Heidi Hanson.
Heidi Hanson, leader of Oak Ridge National Laboratory’s Biostatistics and Multiscale Systems Modeling Group, has been asked to speak Jan. 26 as part of the federal Cancer Moonshot seminar series.
The series showcases research from the various Cancer Moonshot initiatives to inform the scientific community about the progress of Cancer Moonshot–funded projects, enhance discussions and collaborations, and promote the sharing of data.
The Cancer Moonshot, a federally funded program launched in 2016, focuses on bringing large communities of researchers, clinicians and patients together to share information, conduct new research and improve access and quality of care. The overarching goal is to improve the lives of people with cancer, of those at risk and of their loved ones. Last year brought the effort into a new phase, with an emphasis on collaborative educational and clinical efforts among federal agencies, research groups and lay people.
Hanson’s efforts as a researcher and team leader, as well as her upcoming policy recommendations at the Cancer Moonshot seminar series, support and exemplify this collaboration.
She serves as the technical lead on the DOE-National Cancer Institute’s Modeling Outcomes Using Surveillance Data and Scalable Artificial Intelligence for Cancer program, or MOSSAIC, which was developed and led by Georgia Tourassi, director of the National Center for Computational Sciences and the Oak Ridge Leadership Computing Facility. MOSSAIC focuses on advancing computing, predictive machine learning models and large-scale computational simulations for cancer research.
An interdisciplinary team of ORNL scientists collaborating on the project have advanced the research over the past few years.
“A lot of our work is focused on automating processes that are time-consuming when done manually,” Hanson said.
The Surveillance Epidemiology and End Results program, developed by the ORNL team, illustrates the value of this automation. The program seeks to capture large amounts of cancer data from representative portions of the U.S. population to study long-term trends and the effectiveness of healthcare and screening interventions. But researchers and clinicians are always a step behind without faster information processing.
“Because the work is commonly done manually, there is generally a two-year lag between diagnosis and release of reporting statistics,” Hanson said. “So when big events like COVID-19 happen, we don’t really know how they affect cancer screening and diagnosis until it is too late for a response or intervention.”
She and her team are using deep learning algorithms to automate the program’s information processing and have automated processing for roughly 30% of incoming reports so far.
Hanson believes larger data pools are essential to improving cancer outcomes in the U.S.
“One of my biggest recommendations is that we really start thinking about monitoring cancer nationwide and on a micro scale, which would give us more information about the social and environmental factors that contribute to changes in cancer incidence,” she said.
Hanson emphasized the value of sharing code between institutions and agencies, of allowing for distributed training of AI models, and of ultimately improving accuracy and predictive power in cancer studies.
These recommendations underscore the power of AI automating tools to enable cost-effective nationwide cancer surveillance, which she will advocate in her talk at the Cancer Moonshot seminar series.
UT-Battelle manages ORNL for DOE’s Office of Science, the single largest supporter of basic research in the physical sciences in the United States. DOE’s Office of Science is working to address some of the most pressing challenges of our time. For more information, visit https://energy.gov/science.— Galen Fader
