AI for blood‑smear review

Reading a blood smear still often means a human scanning a stained slide by eye, and a 2024 paper described an artificial-intelligence workflow built to automate more of that review. (mdpi.com) The paper, published January 25, 2024 in the *Journal of Molecular Pathology*, came from researchers at the Medical University of Vienna and KML Vision GmbH in Graz, Austria. The team used the web-based platform IKOSA to segment and identify cells in DAPI-Giemsa co-stained blood smears. (mdpi.com) A blood smear is a thin layer of blood spread on glass and stained so different cells stand out under a microscope. Pathologists and laboratory staff use it to inspect white blood cells, red blood cells, and platelets for changes linked to infections, leukemia, anemia, and other disorders. (ashpublications.org; pmc.ncbi.nlm.nih.gov) That work is still heavily manual in many settings. Reviews of peripheral blood smear practice describe the process as tedious, time-consuming, labor-intensive, and dependent on trained experts, even as automated analyzers handle much of routine blood counting. (pmc.ncbi.nlm.nih.gov; pmc.ncbi.nlm.nih.gov; read.qxmd.com) The IKOSA study said its system could automatically detect and classify neutrophils, lymphocytes, eosinophils, monocytes, erythrocytes, and platelets, including young and segmented neutrophils. The authors wrote that this went beyond many earlier algorithms that focused on only one blood-cell type. (mdpi.com) The system also measured cell shape and internal texture, not just cell counts. The paper said those quantitative features included entropy and gray-level co-occurrence matrix metrics, which are image-analysis methods used to capture how pixel patterns vary inside a cell. (mdpi.com) That matters because smear review is still used when speed and morphology both count. The American Society of Hematology’s *Blood Advances* said morphologic evaluation can support timely diagnosis and rapid treatment for findings such as intracellular parasitic infections, hemolytic anemias, suspected thrombotic microangiopathies, and acute leukemias. (ashpublications.org) Giemsa staining also remains central far beyond hematology. The World Health Organization says microscopy of Giemsa-stained blood films is a recommended and reliable procedure for malaria diagnosis, where correct slide reading affects parasite detection, species identification, and parasite counts. (who.int; who.int) The Vienna-Graz paper did not say artificial intelligence should replace expert review. It concluded that automated evaluation could support routine diagnostics by adding quantitative shape and structure parameters to classical leukocyte counts. (mdpi.com) That leaves the near-term use case looking more like screening and triage than a fully autonomous microscope: software sorting slides, flagging unusual cells, and sending the hardest calls back to specialists. (mdpi.com; pmc.ncbi.nlm.nih.gov)