A novel framework for automated targeting of unstained living cells in bright field microscopy

This paper presents a unified framework aimed at detecting unstained living cells in bright-field (BF) microscopy and finding suitable microinjection points within their surface. Automatic localization of cells is a critical step in improving the procedure of microinjection that, so far, is still conducted manually by trained operators. This work compares different state of the art image processing approaches and combines them into a novel integrated algorithm. Three techniques, operating at three different focus levels, are described: an anisotropic contour completion (ACC) method, a local intensity variation background-foreground classifier and a threshold based segmentation. Furthermore, we point out the benefits and drawbacks of each method in finding viable injection points. Experiments carried out on real images of 10 to 50µm CHO-K1 adherent cells showed that the combination of these microscopy methods resulted in remarkably high cell identification performance and correct targeting rates close to 97%.