Imaging flow cytometry combined with spatial light modulator

Abstract

Imaging flow cytometry (IFC), an instrument that combines the features of microscopy with image acquisition and microfluidic chips, enables the rapid analysis of single-cell image information at high throughput. Currently, imaging flow cytometers have been integrated with deep learning for classification purposes. However, due to the rich information contained in images, deep learning networks require complex calculations, often resulting in lengthy computation times and substantial energy consumption. In this work, we propose a method that integrates optical neural network (ONN) to enhance classification efficiency in imaging flow cytometry while reducing computational complexity and processing time. We introduce a spatial light modulator (SLM) into the optical system of the imaging flow cytometer to preprocess the forward scattering signals of cells within the optical neural network framework, thereby lowering the subsequent computational burden on images of cells. A comparison of bright-field images of HepG2 and SW480 cells, along with the SLM-processed images, demonstrates a significant improvement in both computational efficiency and accuracy. Following training of 400 samples and 1,000 training epochs, the model achieved an accuracy of 96%.