Lens-free microfluidic-based method for detecting circulating tumor cells

Accurate identification of Circulating tumor cells (CTCs) can provide valuable information for early diagnosis and therapeutic monitoring of cancer. Here, we present a method of lens-free imaging combined with microfluidic technology to accomplish label-free classification of CTCs and blood cells (BCs). In this approach, Cell images in a microfluidic chip are recorded and classification of the images is accomplished using a Yolo-V4 convolutional neural network. The recognition precision of CTCs and BCs reached 99.21% and 98.26%, respectively. The 8 microfluidic channels used allow a high throughput of 70,000 cells per minute without clogging. It can be seen that the method has great advantages in detecting rare cells.