Until now, residual leukemia cells have been detected using flow cytometry. While highly effective, this method is time-consuming and requires considerable expertise. Researchers at Amsterdam UMC therefore investigated whether part of the routine clinical analysis in hematology could be automated using an AI algorithm they developed. Their study shows that this is indeed possible: the analysis can be completed in just three seconds. In addition, unlike manual analysis, AI always provides a consistent result.
Reliable predictor
The algorithm predicts relapse just as effectively as the traditional manual method. This was demonstrated in a study by Tim Mocking and Costa Bachas from Jacqueline Cloos’ team, involving 399 AML patients. Mocking explains: “Patients in whom we detected residual leukemia cells relapsed at roughly the same rate, regardless of whether we used the manual or AI method. This shows that the predictive value of AI analysis is comparable to the current standard. It marks an important first step toward clinical implementation of AI.” Further validation studies are required before the algorithm can be applied in daily practice.
Time savings
Manual analysis is labor-intensive and always requires review by two analysts to minimize errors. At the same time, new equipment at Amsterdam UMC will generate increasingly larger datasets in the future, making manual analysis less practical.
“Thanks to this AI innovation, analyses can now be performed more quickly and efficiently, freeing up time for other important tasks,” says Mocking. “For example, analysts can devote more time to specialized analyses that require their expertise, or to performing additional measurements. Moreover, the AI algorithm may eventually be able to detect new cell types related to the disease—cells we don’t yet recognize.”
The findings of Tim Mocking and colleagues have been published as a ‘Letter’ in the leading scientific journal Leukemia: Computational measurable residual disease assessment in acute myeloid leukemia: a retrospective validation in the HOVON-SAKK-132 trial | Leukemia