Abstract
Cellular phenotypes are traditionally identified using biomolecular markers, such as cell-surface proteins and quantified by fluorescence-based flow cytometry. Despite their effectiveness, these approaches rely on extensive sample preparation, can influence cell viability, and frequently fail to resolve rare or phenotypically dynamic subpopulations such as metastatic cancer cells, drug-resistant cells, and immune cells undergoing dynamic state transitions due to the lack of reliable molecular markers. In contrast, intrinsic cellular biophysical properties, including size, deformability, electrical physiology, and organelle structure, are closely linked to cellular function and disease state, and can be measured in a label-free manner with minimal perturbation. Leveraging these properties provides an opportunity to enable high-throughput, viable single-cell analysis and separation beyond the restrictions of marker-based cytometry.
This dissertation develops and integrates impedance-based biophysical cytometry platforms with microfluidic deformation and separation techniques to enable multiparametric, label-free characterization and monitoring of heterogeneous cell populations. In Chapter 1, the scientific background and motivation for this work are presented, including an overview of biophysical cytometry, impedance-based single-cell analysis, microfluidic deformation, and machine-learning approaches relevant to this dissertation. As Aim 1 (Chapters 2–4), impedance signal analysis is advanced for multiparametric biophysical cytometry by combining frequency-resolved electrical measurements with viscoelastic microfluidic flows to quantify cell size, viability, deformability, and recovery dynamics at high throughput. A neural network-based impedance signal templating framework is introduced to extract deformability metrics directly from raw temporal impedance signals acquired during controlled extensional deformation, enabling robust, size-independent quantification of cell shape anisotropy. These electrical deformability metrics are validated against image-based measurements and integrated with impedance-derived electrical physiology to classify biologically relevant cell states, including pancreatic cancer cells across epithelial-mesenchymal transition and cutaneous T-cell lymphoma subtypes. Deep learning approaches further demonstrate improved classification performance and robustness compared to conventional feature-based methods.
As Aim 2, (Chapters 5–6), impedance-based biophysical cytometry is integrated downstream of active and passive microfluidic separation techniques, including dielectrophoresis and deterministic lateral displacement, to enable inline monitoring and phenotypic validation of separated cell populations. These platforms provide real-time assessment of cell viability, size distribution, and electrical phenotype before and after separation, allowing optimization of separation conditions and evaluation of separation-induced cell damage or bias.
Collectively, this work establishes a scalable framework for impedance-based biophysical cytometry that bridges single-cell measurement, machine learning-enabled analysis, and microfluidic separation. The established approaches enable high-throughput, label-free identification and monitoring of cancer and immune cell subpopulations, and position biophysical cytometry as a complementary and enabling tool for functional single-cell analysis and downstream therapeutic applications.
