In a major milestone for cancer research, the Department of Energy's Oak Ridge National Laboratory helped reduce the time between cancer diagnosis and pathology report processing from 22 months to just 14 months, utilizing advanced artificial intelligence technology developed by ORNL's research team.
This significant step forward, announced by the National Cancer Institute's Division of Cancer Control and Population Sciences (DCCPS), could revolutionize the speed at which cancer trends are identified and addressed.
This work was part of the Modeling Outcomes Using Surveillance Data and Scalable Artificial Intelligence for Cancer program, or MOSSAIC , which serves as a partnership between NCI and ORNL to advance precision oncology and scientific computing. ORNL has made significant contributions to this effort as well.
Historically, cancer registries are updated manually by sorting through stacks of pathology reports. This slowdown can create as much as a two-year gap between the cancer diagnosis and its reporting. That means if there is an increase in cancer rate nationally, researchers might have to wait two years before recognizing this uptick in diagnoses.
ORNL's Heidi Hanson serves as a senior research scientist and leads the Biostatistics and Biomedical Informatics group. She described this technology, although a different architecture, as similar to a large language model such as ChatGPT.
"We can reduce the time it takes to process reports by using artificial intelligence to autocode records. Our MOSSAIC team is utilizing a large volume of sensitive medical text to train a model to learn what features inside that text are important for predicting a classification of disease," Hanson said. "These classifications, or common data elements, are then used in downstream statistical analyses to learn more about what types of disease are present in our populations."
Hanson said the Information Extraction API, or OncoIE, a deep learning algorithm developed by ORNL researchers, played a critical role in such a drastic improvement in reporting time.
"In my opinion, (OncoIE) is really ahead of its time in terms of what can be done with AI in medicine and how deep learning models can be used to facilitate rapid or near real-time response through the extraction of critical information from a large volume of unstructured data," said Hanson.
The achievement was highlighted at an event celebrating the retirement of Lynne Penberthy of NCI, who served as the associate director for the Surveillance Research Program within the Division of Cancer Control and Population Sciences at NCI.
The NCI's Surveillance, Epidemiology, and End Results, or SEER, program provides information on cancer incidence and survival in the United States, collecting and publishing cancer incidence and survival data from population-based cancer registries covering approximately 48% of the U.S. population. Researchers use this information to organize pathology reports to better understand increases in cancer diagnoses in a particular area or population. All the algorithms used by SEER registries were developed through the MOSSAIC project.
Hanson said improving this nearly two-year difference in reporting is a game-changer in cancer prevention.
"When you're 22 months behind in even knowing what's happening on the ground, you can't prevent it," Hanson added.
As part of the event, Goddard announced the success in reducing the reporting time, but Hanson stressed they have much more ambitious milestones.
"They discussed the next goal on the horizon, trying to reduce this amount of time down from 14 months to two months," she said. "We want to prevent disease, so we want to identify changes in disease trends early on so that we can stop whatever's happening and reduce the number of cancer diagnoses."
"Having this two-month goal is ambitious, but can you imagine what we could do if we had information about an increase in cancer rates in time to quickly investigate potential causes and change screening practices, rather than reacting when it might be too late," she added.
UT-Battelle manages ORNL for the Department of Energy's Office of Science, the single largest supporter of basic research in the physical sciences in the United States. The Office of Science is working to address some of the most pressing challenges of our time. For more information, please visit energy.gov/science . - Mark Alewine
