A Semi-automatic Cell Boundary and Adherens Junction Layers Detection Approach Using Level Set Method
MetadataShow full item record
Cell boundary detection is a very important task in bioinformatics. It is the basis of plenty of future work towards cell analysis. The connection parts of different cells are call adherens junctions. Recent work on detecting junctions has the ability to locate them with high accuracy but fails to draw the whole layer. In this thesis, cell boundary and adherens junctions detection of two-cell clusters from several image dataset of MDCK cells is finished by using a semiautomatic approach with the classic contour evolving technique, the level set method, which is able to emphasize the whole junction layer. A data driven stopping criteria is studied to stop the level set contour evolving. Besides level set method, threshold and filter techniques and noise removal are added to help improve the performance. To divide cluster into two cells, a method using linear regression is used to remove noises so that level set method can be performed within clusters. In order to evaluate the accuracy of this method, the results from proposed method are compared with manually drawn boundaries by calculating accuracy and Dice coefficient, showing high precision rates.