Microbial Cell-Adaptive Segmentation in Quantitative Phase Imaging

Abstract

Quantitative phase imaging (QPI) enables label-free microscopy with exceptional signal-to-background contrast and provides direct measurements of key biophysical parameters such as cell mass. Despite these advantages, QPI lacks a generalized, accurate, and computationally efficient segmentation method, limiting its use in high-throughput single-cell biology. Here, we introduce microbial Cell-Adaptive Segmentation (mCAS), a purpose-built framework for QPI. mCAS is an object-based, modular pipeline that leverages the robust phase-edge signal to achieve accurate segmentation without Fourier filtering, deconvolution, or deep learning. It runs efficiently on standard hardware, is readily implemented in common software platforms, and requires no curated training data. mCAS consistently outperformed both intensity-based and deep learning approaches, achieving over 98% accuracy with up to tenfold faster processing. By eliminating the segmentation bottleneck, mCAS enables scalable, high-fidelity analysis of QPI datasets and provides a broadly applicable foundation for label-free, high-throughput single-cell biology. We validate its generality across diverse bacterial species, morphologies, and intracellular complexities. Significance StatementQuantitative phase imaging (QPI) enables label-free, noninvasive measurement of cellular biophysics, yet its broader adoption has been limited by the absence of a generalized and computationally efficient segmentation strategy. We present microbial Cell-Adaptive Segmentation (mCAS), a framework purpose-built for QPI that achieves high accuracy without training, specialized hardware, or intensive preprocessing. mCAS generalizes across microbial species and morphologies, outperforming state-of-the-art deep learning approaches while processing thousands of single-cell images in minutes. By eliminating the segmentation bottleneck, mCAS makes QPI broadly accessible for high-throughput, label-free single-cell biology and expands the computational toolkit for quantitative imaging in microbiology.