Machine Learning for Identification of Cancer Cells Based on Electrical Property Using Microfluidic Impedance Flow Cytometry

Abstract Microfluidic impedance flow cytometry (IFC) has become an essential tool for cell analysis, providing a rapid and non-invasive approach to the early diagnosis of cancer. However, a large amount of data was generated during microfluidic IFC, which requires highly efficient data processing tools. In recent years, machine learning has emerged as an efficient tool for data analysis. Here we present a microfluidic IFC chip combined with an enhanced deep neural network for the detection of cancer cells based on electrical properties. The effectiveness of the enhanced deep neural network is illustrated by achieving a high identification accuracy of 93%, surpassing the 86% accuracy of a conventional fully connected neural network.