Cell splitting using dynamic programming

Cell detection and segmentation is an essential step in many biological studies. Unfortunately, automatic splitting of merged cells continues to be one of the most challenging problems. We present an algorithm for splitting two adjacent cells using geometric analysis as well as a dynamic programming approach to find the optimum path. The algorithm is compared with Al-Kohafi´s algorithm, which uses a graph-cut method to split the merged cells, and Dzyubachyk´s algorithm which uses a three dimensional Radon transform. The performance was analyzed using images from Dzyubachyk´s dataset. The results show a mean improvement of 27.2% versus the other two methods.