Deep Learning Assisted Mechanotyping of Individual Cells Through Repeated Deformations and Relaxations in Undulating Channels

Abstract

Mechanical properties of cells are important features that are tightly regulated, and are dictated by various pathologies. Deformability cytometry allows for the characterization of mechanical properties of hundreds of cells per second, opening the way to differentiating cells via mechanotyping. A remaining challenge for detecting and classifying rare sub-populations is the creation of a combined experimental and analysis protocol that would assure classification accuracy approaching 100%. In order to maximize the accuracy, we designed a microfluidic channel that subjects each cell to repeated deformations and relaxations. We also track the shape dynamics of individual cells with high time resolution, and apply sequence-based deep learning models for feature extraction. HL60 cells with and without treatment with cytochalasin D (cytoD), a reagent previously shown to perturb the actin network, were used as a model system to understand the classification potential of our approach. Multiple recurrent and convolutional neural network architectures were trained using time sequences of cell shapes, and shown to achieve high classification accuracy based on cytoskeletal properties alone. The best model classified the two sub-populations of HL60 cells with an accuracy of 95%. This work establishes the application of sequence-based deep learning models to dynamic deformability cytometry.