Defense Innovation Unit Develops Algorithms to Assist Cancer Diagnosis

The Defense Innovation Unit’s Predictive Health program is prototyping artificial intelligence and machine learning tools to facilitate early cancer diagnosis.

DIU explained that the current process, where pathologists search for cancer signs on biopsy samples, is time-consuming and can lead to errors. It also noted that the number of personnel qualified to evaluate samples for cancer is trending downward.

According to the agency, pathologists supported by AI and ML locate cancer faster and more accurately. To obtain feedback from such personnel, the algorithms were integrated into workflows at some Department of Veterans Affairs and military treatment facilities domestically and abroad.

Pathologists used the AI and ML capabilities through augmented-reality microscopes, which create inference overlays on the field of view.

The PH program has so far produced four new machine-learning algorithms to support cancer detection on pathology slides. 

To sustain ARM testing efforts, work is underway alongside Germany-based optic technology maker Jenoptik to expand adoption by selling the devices to the public sector, DIU said.