There are numerous error checks and quality assurance steps involving many devices and software packages. A large portion of this process is done manually and error propagation can contribute to poor outcome.

Our aim is to make meaningful productivity improvements by efficient proactive error reduction and continuous quality improvement through-out the entire radiation therapy process. We are applying an array of AI/ML tools to large amounts of clinical data and medical images from more than 50 data points. This collaborative project involves, RadAmerica of MedStar Health, RadPhysics Inc., Georgetown University Medical Center and North Carolina State University.

Imaging Component of APOLLO Network Project: 

The Applied Proteogenomics OrganizationaL Learning and Outcomes (APOLLO) network is a collaboration between NCI, the Department of Defense (DoD), and the Department of Veterans Affairs (VA) to incorporate proteogenomics into patient care as a way of looking beyond the genome, to the activity and expression of the proteins that the genome encodes. The emerging field of proteogenomics aims to better predict how patients will respond to therapy by screening their tumors for both genetic abnormalities and protein information, an approach that has been made possible in recent years due to advances in proteomic technology. We are supporting projects dealing with diagnostic imaging and pathologic imaging and related research such as quantitative imaging.

Clinical Trial of Imaging AI Products for FDA Clearance:

In the past 30 years, we have been working on medical imaging CAD/AI research and have collaborated with our industrial partners on the development of medical imaging AI products. We also helped industrial partners perform AI product evaluation studies as a part of clearance processes required by the FDA. As the evaluation team, we consult with the sponsor to develop an evaluation project plan including (1) assist the sponsor in developing a Q-submission to discuss with FDA if appropriate, (2) assist the sponsor in developing product claims [intent to use], (3) select proper study methodology, (4) check-up image data quality that meets inclusion criteria, (5) estimate data size and number of readers needed to obtain statistically significant results, (6) develop a complete and excusable study protocol, (7) apply an IRB for the study, (8) register, (9) assist the sponsor in developing informed consent form for study participants, (10) recruit experts to confirm and establish the truth for the test data-set, (11) review the test data-set to ensure proper distribution of the data [mainly done by the clinical director], (12) recruit study readers, (13) perform stand-alone machine study with the test data-set, (14) assist the sponsor in designing sequence of study functions on the computer workstation,  (15) prepare and set-up a proper study room for each reader to perform the study, (16) perform rehearsal of the reader study [training session followed by study session], (17) execute the reader study [training session followed by study session], (18) submit output data of the study for statistician's analysis, (19) write a complete study report based on the study protocol and study results, (20) assist the sponsor in filing the FDA clearance application, (21) submit output data of the study to FDA per FDA's request, and (22) answer FDA's questions.  We welcome comments and inquiry about potential medical image related evaluation study and clinical trial. 

Research Activities:


Convolutional Neural Networks (CNN) with Symmetric Kernels for Medical Imaging Applications: 

The structure of a typical convolutional neural network consists of two major sections: convolution and classification.  Kernels associated with multiple channels in each layer of the convolution section are the parameters to be trained.  Weights associated with connecting lines between nodes of adjacent layers in the classification section are the parameters to be trained.  Unlike alphanumerical writing and some nature image recognition, most of medical image patterns are orientation independent. Hence the symmetric kernel are a better choice for the stabilization of CNN performance. We have developed transformation-identical convolutional neural network (TI-CNN) and geared rotation-identical convolutional neural network (GRI-CNN) structures. The algorithms of TI-CNN has been tested using Darknet and Caffe platforms. Potential applications of the symmetric kernel CNN include, but not limit to, disease pattern recognition, predictive model based CT and MRI image reconstruction, and general image segmentation.

Prostate Cancer Outcomes Research:

We have developed a web-based tool to estimate the likely outcomes of CyberKnife radiation therapy treatment of prostate cancer based on the outcomes data that Georgetown University Medical Center has collected for approximately 850 patients over a 5 year time period. The web-based tool can be used by patients and oncologists. After initial clinical deployment by Dr. Anatoly Dritschilo, Chair of Department of Radiation Medicine of Georgetown University Medical Center, Dr. Xiaofeng Zhu of AIC is currently working to improve the useability of the system. The software has been released as open source code and additional information can be found here.

Radiation Induced Immunotherapy:

Yanni Li, MD of AIC is working with Anatoly Dritschillo, MD, of Georgetown University Medical Center to explore Radiation Induced Immunotherapy.  The project has collected the data from prostate cancer patients who relapsed after treatment and those with other related complications from the record. Statistical Analysis is used in order to find the change tendency of the proteins in the blood following the timeline. Then, by studying the change curves of proteins that are truly correlated and combining the actual functions of proteins and pathway analysis, proteins related to recurrent prostate cancer and other complications of prostate cancer are screened out. Lastly monoclonal antibodies are used to test the validity in the lab. For now, the experiment remains at the laboratory verification stage.

Blockchain Enabled Next Generation Longitudinal Personal Health Record (LPHR):

Blockchain Technology can offer uniquely powerful tools in managing health information in highly heterogeneous environment with many systems, stakeholders and users.  Yibin Dong, PhD student of Professor Joseph Wang of Electrical and Computer Engineering Department of Virginia Tech is developing a Smart Contract System using IBM's Hyperledger Fabric environment to enable the establishment of longitudinal personal health records, a high priority of Department of Health and Human Services as a means to promote interoperability in health IT.  This is a collaborative project with Bradley Department of Electrical and Computer Engineering of Virginia Tech. 

Longitudinal patient health record (LPHR) is a full record of a patient that would lead to a more accurate diagnosis and effective treatment. However, LPHR adoption rate is low mainly due to low security of lacking privacy control by patients. The objective of this research is designing a framework for implementation of a secure LPHR authorization management system based on Next Generation Access Control standard and Hyperledger Fabric permissioned blockchain technology. The solution will improve patients’ confidence in security and privacy control via fine granular data access control by patients, leading to higher user adoption rate of LPHR. The solution is distributed, decentralized, and enterprise-level scalable. We also plan to address some open issues of blockchain in healthcare applications.

Analysis of Open Source Software Communities:

While open source business model is getting wider acceptance in the private sector, the US government agencies struggle with such new concept in spite of issuance of open source policy by the White House in 2018. Our research partner has a set of rich raw material and community members in and out of government with experiences in evolving two distinctly different open source operations involving two government agencies. We have partnered with Professor Donald Wynn of University of Dayton to study these to development pathways. 

Evaluating AI Applications for Combat Casualty Care:

Combat casualty care has been research interests of Professor Kenneth Wong. It is a subfield of emergency medicine that requires intense situational awareness, encyclopedic knowledge, split-second decision making, and high-performing technology. Training medics with these skills requires much time and effort, yet even with the best training, medics can still experience numerous challenges. Artificial intelligence (AI) could offer numerous positive benefits in combat casualty care, but also has significant drawbacks and pitfalls. As a result, there is a vast, multi-dimensional space of possible AI systems and implications to be investigated. Given this context, it would be beneficial to develop a framework for guiding research and development efforts in this arena. This framework, grounded in Multiple Attribute Decision Making (MADM) methods, should benefit the field of combat casualty care in at least two ways. First, the framework will support a comprehensive and holistic view of AI applications. Second, it should help to prioritize areas and techniques for future research investments.