Research On Automatic Segmentation Of Bladder Wall OCT Images

OCT (Optical Coherence Tomography) is a biomedical imaging method using low-coherence interferometry to produce an image from biological tissues with the characteristics of real-time, high resolution and harmless to biological tissues. OCT has been widely used in medical imaging area. In this paper, bladder wall images based on OCT system have been segmented. The thickness of bladder wall is crucial for the study and diagnosis of bladder diseases. Two effective methods for three-layer bladder wall OCT images have been proposed in this study.The background of OCT system, segmentation methods based on OCT images and the research status has been introduced in this paper. Manual segmentation has been employed in previous bladder wall segmentation studies. However, due to the time consuming and subjective nature of manual segmentation, two automatic segmentation algorithms have been proposed in this study. By using method based on graph theory and dynamic programming, boundaries of each layer are detected clearly. An Image is presented by a graph of nodes; appropriate weights are assigned on each edge; a path travels across the graph form the start node to the end node with the minimal total weigh sum is the preferred cut. By using three thresholds technology applied to canny detector, the accuracy of edge information has been improved and the useful edge information missed due to high threshold has been retained. Meanwhile, the axial gradient information with higher anti noise property than light intensity only is used as a parameter. Good result has been presented by using curve fitting to optimize the boundaries.The result has shown that the segmenting process time has been reduced enormously by comparing to the result of traditional methods. Consequently, two effective, high-accuracy and high anti noise automatic algorithms have been proved.OCT system is non-invasive, non-radiation with high resolution. It has a very impressive prospect in field of bladder imaging.