Application of artificial neural networks for automatic measurement of micro-bubbles in microscopic images

Abstract

A novel approach for quantitative segmentation and measurement of oxygen microbubbles in microscopic images is presented. In this approach, ellipse-based models were first built using moment parameters as rough approximations of oxygen microbubbles. Artificial neural networks were then developed and trained for segmentation refinement. The results show that the proposed approach achieved high accuracy of microbubbles measurement with less than 8% measurement error.<>