Algorithm Research On Accurate Kidney Automatic Segmentation And Accurate Kidney Tracking In The Minimally Invasive Surgery Guided By Real-time Ultrasound And MRI

Recent years, advances in transducer design have led to the great improvement ofspatial/temporal resolution of ultrasound images. Improvement of the image quality makeultrasound applied in not only its traditional area of diagnosis, but also image-guided kidneyminimally invasive therapy. However, the intrinsic low SNR property of the ultrasoundimages makes the ultrasound image segmentation and tracking a challenge job.During the proposing process of new image segmentation and tracking algorithms,many people apply the level set segmentation in the field of ultrasound image segmentationand tracking. Since the level set method was firstly proposed by Osher and Sethian, peopleintroduce many models in this framework, such as texture model, gray scale, statistics model,Markov random field model, shape model etc. The introduction of model is mostly toovercome the bad influence of the speckle noise in the ultrasound image. Due to the low SNRproperty of the ultrasound image, many level set methods with complicated model often getlocal minimum and have low robust property.Through a lot of experiment, we propose novel ultrasound image segmentation andtracking algorithms to meet clinical need. To solve ultrasound kidney segmentation problem,our algorithm framework combined US denoising, US segmentation and shape prioroptimization. Compared with common level set method based on shape prior, our algorithmonly find the similar shape once in the whole algorithm framework, leading to the reductionof the cost time of the segmentation. And setting the finding of the similar shape in the outsideof level set iteration makes our algorithm enjoy good convergence property. To solve theultrasound kidney tracking problem, our algorithm framework combined histogram searchand our US segmentation. Compared with common tracking algorithms, our algorithm havebetter location robust property by histogram search. And, the application of ultrasoundsegmentation helps us get accurate contour. We have demonstrated the effect of our methodthrough qualitative and quantitative experiment.Our new method can make the registration between US image and other high solutionimage become possible. Furthermore, it can lead the application of the US image-guidedkidney minimally invasive therapy. It will be meaningful to the field of US image-guidedtherapy and teaching and become a breakthrough in this field. This way not only reduce therisk of the surgery, improve the security and efficient of the surgery, but also reduce the cost,owing significance to the national health.