Machine learning for image-based cell classification, detection, and segmentation in microfluidic biosensors: A computer vision perspective

Abstract

Microfluidic biosensors, when combined with microscopy imaging, provide continuous observation of cells during growth and division, generating vast quantities of visual data. The integration of machine learning (ML) enables real-time analysis of these images, enhancing the monitoring of cell behavior and dynamic changes. This offers critical insights for cell biology and disease research, representing a typical problem in computer vision (CV). This paper specifically focuses on cell analysis tasks, including cell classification, detection, and segmentation, within the microfluidic biosensor framework. We review the application of ML techniques for these three core tasks, discussing both traditional methods and end-to-end deep learning (DL) models. Emphasizing how ML improves the accuracy and efficiency of microfluidic biosensor imaging, we highlight the growing potential for personalized medicine, disease diagnosis, and drug development. Finally, we analyze current technological trends and propose recommendations for future research in cell analysis using microfluidic biosensors.