Office of Research & Development |
|
Scientists at seven VA Medical Centers have begun a study that will examine the effectiveness of AI and machine learning to predict treatment response for Veterans with mouth and throat cancer. (Photo © iStock_Nico El Nina)
April 22, 2024
By Erica Sprey
VA Research Communications
"We will put to work the newest AI tools to help improve treatment and quality of life for patients with OPC."
After two decades of exponential increases, U.S. Veterans are now three times more likely to develop oropharyngeal cancer (OPC) than the general population and have shown persistently lower survival rates. VA scientists are looking to change this tragic trend with an innovative use of artificial intelligence (AI).
Because cancer treatment can be physically challenging and toxic to the body, a team of VA scientists across seven VA Medical Centers have begun a study that will exam the use of AI to help select treatment type and intensity for Veterans with OPC: a type of cancer that occurs in the mouth and throat.
“OPC biology and treatment response in the Veteran population differs from what is encountered in the broader U.S. population,” said Dr. Vlad C. Sandulache, co-lead for the study and surgeon at the Michael E. DeBakey VA Medical Center in Houston. “This multi-institutional effort is designed to leverage collaborations between oncologists, pathologists, radiologists and computer scientists to develop more effective precision oncology approaches to this deadly disease.”
The research team will use data on 1,000 Veterans with OPC and information from the Computer Vision and Machine Learning in Precision Oncology (CoMPL) hub to build and train an AI algorithm to assess chemo-radiation treatment response. The new platform will be called the AI-based Risk Stratification for Oropharyngeal Carcinomas (AIROC).
Cancer is diagnosed in part through diagnostic images, like CT scans, and pathology slides made from tissue biopsies. Because the AIROC platform will use data that is collected in the normal course of cancer care for Veterans, when/if it is validated, it could be deployed across the VA health system with minimal costs or training.
The goal for AIROC is to develop algorithms that will assist pathologists and radiologists with their diagnoses and make it easier and more effective to assess disease pathology. According to Michael Gilkey, a data scientist at the Atlanta VA, computers make it possible to “zoom in” on information that the human eye can’t see, like the twistedness of blood vessels or the inside of a tumor.
A type of head and neck cancer, OPC is most often caused by infection with the human papilloma virus. HPV infection, which is a common sexually transmitted disease, causes 70% of OPC cancer cases, according to the Centers for Disease Control.
Some individuals can safely undergo less-intensive treatment for their OPC and still have good outcomes, but deciding who might benefit from that approach is difficult. That is why the research team will develop a special computing platform that can review biomarkers like CT scans and pathology slides and integrate that information with Veterans’ clinical data to predict treatment response. These programs will be able to review medical data and select patients who have a greater likelihood of benefiting from a lower treatment intensity for their cancer.
In the future, it might even mean the difference between forgoing radiation treatment, getting a lower dose of chemotherapy, or intensifying treatment. In all cases though, it will still be the cancer specialists who guide Veterans’ therapy to ensure they receive optimal care.
“This work will use Veteran data in the service of Veterans, putting to work the newest artificial intelligence and machine learning tools to help us improve treatment effectiveness, survival and quality of life for patients with oropharyngeal cancer,” Sandulache said.
Across the VA Office of Research and Development’s (ORD) Precision Oncology Program, AI is being used to study and improve treatment approaches for a number of different cancers, reinforcing its goal to address the real-world health care needs of Veterans. In addition to refining treatment approaches for Veterans with cancer, the development of precision clinical tools also supports ORD’s Precision Oncology research portfolio.
Current VA studies that use AI include the development of a “Lung Imaging based Risk Score,” “A Novel Radiomics Toolkit to Predict and Characterize Response to Immunotherapy in Stage III Non-Small Cell Lung Cancer,” and using AI to “Predict Metastatic Progression of High Risk Localized Prostate Cancer.”
Dr. Anant Madabhushi is a research career scientist at the Joseph Maxwell Cleland Atlanta VA Medical Center, director of the Empathetic AI for Health Institute at Emory University, and co-lead on the AIROC study. He is a recognized leader in AI and machine learning who has dedicated his career to developing AI applications that use multimodal datasets to find disease characteristics to predict risk, prognosis, therapeutic response, and outcomes in patients. As a young man, he lost his aunt to an aggressive breast cancer, and resolved to find a better way to help physicians treat the disease.
“We are excited about the use of multimodal imaging AI for better risk stratification and outcome prediction for Veterans with head and neck cancer,” Madabhushi said. “These approaches will pave the way for better treatment management of Veterans with cancer. It will also encourage and generate innovative research in AI within the VA.”
*Listen to Madabhushi’s Tedx Atlanta Talk on “Could AI find your cancer?”
VA Research Currents archives || Sign up for VA Research updates