Artificial Intelligence Enables the Label-Free Identification of Chronic Myeloid Leukemia Cells with Mitochondrial Morphological Alterations

Abstract

Long-term tyrosine kinase inhibitor (TKI) treatment for patients with chronic myeloid leukemia (CML) causes various adverse events. Achieving a deep molecular response (DMR) is necessary for discontinuing TKIs and attaining treatment-free remission. Thus, early diagnosis is crucial as a lower DMR achievement rate has been reported in high-risk patients. Therefore, we attempted to identify CML cells using a novel technology that combines artificial intelligence (AI) with flow cytometry and investigated the basis for AI-mediated identification. Our findings indicate thatBCR-ABL1-transduced cells and leukocytes from patients with CML showed significantly fragmented mitochondria and decreased mitochondrial membrane potential. Additionally,BCR-ABL1enhanced the phosphorylation of Drp1 via the mitogen-activated protein kinase pathway, inducing mitochondrial fragmentation. Finally, the AI identified cell line models and patient leukocytes that showed mitochondrial morphological changes. Our study suggested that this AI-based technology enables the highly sensitive detection ofBCR-ABL1-positive cells and early diagnosis of CML.