Supervised machine learning in microfluidic impedance flow cytometry for improved particle size determination

Abstract

The assessment of particle and cell size in electrical microfluidic flow cytometers has become common practice. Nevertheless, in flow cytometers with coplanar electrodes accurate determination of particle size is difficult, owing to the inhomogeneous electric field. Pre-defined signal templates and compensation methods have been introduced to correct for this positional dependence, but are cumbersome when dealing with irregular signal shapes. We introduce a simple and accurate post-processing method without the use of pre-defined signal templates and compensation functions using supervised machine learning. We implemented a multiple linear regression model and show an average reduction of the particle diameter variation by 37% with respect to an earlier processing method based on a feature extraction algorithm and compensation function. Furthermore, we demonstrate its application in flow cytometry by determining the size distribution of a population of small (4.6 ± 0.9 μm) and large (5.9 ± 0.8 μm) yeast cells. The improved performance of this coplanar, two electrode chip enables precise cell size determination in easy to fabricate impedance flow cytometers.

When microplastics meet electroanalysis: future analytical trends for an emerging threat

“…157 For electroanalytical methods, there is a strong need for high-throughput real-time signal processing methods; in particular, the use of ML is of great interest to researchers, especially for the analysis of signals obtained with IFC. [158][159][160][161] While most of the studies are applied to the determination of electrical parameters of blood cells, cancer cells and bacteria, the use of these tools for the analysis of MPs has not yet been investigated. The analysis performed with IFC coupled with ML tools shows a promising future for the analysis of MPs for highthroughput applications.…”

Enhancing single-cell biology through advanced AI-powered microfluidics

“…AI models in single-cell biology aim to extract cellular features from different modalities, such as fluorescence microscopy images, 8 holographic flow cytometry images, 9 electrical signals, 10 RNA-seq, 11 or a combination of these. 12 According to whether the data are labeled, the model's training method is divided into supervised and unsupervised.…”

Recent Approaches to Design and Analysis of Electrical Impedance Systems for Single Cells Using Machine Learning

“…Considering cytometric measurement limitations, several papers in recent years have worked to establish correction factors to monitor cell location during measurement and improve the classification of particles based on positional compensation. These methods can either rely on the peak amplitude and spacing properties of the time domain cytometric measurement [31] or extracted parameters calculated from the initial measurements, such as opacity [28,30]. For these methods, the accuracy of the model is typically defined as the closeness to the distributive values of the measured parameters.…”