Clustering and Classification of Fungal Cells and PMMA Microparticles: Unsupervised and Supervised Learning using K-Means, PCA, Logistic Regression and Spiking Neural Networks on Event-Based Cytometry Datasets

Abstract

Imaging flow cytometry (IFC) is a powerful analytical technique used for rapidly categorizing small cells and micro-particles. Unlike traditional microscopy, IFC instruments handle a high throughput of cells. Typically, photodetectors, photomultiplier tubes, or high-speed frame-based cameras are employed for this task. This study explores the potential of neuromorphic cameras, also known as event-based sensors (EVS), as a detection mechanism for cytometers. While previous work focused on PMMA microparticles, this investigation centers on fungal cells. Moreover, while earlier research leaned towards supervised learning algorithms like logistic regression and spiking neural networks, our approach in this paper employs k-means, an unsupervised learning paradigm. We demonstrate that training such a simple algorithm in conjunction with PCA achieves 100% classification accuracy without relying on training labels.