Skip to Main Content

Cancer Detection is Making Strides Thanks to Predictive Analytics

Public health researchers are developing methods for better detecting and predicting dangerous forms of cancer.

5m read
Written by:
3d illustration of AI(Artificial Intelligence), Artificial General Intelligence (AGI) or deep learning or machine learning concept representing how the department of veterans affairs and nation institutes of health are overseeing programs to improve the outcomes for dangerous cancers using machine learning and artificial intelligence to improve detection rates
Photo Credit: shutterstock:Pdusit

The Department of Veterans Affairs and National Institutes of Health are overseeing programs to improve the outcomes for particularly dangerous cancers using machine learning and artificial intelligence to improve detection rates.

One of the technical priorities of both agencies has been developing methods for collecting and standardizing data used in AI-backed research, launching collaborative initiatives that draw medical informatics from multiple sources in order to create the data pools necessary to develop predictive models.

The NIH National Cancer Institute has put a particular focus on using natural language processing to standardize doctor’s notes and other sources of commentary on patient health to draw discrete information on cancer progression — particularly with the intention of early detection and better predicting a course of treatment.

Much of this work centers on using advanced capacities to better leverage existing resources, or to make information sources actionable in ways that might not have previously been possible.

“Growing the data set by something like natural language processing is an example of where this resource link can be helpful to improve the amount of data we can use,” said Amanda Krauze, NCI radiation oncologist, during the Department of Veterans Affairs’ National Artificial Intelligence Institute’s BRAIN Summit in Washington, D.C., Wednesday.

Leveraging new medical information sources leaves considerable variation in the relative structure of the data, with certain kinds of data much more neatly regimented and ready to use, while other sources require a certain degree of curation to standardize — which these newly capacities are helping accomplish.

Krauze emphasized that federal agencies are focused on mapping data inputs to specific markers of cancer progression, and that “the goal is to connect our resources and technology to the disease.” This standard has become the unifying benchmark for applying these technologies in a clinical setting, and for evaluating their overall efficacy at improving patient outcomes and care.

Much of this centers on analyzing tumor progression, which has become a primary focus of federal AI health care researchers and is currently one of the major discoveries driving the field forward.

“There are a number of AI-driven approaches to try to solve this particular question — what is progression,” Krauze said. “When you’re in a clinic, your patient really just wants to know is it growing or is it stable, and is the treatment successful?”

This has culminated in shifts across VA and Defense Department health care practice, where information and knowledge sharing between clinicians and researchers has streamlined. Broadly referred to as a “learning health care system,” the practice involves a dynamic approach to technical advancement where discoveries are facilitated via a continuous feedback between these nodes of collaboration.

“The idea of a learning health care system is to improve care by delivering knowledge in real time. Where clinical decision-making occurs,” said Nathanael Fillmore, VA associate director for machine learning. “We want to try to immediately deliver knowledge to the bedside or to the providers who are taking care of our veterans.”

These collaborative partnerships, combined with the growing input of patient data from multiple sources such as the Precision Oncology Data Repository (PODR), is allowing for certain kinds of advanced detection that would have previously been impossible.

One of the most promising horizons rests on complex analysis of patient histories to allow for the advanced detection of simultaneously aggressive, yet hard to notice, conditions like pancreatic and brain cancers. The greater the data inputs, the more sophisticated these kind of AI and machine learning processes can become.

These innovations are now also merging with electronic health record modernization, with data powering these algorithms, while also scanning EHRs for warning signs of early stage cancer.

“Passively surveilling the electronic medical record allows us to identify candidates for more traditional screening,” Fillmore said.

The ultimate outcome will be potentially unprecedented breakthroughs in treatment for conditions that had previously been considered fatal upon detection — advancing quality of life for both America’s veterans and cancer patients as a whole.

Related Content
Woman typing at computer

Stay in the know

Subscribe now to receive our curated newsletters

Subscribe