The Hetairos Artificial Intelligence System
Developed by a team led by Moritz Gerstung of the German Cancer Research Center (DKFZ) and Felix Sahm of Heidelberg University's Medical Faculty and University Hospital Heidelberg, the Hetairos AI system can identify over 100 molecular subtypes of central nervous system tumors in just minutes using standard histological slides. Trained on more than 11,000 digitized tissue samples from 9,606 patients across eleven medical centers on four continents, Hetairos distinguishes 102 distinct molecular tumor subtypes. When operating with high confidence—which occurs in 50-70% of cases—it achieves a prediction accuracy of 87-88%.
A study comparing Hetairos against five experienced neuropathologists across 210 cases found the AI significantly outperformed its human counterparts. Hetairos posted a 68% accuracy rate, while the neuropathologists averaged just 30%. When considering the top three most likely diagnoses, the AI's accuracy climbed to 84%, compared to only 50% for the specialists.
The Potential of the Hetairos System
DNA methylation analysis is considered the gold standard for classifying brain tumors, but traditional molecular tests take roughly two weeks to deliver results. In this context, Hetairos could become a vital tool to support diagnostic workflows.
Felix Sahm: 'We developed Hetairos primarily as a diagnostic support tool. It is not intended to replace molecular analysis, but rather to complement it in a targeted way and speed things up.'
Moritz Gerstung also noted that 'Currently, diagnosing very rare tumor types remains a major challenge for Hetairos; in this regard, experienced neuropathologists appear to be at least on par. However, we expect the system's performance to improve further with larger and more diverse datasets.' This suggests Hetairos has the potential to significantly enhance the diagnosis of brain and spinal cord tumors.
By providing doctors with fast, accurate recommendations based on cutting-edge technology, Hetairos could transform how central nervous system tumors are diagnosed. Because traditional methods require considerable time, adopting such technologies could dramatically reduce wait times for patients—a critical factor in cancer care.
Looking ahead, as algorithms improve and databases expand, Hetairos may become even more effective, particularly in handling rare tumors where diagnosis remains challenging. This could lead to better treatment outcomes and improved quality of life for patients.