Image-based machine learning quantitative evaluation of bead-cell binding interaction

• Published in: Sensors and actuators. A. Physical. 2025-02, Vol.382, p.116123, Article 116123
• Main Authors: Phan, Hoang Anh, Thi Nguyen, Anh, Do Quang, Loc, Bui Thanh, Tung, Jen, Chun-Ping, Chu Duc, Trinh
• Format: Article
• Language: English
• Subjects: Bead-cell interaction; Bright-field optical microscopy; Image processing; Machine learning
• ISSN: 0924-4247
• DOI: 10.1016/j.sna.2024.116123

This study presents an image-based method for detecting magnetic beads (MBs) attached to biological cells using A549 lung cancer cells as a case study to demonstrate its effectiveness. The approach leverages magnetic beads of different sizes to evaluate binding interactions employing advanced imaging techniques such as bright-field and fluorescence imaging for precise quantification. The proposed method uses advanced image processing and machine learning to analyze bead coverage on cells, yielding critical data regarding the quality of binding interactions. This coverage value is essential for evaluating the efficiency of immunomagnetic separation processes, particularly in clinical applications where precise cell detection is crucial. The models achieved high accuracy across all three magnetic bead sizes: 1.36 µm, 3.0 µm, and 4.5 µm. The proposed method provides a strong foundation for broader biomedical applications, demonstrating its potential for analyzing various cell-particle interactions in diverse biological systems, further extending its utility to multiple areas of clinical and research settings. [Display omitted] •Proposing a machine learning-based platform for analyzing bead-cell interactions with A549 lung cancer cells.•Implementing automated image segmentation with YOLOv8 and K-means clustering for bead-cell quantification.•Evaluating the binding efficiency of magnetic beads of 1.36 µm, 3.0 µm, and 4.5 µm sizes.