New AI Tool Improves Blood Cancer Diagnosis
Posted on 16 Jun 2025
Traditionally, diagnosing blood and bone marrow cancers has involved doctors manually examining and classifying hundreds of cells under a microscope—an approach that is time-consuming but essential for identifying disease and determining cancer stage. A newly developed artificial intelligence (AI) tool now offers a way to automate this process, according to findings published in Science Translational Medicine.
Developed by researchers at Memorial Sloan Kettering Cancer Center (MSK, New York, NY, USA) along with collaborators, the AI tool, named DeepHeme, was trained on nearly 50,000 labeled digital cell images. It was later evaluated using previously unseen cases, where it demonstrated performance that matched or surpassed that of expert pathologists. DeepHeme achieves expert-level diagnostic accuracy and reduces the time needed to analyze blood and bone marrow slides from over 30 minutes to just seconds. It is capable of evaluating both types of samples and has potential applications in advancing personalized medicine. Additionally, DeepHeme may support the discovery of new morphology-based biomarkers—visual cellular features that can aid in diagnosing disease, monitoring treatment responses, and forecasting disease progression. Identifying such biomarkers is key to furthering personalized treatment strategies.
At MSK, teams of researchers and technologists are actively exploring how AI can be leveraged to enhance patient care. These tools are designed not to replace clinical expertise but to augment it by increasing the speed and precision of data interpretation and enabling pattern recognition across large patient populations and disease variations. AI is already proving helpful in areas like mammogram interpretation, processing complex genomic data, and analyzing patient-reported treatment side effects. DeepHeme is part of a broader initiative to establish MSK as the first fully digital hematopathology service in the U.S., transitioning from physical glass slides and microscopes to digital formats. Beyond enabling AI use, digitized slides offer other advantages—they can be instantly reviewed from any location, eliminating the need for physical transport.
Digitized slide images can also be stored in a patient’s medical records. If follow-up tests are required, even years after the initial biopsy, clinicians can instantly access and compare past and present slides without retrieving physical specimens from archives. In addition to its clinical applications, digital pathology data supports research efforts such as MSK’s Cancer Data Science Initiative. This project collects millions of anonymized data points in near real time from patients receiving treatment at MSK. The resulting vast dataset provides researchers with valuable insights into which therapies are most effective for specific patient profiles and scenarios. Though the data is de-identified, the findings can help refine treatment decisions for both current and future patients. Following further validation, MSK intends to introduce DeepHeme into clinical use and may extend its availability by licensing it to other healthcare institutions.
“DeepHeme is just one example of MSK’s broader digital transformation,” said physician-scientist Ahmet Dogan, MD, PhD, Chief of MSK’s Hematopathology Service. “Now that we have high-resolution scans of every slide, we can use them to learn more about blood cancers in general.”
Related Links:
MSK
