CCNet: a segmentation network based on U-Net for calculating cell confluency

Cell confluency is a critical indicator for determining the optimal collection time of cultured stem cells in vitro. Currently, assessing cell confluency heavily relies on the subjective experience of researchers, which can introduce significant variability. To achieve a more accurate calculation of cell confluency, we proposes an image segmentation network——cell confluency network (CCNet), based on the U-Net architecture. CCNet incorporates attention modules and a residual structure, thereby improving the network’s learning capacity and accuracy. By calculating the pixel ratio (the ratio of cell pixels to total pixels) from the images processed by CCNet, the cell confluency value is obtained. Experiments indicate that the proposed CCNet provides more accurate segmentation results for stem cell images compared to the previous version of U-Net, thereby demonstrating the effectiveness of CCNet.