Research On Precise Tumor Location Technology Based On CT Image Guidance

Precise radiotherapy is one of the important methods for treating tumors.The medical robotic arm plays the role of supporting the patient in radiotherapy system,and how to align the patient’s tumor area to the beam stream efficiently and precisely is one of the challenges in the industry.The use of CT images to give the location of the target area under the spatial coordinate system through segmentation and alignment,and guide the movement of medical robotic arm to achieve automatic and accurate positioning of tumor,is of great significance to improve positioning efficiency,reduce preoperative manual intervention,save medical resources and improve radiotherapy effect.At present,the domestic research in the field of CT image guided automatic positioning technology is still in the early stage.In this paper,we study the preliminary localization of tumor location based on improved image pre-processing algorithm and autonomous segmentation network model ADLNet,compensate for tumor errors by improved alignment algorithm,verify the accuracy and feasibility of the above method through experiments,and realize image-guided precise localization.The main research contents and results of this paper are as follows:(1)The composition of CT images is studied in depth and its preprocessing methods are profiled;K-Means clustering algorithm and diffuse filling method based on morphological operations are introduced to preprocess CT images by removing nonessential organs and external interference and retaining important lung features.The common semantic segmentation algorithms based on deep learning are reproduced and improved to provide the theoretical basis for subsequent research.(2)Based on the experimental comparison of network nonlinear expression ability,output feature refinement effect,and perceptual field expansion degree,an autonomous segmentation ADLNet network model integrating the advantages of each module is proposed;it is more accurate than other models in the LIDC-IDIR open dataset segmentation task,and the network migration performance is good.The final Dice similarity coefficient reached 0.9125,the accuracy reached 91.75%,and the recall rate reached 90.12%,which is a big improvement compared with the previous ones,and the preliminary localization of tumor lesions was performed on this basis.(3)Use digital reconstruction radiographic image algorithm to obtain DRR images at any angle;combine the continuous image representation of cubic B-spline curve and the improved algorithm of Parzen histogram estimation.At the same time,the image pyramid algorithm is added in the optimization process to increase the registration speed,Improve real-time performance,and verify the effectiveness of the method through experiments.Compared with the previous algorithm,the displacement accuracy in the X and Y directions of the method in this paper is increased by 17.65%and 13.79%,respectively,and the registration time is further increased by 19.64%,which can better correct the patient’s tumor position deviation during image guidance.(4)Synthesize the above research and design the architecture of the medical robotic arm automatic positioning system within a working framework that is consistent with the actual treatment process.Combine the hardware and software system to develop independent test software and validate it on a test platform in a hospital.The results show that the precise positioning technology based on CT image guidance studied in this paper can meet the requirements of preoperative automatic positioning of medical robotic arm.The research in this paper provides a technical solution for improving tumor positioning accuracy and positioning efficiency in the space coordinate system,and provides a reference for further in-depth research on other surgical robots’ automatic positioning technology.