Artificial intelligence is becoming an important tool in medicine, helping doctors analyze large amounts of information more quickly.
A new AI system called HemaGuide could soon make it easier for doctors to treat people with blood cancers, especially in hospitals that do not have teams of highly specialized experts.
The system was developed by researchers from the German Cancer Research Center (DKFZ), the HI-STEM Stem Cell Institute, and Heidelberg University Hospital. Their work was published in Nature Medicine.
Treating blood cancers such as leukemia, lymphoma, and myeloma has become increasingly complicated.
Doctors must consider a patient’s age, medical history, previous treatments, laboratory results, genetic changes in the cancer, other illnesses, and the latest treatment guidelines. New medicines are also becoming available every year, making treatment decisions even more challenging.
Large cancer centers often solve these difficult cases through tumor boards. During these meetings, doctors from different specialties work together to choose the best treatment for each patient. However, these meetings require time and many experienced specialists. Smaller hospitals often do not have enough experts, particularly for rare blood cancers.
To help solve this problem, the researchers created HemaGuide. The AI reads doctors’ reports, organizes important medical information, compares it with current treatment guidelines, reviews more than 2,000 previous tumor board cases, and searches the latest scientific studies.
It then produces a treatment recommendation together with a clear explanation of how it reached its conclusion.
One of HemaGuide’s strongest features is its ability to analyze genetic changes inside cancer cells. This process normally requires experts and may take several hours. HemaGuide completes the task in less than one minute while following internationally accepted medical standards. It can also suggest targeted treatments when appropriate.
Another advantage is patient privacy. The system can run entirely on a hospital’s own computer servers, so confidential patient information does not need to be uploaded to outside companies or cloud services.
The researchers carefully tested HemaGuide. In one study involving 45 particularly difficult cases, experienced blood cancer specialists rated its recommendations more highly than those produced by general AI language models.
In another study involving 555 tumor board cases covering 47 different blood cancers, the AI agreed with expert treatment decisions in nearly 82% of cases. During a one-month real-world trial, agreement increased to almost 83%.
The system also helped less experienced doctors. Resident physicians using HemaGuide performed almost as well as experienced senior specialists in simulated decision-making exercises. Importantly, the AI did not incorrectly label dangerous cancer-related genetic changes as harmless.
The researchers emphasize that HemaGuide is designed to support doctors rather than replace them. Physicians remain responsible for the final treatment decision. The AI simply provides expert-level information much more quickly.
The findings are promising because they suggest advanced cancer expertise could become available to many more hospitals. However, the current research mainly measured agreement with expert decisions rather than patient outcomes.
Clinical trials are now being planned to determine whether using HemaGuide actually improves survival and quality of care. If those studies are successful, this technology could become an important partner for doctors treating blood cancer patients.
If you care about cancer, please read studies that artificial sweeteners are linked to higher cancer risk, and how drinking milk affects risks of heart disease and cancer.
For more health information, please see recent studies about the best time to take vitamins to prevent heart disease, and results showing vitamin D supplements strongly reduces cancer death.
Source: German Cancer Research Center (DKFZ).
