Research Of Navigation System Based On Image-Guided Percutaneous Surgery Robot

With the advantages of less invasiveness and faster recovery,minimally invasive surgery gradually becomes a hot spot in the field of malignant tumor treatment in recent years.The stability and accuracy of a puncture can be improved by applying a robot to the puncture.The thesis finishes research of a puncture surgery robot navigation system.The robot mainly focuses on lung cancer puncture surgery.A method based on static CT images and dynamic DR images can be used for localization of lung lesions;puncture needle identification and azimuth tracking during surgery are realized.The main contents of this thesis include the following parts:(1)Pulmonary lesions segmentationWe transform the three-dimensional data segmentation problem into two-dimensional continuous image segmentation problem.The two-dimensional full-convolution neural network is used to segment each slice of CT images.We have performed some simulation experiments to verify whether the network can segment pulmonary lesions,and the results show that this segmentation network can accurately segment the pulmonary lesions in most 2D images.(2)2D/3D registrationThe object of 2D/3D registration is to get the spatial position relationship between the two-dimensional lung X-ray image and the three-dimensional lung CT image.In this paper,we extended the single image registration method to double image registration method.We have performed some simulation experiments to verify the effect of the registration method,and the experimental results show that the registration method can obtain accurate spatial position relationship between X-ray image and CT image by using two images at any angle.(3)Localization of pulmonary lesions and tracking of puncture needle in X-ray imageFor the localization of lung lesions,according to the results of regional segmentation of lung lesions and the registration results of X-ray images and CT images,the lung lesion area in CT images was projected onto the X-ray images by the triangular similarity principle.We have performed some experiments to verify the result of localization,and the experimental results show that the method can correctly locate the lung lesions in X-ray images.For the recognition and tracking of puncture needle,we used a semi-automatic algorithm based on Hough line detection and prior knowledge.We have perform some experiments on offline videos to verify the proposed algorithm.Simulation experiments show that the algorithm can correctly track the position of the puncture needle and draw the extension line according to the reference line originally marked by human.