Cell Segmentation Using Level Set Method

In this thesis, the level set method and its application in cell segmentation are investigated. The main task is to segment images with clustered cells provided by FLLL-Hagenberg. Starting with the traditional level set method, the formulation of the level set function and its property are introduced in Chapter 1. The traditional method appears suitable for segmenting cells that don’t stick together. In Chapter 2, a multiphase level set method is set up to segment images with clustered cells. Combined with the iterative voting method, the multiphase level set method can separate clustered cells into individual cells and calculate the area of each cell. In Chapter 3, the Chan-Vese Model is introduced. Instead of solving n PDEs (Partial Differential Equations) as in the previous method, this model only needs two PDEs. All models introduced in the thesis are implemented and tested with real cell images. Based on these tests, we draw the conclusion: the multiphase level set method combined with iterative voting method is best suited for segmenting given images with clustered cells. The drawbacks of this method are the time complexity and the space complexity. The Chan-Vese model reduces the operation time and can deal with images that do and do not contain edges. However, it has some limitations in cell segmentation because of the sub-structures present inside cells. A preprocessing method is proposed in this work; better suited preprocessing methods are still under research.