Cellular Rhythm for Label-Free Tracking of Apoptotic Progression in Leukemia Cells

Leukemia is a hematological malignancy originating from hematopoietic stem cells with diverse clinical subtypes. Accurate assessment of apoptosis is integral to leukemia research and therapeutic evaluation, providing the foundation for precise diagnosis and optimized therapeutic strategies. Conventional techniques, including flow cytometry and immunofluorescence, have played a pivotal role in advancing apoptosis research. However, their dependence on staining and antibody labeling renders them labor-intensive and time-consuming and introduces the risk of cytotoxicity. This study proposes a label-free detection strategy for leukemia cells based on cellular rhythmicity, implemented through a scattering interferometric microscopy system that enables interference between coverslip-reflected and cell membrane-scattered light. Nanometer-scale membrane displacements, driven by cytoskeletal traction, membrane coupling, and transmembrane ion fluxes, modulate scattering intensity, which is reflected in the interference signals as temporal rhythmic patterns indicative of apoptotic states. Rhythmic patterns were extracted using the Farneback dense optical flow algorithm, while spectral features were obtained from snapshot hyperspectral imaging. These two types of features were integrated through a dual-stream neural network to enable label-free identification of apoptosis progression in multiple leukemia cell types. Experimental results demonstrated a classification accuracy of 95.45% across four leukemia cell lines (K562, MEC-1, Jurkat, and THP-1) under three apoptotic states (viable, early apoptosis, and late apoptosis), with a per-cell detection time of 160 ms. This study proposes a label-free approach that leverages cellular rhythmicity for the recognition of apoptosis progression, providing a novel paradigm for leukemia research and clinical evaluation.