Artificial Intelligence to Improve Leukemia Diagnosis

Malignant hematological diseases include various types of blood cancer, such as leukemias and lymphomas. These diseases result from the abnormal growth and proliferation of blood cells, which can lead to a variety of health problems. Those affected often experience fatigue and reduced performance, increased bleeding, and are more susceptible to infections.

The BELUGA study is investigating whether the use of artificial intelligence (AI) can improve the diagnosis of blood cancer, particularly leukemias. In this context, artificial intelligence refers to computer-based methods that can learn autonomously from large amounts of data—similar to how a human learns from experience. The AI is trained to detect pathological changes in the blood by first being “fed” thousands of previously analyzed blood smears and flow cytometry data. Specifically, the study uses two key diagnostic methods: first, blood smears, in which cells are examined under a microscope, and second, flow cytometry, a laboratory test in which cells are identified based on their surface characteristics. Both methods yield a wealth of complex data—and this is precisely where the AI comes in. It is trained to recognize typical patterns of leukemia in this data and then apply them to new patient samples. A total of 25,000 digitized blood smears and an equal number of flow cytometry data sets from the Munich Leukemia Laboratory (MLL) are being used to develop a so-called “deep neural network” —a particularly powerful form of AI that processes structures in a manner similar to the human brain. In the next step, researchers will test how well this system performs compared to standard diagnostics conducted by experienced physicians. Among other factors, the evaluation will assess sensitivity (accuracy in identifying positive cases), specificity (accuracy in identifying healthy samples), and the speed of diagnosis. The hope is that AI can make diagnostics not only faster but also more precise—for example, by detecting small details that even experienced doctors might overlook. This could significantly improve care, especially in regions with limited access to specialized physicians. This study is the first prospective trial worldwide to test the clinical use of AI in blood cancer diagnostics under real-world conditions. If successful, it could fundamentally change the way leukemias are diagnosed in the future.

The study is scheduled to run until approximately 2031. The study does not test any novel treatment options. For participating patients, potential enrollment in the study will not change the type or scope of the tests performed to establish a correct diagnosis, nor will it change how treatment is administered.