Machine Learning-Enhanced Microfluidic Impedance Platform for Rare Cell Analysis

Rare cells, despite constituting only a small fraction of the population, play a critical role in health and disease. For instance, a minor subset of cancer cells can drive tumor initiation and progression. However, detecting and analyzing rare cells remain challenging due to their low abundance and the need for high-precision identification methods. Microfluidic impedance-based flow cytometry (μIFC) provides a label-free and highly informative approach for single-cell characterization. In this study, we present a multifrequency μIFC platform enhanced by a support vector machine (SVM) learning strategy to accurately distinguish rare cancer cells from white blood cells (WBCs). Our platform achieves over 99% accuracy in differentiating three cancer cell lines─MDA-MB-231 (breast), A549 (lung), and HeLa (cervical)─from human immortalized T lymphocytes (Jurkat) or peripheral blood mononuclear cells (PBMCs). To further improve detection reliability, we introduce a postprediction correction strategy that reduces false identification rates. Additionally, we demonstrate the platform's capability to detect rare cancer cells within WBC populations at concentrations of 1-10%, with results highly consistent with conventional flow cytometry. This work establishes a robust, label-free, machine learning-enhanced μIFC platform for rare cancer cell analysis, paving the way for broader applications in rare cell characterization.