Research And Implementation Of Key Technologies Of Medical Image Aided Diagnosis

With the widespread use of medical imaging examinations such as X-ray photography and computer tomography(CT),the massive data generated by imaging examinations help doctors improve the accuracy of diagnosis,but increase their workload as well.In order to improve the work efficiency and level of doctors,many medical imaging diagnostic systems have been developed at home and abroad.However,the system currently used in China only contains the common 2D image browsing function,but lacks 3D reconstruction post-processing technology.Although some foreign software has corresponding functions,it is expensive and often tied to inspection equipment,which is not helpful to other clinical departments except radiology.Meanwhile,although many scholars now use deep learning for disease detection,most of them stay in the algorithm research stage and have not been applied to actual software.Therefore,in view of the above problems,target detection,image segmentation and 3D visualization technology needed in the medical image diagnosis system will be researched in thie study.In addition,based on the existing medical image browsing software in the laboratory,automatic breast mass detection,lung parenchymal segmentation and 3D visualization as well as dental implant simulation have been designed and achieved.The main work is as follows:1.Using YOLOV3 model as the target detection framework,a deep learn-based breast mass automatic detection module was designed and implemented.In order to facilitate the online training optimization of the target detection model and further improve the detection accuracy,a Windows graphical user interface(GUI)was designed for model optimization training using the newly labeled data set.And the validation indicators of lesion detection are visualized,which can assist researchers in clinical evaluation and research.2.Through the research and analysis of traditional regional growth algorithm and deep learning u-net segmentation algorithm,the automatic segmentation of lung parenchyma is realized by many methods.Combined with the 3D visualization technology of volume rendering,the 3D reconstruction function of lung parenchymal segmentation results is designed and realized,which can effectively assist doctors to determine the size and location of pulmonary nodules and make operation plan.3.In order to effectively simulate the implant surgery process,a dental implant simulation module was designed and implemented on the Activiz.Net visualization platform based on the principles of computer graphics and 3D visualization technology.First,surface reconstruction was used to complete the generation of a panoramic image of the facial jaw position and the delineation of the alveolar nerve tube;then the collision detection technology was used to design the automatic collision detection between the implant and the nerve tube to prevent the implant from being implanted too deeply and causing nerves.Tube damage,which can effectively assist the doctor in determining the dental implant surgery plan.The three newly added functional modules in this article have made up for the lack of image post-processing technology in the application of domestic medical imaging auxiliary diagnosis system to a certain extent.It can effectively promote the development of the medical imaging field towards a more efficient,smarter and more precise direction.