A Sparse Constrained Optimization Method for Resolving Coincident Single-Cell Events in Microfluidic-Based Impedance Sensing.

Abstract

Label-free electrical impedance-based single-cell detection has been widely applied in cell sorting, electrical phenotyping, and monitoring of cell growth status. However, when high-concentration cell suspensions pass through the sensing region simultaneously, coincident events frequently occur, which leads to inaccurate segmentation of cell events and distorted identification of single-cell waveforms. As a result, statistical errors in electrical phenotyping are introduced. In this work, we propose a two-step sparse-constrained optimization algorithm based on $\ell _{1}$-norm regularization, which addresses this challenge without requiring any structural modification to the microfluidic chip. The raw signal is processed using this two-step framework: first, a waveform detection dictionary is constructed to segment the signal; subsequently, a de-coincidence dictionary is applied to resolve coincident waveforms. Experimental validation on synthetic data streams demonstrates robust counting accuracy from 2×10⁵ to 5×10⁶ particles/ml (99.9%-98.4%), with only a 5.1% reduction under five levels of additive noise at 2×10⁶ particles/ml. Analysis of polystyrene beads of two sizes and T cells at three concentrations demonstrates enhanced size discrimination, improved statistical accuracy, and consistent counting performance compared with conventional algorithms. The proposed method effectively segments and decomposes coincident signals into individual cell events by employing sparse optimization techniques. This algorithm is well suited for applications that demand accurate counting and classification of cell/particle suspensions across a wide concentration range.