Abstract
The biophysical and biochemical phenotype of a single cell serves as a comprehensive indicator of its functional state, yet conventional cytometry techniques typically assess mechanical and molecular attributes separately and at limited throughput. This thesis presents the development of a microfluidic platform based on multi-stage constriction channels for real-time, label-free single-cell biomechanical and biochemical phenotyping and sorting.A nonlinear stress–strain model was formulated to relate cell transit velocity and deformation to elastic modulus during passage through constriction channels. This model was embedded in a deep learning-based visual velocimeter capable of tracking cells directly from video recordings. Using this optomechanical chip, EpCAM-positive breast cancer cells were identified with 96.4% accuracy without fluorescent labeling.
Building on this foundation, a parallel three-stage serpentine microchannel architecture was introduced. By coating channel walls with specific antibodies, the system simultaneously resolved single-cell mechanical properties and surface antigen expression levels. To eliminate reliance on microscopy and automate stiffness quantification, differential impedance cytometry was integrated into straight constriction channels. Combined analysis of voltage signals and transit time enabled real-time estimation of cell diameter and elasticity at a throughput of ~10⁴ cells per minute, distinguishing benign MCF-10A cells from malignant MCF-7 and MDA-MB-231 cells with over 95% sensitivity and specificity. Drug response experiments confirmed the platform’s ability to detect cytoskeletal changes induced by cytochalasin D, lysophosphatidic acid, and cetuximab, consistent with conventional assays.
Finally, an automated sorting system was developed, incorporating a neural network for real-time signal feature extraction and pneumatic microvalves for cell routing. The system processed impedance pulses within 1 ms and achieved 79.5% sensitivity in sorting live versus fixed breast cancer cells at a rate of 1000 events per second.
Together, these contributions establish a robust and scalable framework for integrated single-cell cytometry, offering high-throughput, multidimensional analysis with significant potential for cancer diagnostics, drug screening, and personalized medicine.
| Date of Award | 6 Nov 2025 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Hiu Wai Raymond LAM (Supervisor) |