Researchers at the German Cancer Research Center (DKFZ) and Heidelberg University Hospital have developed an artificial intelligence system, “Hetairos,” capable of classifying brain tumors by molecular subtype using only standard microscopic tissue sections. By analyzing digitized samples, the AI delivers diagnostic classifications in minutes, potentially reducing the wait time for molecular results from weeks to days.
How Does Hetairos Improve Brain Tumor Diagnosis?
Hetairos accelerates the diagnostic process by identifying molecular signatures that were previously only detectable through complex, time-consuming DNA methylation analysis. According to the study published in Nature Cancer, the system was trained on more than 11,000 digitized tissue sections from 9,606 patients across four continents. Lead authors Darui Jin and colleagues report that the AI identifies 102 distinct molecular tumor subtypes, covering nearly the entire spectrum defined by the World Health Organization (WHO) for central nervous system tumors.
How Does the AI Compare to Human Specialists?
In head-to-head testing, Hetairos demonstrated higher diagnostic accuracy than human neuropathologists. According to data provided by the DKFZ, five experienced specialists correctly diagnosed 30 percent of 210 test cases, while the AI achieved an accuracy rate of 68 percent. When allowing for the three most likely diagnoses, the specialists reached 50 percent accuracy compared to the AI’s 84 percent.
Felix Sahm, a neuropathologist at Heidelberg University Hospital, notes that the system identifies subtle morphological patterns that are difficult for human eyes to distinguish. However, researchers acknowledge that the AI currently faces challenges with extremely rare tumor types, where human experts remain on par with the software’s performance.
Why Is This Technology Important for Global Healthcare?
The primary advantage of Hetairos is its reliance on standard histological stains rather than specialized, expensive laboratory equipment. Moritz Gerstung of the DKFZ states that the system is designed to support, not replace, traditional molecular diagnostics. By narrowing down the potential tumor subtypes in unclear cases, the AI provides a triage tool that allows clinicians to prioritize further testing.
This approach could address significant disparities in cancer care. In many regions, the infrastructure required for standard DNA methylation testing is unavailable or prohibitively expensive. Because Hetairos utilizes existing tissue preparation workflows, it offers a scalable method for improving diagnostic speed in resource-limited settings.
Pro Tip: Understanding AI Confidence Scores
Hetairos does not simply provide a diagnosis; it provides a confidence interval. In 50 to 70 percent of cases, the system reports a high level of certainty. When the AI is uncertain, it still provides a narrowed list of candidates, which significantly reduces the cognitive load for pathologists and streamlines the path to a confirmed diagnosis.
Frequently Asked Questions
Is this AI intended to replace human doctors?
No. According to Felix Sahm, the system is designed to assist neuropathologists by complementing existing molecular analyses and accelerating the diagnostic workflow.
How long does the diagnosis process take with Hetairos?
The AI generates findings in approximately twelve minutes on standard hardware. When including the time required for tissue preparation and digitization, results are typically available within 24 to 48 hours.
What types of tumors can Hetairos classify?
The system is trained to distinguish 102 different molecular subtypes of central nervous system tumors, encompassing almost the entire range of the current WHO classification system.
Are you interested in how digital pathology is changing the future of oncology? Subscribe to our newsletter for the latest updates on medical AI research and diagnostic innovations.
