ARTIFICIAL INTELLIGENCE IN FLOW CYTOMETRY FOR THE DIAGNOSIS OF ACUTE MYELOID LEUKEMIA: DIAGNOSTIC PERFORMANCE, ANALYSIS TIME AND CLINICAL FEASIBILITY

Abstract

Acute myeloid leukemia (AML) is an aggressive clonal neoplasm whose incidence is rising with population aging and whose diagnosis depends on expert interpretation of flow cytometry. This study aimed to analyze the performance of artificial intelligence (AI) models applied to automated flow cytometry analysis in the diagnosis of AML, considering accuracy, processing time and feasibility of implementation. This is a narrative review, structured according to the SANRA scale, with searches in the PubMed and LILACS databases in March and July 2026, which included five original studies published between 2018 and 2026. Sample sizes ranged from 200 to 5,333 and included comparisons with non-neoplastic conditions, healthy individuals and other acute leukemias. The algorithms analyzed samples in 7 to 84 seconds, compared with 15.6 to 20 minutes for a hematologist, and achieved a specificity of 99.79% and a precision of 92.31% in detecting AML in good-quality samples, as well as an accuracy of 90% in detecting minimal residual disease. AI is a high-performance support tool that does not replace the specialist, and its adoption in Brazil requires local validation, especially in the public health system.