AI Decision Support System Guides Treatment Selection for Complex Blood Cancers

Treatment selection for hematologic malignancies often requires clinicians to synthesize clinical histories, genomic alterations, prior therapies, and rapidly evolving drug options. These complex decisions are typically resolved through multidisciplinary tumor boards, which smaller centers may find difficult to convene consistently. As evidence expands and case heterogeneity increases, recommendations can become more variable. Researchers now describe an artificial intelligence (AI) approach designed to help clinicians navigate treatment decisions for complex blood cancers.

The German Cancer Research Center (DKFZ), together with the HI-STEM Stem Cell Institute and Heidelberg University Hospital (UKHD), developed HemaGuide, an AI clinical decision-support assistant for malignant hematology. The system is intended to aid physicians with difficult treatment choices by providing case-specific recommendations accompanied by transparent reasoning. According to the developers, HemaGuide can be deployed on local hospital servers so that sensitive patient data remain on site.

HemaGuide analyzes unstructured physician reports, organizes the extracted information, and integrates it with current treatment guidelines, a repository of more than 2,000 real-world tumor board cases, and the latest scientific literature. The platform can function as a molecular tumor board by interpreting tumor genetic alterations against internationally established standards, automatically surveying relevant publications, and proposing targeted therapies where appropriate. The molecular analysis typically completes in less than a minute, a task that previously often required several hours and was available only at a limited number of specialized centers.

Performance was evaluated across multiple assessments. In 45 particularly complex cases, experienced hematologists rated HemaGuide’s recommendations significantly higher than those from conventional AI language models without the system’s specialized architecture, citing better alignment with tumor board decisions and individualized patient factors. In a separate test using 555 tumor board cases from an independent university hospital encompassing 47 blood cancer types, HemaGuide’s recommendations matched expert panel decisions in nearly 82% of cases; in a one‑month prospective parallel run, agreement was just under 83%.

Additional analyses indicated benefits for less experienced clinicians. In a simulated study, residents supported by HemaGuide approached the performance of senior physicians. For automatic evaluation of genetic alterations, the system adhered to international expert standards, and no clearly cancer‑promoting variant was mistakenly labeled benign. The work was published in Nature Medicine on June 30, 2026, and the team is preparing a clinical trial to examine effects on care quality and long‑term outcomes.

“HemaGuide is intended to help improve access to highly specialized cancer care. Smaller clinics in particular could benefit from this support in decision-making,” said Julian Zoller, one of the two first authors at the German Cancer Research Center.

“The system is not intended to replace medical expertise, but to complement it. It can ease the burden on tumor boards and make their knowledge more widely available—but the final decision is always made by the treating physician,” said Mirco Julian Friedrich, study leader at the German Cancer Research Center, HI-STEM, and Heidelberg University Hospital.

Related Links
DKFZ
HI-STEM Stem Cell Institute
Heidelberg University Hospital