Magnetic Bead Conjugated Lung Tumor Cell Binding Efficiency Assessment Based on Deep-Learning Approach

Accurate quantification and isolation of circulating tumor cells (CTCs) are critical for early cancer detection and treatment. Magnetic beads (MBs) have emerged as a promising non-invasive and effective method for capturing CTCs. The success of CTC capture primarily relies on the binding efficiency of MBs to these rare cells through antibody-antigen or aptamer-receptor interaction. This study proposed a deep learning-based approach to automatically evaluate the binding efficiency of the magnetic beads immobilization on lung cancer cells (A549 cell line). This investigation studied two different sizes of magnetic beads (4.5 µm and 3.0 µm) to understand their effect on the binding process between cells and beads. The methodology leverages a combination of object detection and semantic segmentation deep learning models to precisely detect the A549 cells in the acquired microscopic image data and calculate the area of MBs binding to each cell. The object detection model accurately classifies unbound cells, MBs-bound cells, and unbound MBs. Each MBs-bound cell image is then cropped and subjected to further analysis by the semantic segmentation model to delineate cell and binding MBs regions. Integrating these two deep learning models streamlines the evaluation process, reducing manual intervention and offering an automated and efficient solution for quantifying magnetic bead (MB) binding. This approach can potentially serve as a reliable signal readout mechanism in a biochip designed for CTC detection and analysis.