Digitalization is playing an increasingly important role in laboratories as well. The use of artificial intelligence in data analysis significantly expands possibilities in the areas of diagnostics and individualized therapies. Therefore, the German Society for Clinical Chemistry and Laboratory Medicine (DGKL) awards the Digital Laboratory Funding Prize for outstanding scientific achievements. This year, Dr. Jan Middeke and Dr. Jan-Niklas Eckardt, physicians and researchers at the University Hospital Dresden, the Else Kröner Fresenius Center (EKFZ) for Digital Health at TU Dresden, and the National Center for Tumor Diseases (NCT/UCC) Dresden, are receiving the award, endowed with €15,000, for their research on detecting leukemia in bone marrow using AI. They share the award with Annika Meyer from Cologne, who receives the prize for her research as part of her medical state examination on the application of artificial intelligence.

The so-called acute myeloid leukemia, or AML, is a malignant disease of the blood-forming system, in which blood-forming cells multiply uncontrollably at an early stage. The affected cells never develop into functional blood cells but gradually displace them. A rapid and precise diagnosis is crucial for developing an individual and thus successful therapy for those affected.

Dr. Jan Middeke and Dr. Jan-Niklas Eckardt lead the interdisciplinary working group "Artificial Intelligence in Cancer" at the University Hospital Dresden and the EKFZ. With their team, the two doctors research and develop systems for computer-assisted decision-making to improve the diagnosis of blood cancer at the intersection of medicine and computer science. The diagnostic methods currently used for hematological diseases are reaching their limits. "The assessment of bone marrow morphology by experts is crucial for the diagnosis of AML but currently suffers from a lack of standardization. Deep learning can address this issue by analyzing medical imaging data and providing precise predictions," explains Dr. Middeke, who also leads the outpatient clinic and day clinic in hematology/oncology at NCT/UCC Dresden. As a co-founder of the working group "Artificial Intelligence in Hematology and Oncology" of the German Society for Hematology and Medical Oncology (DGHO), Middeke is convinced that the use of deep learning represents a significant advancement that can not only increase efficiency but also improve the accuracy of diagnoses. "This will hopefully lead to an improved treatment of patients in the long run," he says.

As an example, he cites the study "Deep learning detects acute myeloid leukemia and predicts NPM1 mutation status from bone marrow smears," conducted by Jan-Niklas Eckardt and himself together with Prof. Martin Bornhäuser, published in the renowned journal Leukemia, which deals with the preliminary examination known as cytomorphology in AML. This is crucial for the diagnosis, particularly through the assessment of myeloblast counts and their morphology. A frequently mutated gene in AML is Nucleophosmin 1 (NPM1), which occurs in about one-third of cases and plays an important role in disease development. However, the interpretation of cytomorphological data is often very subjective. Therefore, a concept for AI was developed in the study that distinguishes AML cases from healthy samples based on digitized bone marrow images and predicts the NPM1 mutation status precisely by analyzing specific morphological features.

"The results demonstrate the potential of deep learning to derive morphological characteristics that can predict mutation status," concludes Dr. Middeke. "Future studies will focus on other clinically relevant mutations. For machine learning models to be integrated into practice, they must be highly accurate and generalizable, which requires cooperation between doctors and software developers," he says.

Together with his colleague Eckardt and other experts from business informatics, computer science, biology, and medicine, he founded the company Cancilico to develop corresponding AI-powered software tools that make the diagnostic process faster and more accurate.

The award, endowed by the Limbach Group SE, underscores the significance of Dresden as a location for digitalization and the application of AI in medicine.

Study and Method

The study "Deep learning detects acute myeloid leukemia and predicts NPM1 mutation status from bone marrow smears" analyzed 1,251 newly diagnosed AML patients from several multicenter studies and a patient registry, complemented by 236 control samples from healthy bone marrow donors. The samples were prepared according to WHO guidelines and examined for NPM1 mutations. A multi-stage machine learning workflow was developed to segment and classify digital bone marrow images. A human-in-the-loop approach for cell segmentation and various deep learning models were employed to distinguish between AML and healthy samples, as well as to predict NPM1 status.

Publication

Eckardt, JN., Middeke, J.M., Riechert, S. et al. Deep learning detects acute myeloid leukemia and predicts NPM1 mutation status from bone marrow smears. Leukemia 36, 111–118 (2022). https://www.nature.com/articles/s41375-021-01408-w

Contact:

Dr. Jan Moritz Middeke
University hospital Carl Gustav Carus Dresden

^{National Center for Tumor Diseases Dresden (NCT/UCC)
The National Center for Tumor Diseases Dresden is a joint institution involving the German Cancer Research Center (German abbreviation: DKFZ), the Faculty of Medicine at TU Dresden, the Carl Gustav Carus University Hospital Dresden and the Helmholtz-Zentrum Dresden-Rossendorf (HZDR). The NCT Dresden is part of NCT, which has sites in Berlin, Heidelberg, SouthWest (Tübingen-Stuttgart/Ulm), WERA (Würzburg, Erlangen, Regensburg, Augsburg) and West (Essen/Cologne).
The NCT has made it its mission to link research and patient care as closely as possible. It is a long-term collaboration between the German Cancer Research Center (German abbreviation: DKFZ), excellent partners at the university medical center, and other outstanding research partners at various locations in Germany. The NCT aims at translating innovations in cancer research in Germany into studies in a targeted and rapid manner, with the aim of diagnosing and treating cancer while maintaining a high quality of life. In doing so, patients are research partners on an equal footing.
The Dresden center draws on the structures of the University Cancer Center Dresden (UCC), which was founded in 2003 as one of the first Comprehensive Cancer Centers (CCC) in Germany. Since 2007, the Dresden center has repeatedly been awarded the title of “Oncological Center of Excellence” by the German Cancer Aid (German abbreviation: DKH).}

^{Else Kröner Fresenius Center (EKFZ) for Digital Health
The EKFZ for Digital Health at TU Dresden and University Hospital Carl Gustav Carus Dresden was established in September 2019. It receives funding of around 40 million euros from the Else Kröner Fresenius Foundation for a period of ten years. The center focuses its research activities on innovative, medical and digital technologies at the direct interface with patients. The aim here is to fully exploit the potential of digitalization in medicine to significantly and sustainably improve healthcare, medical research and clinical practice.}