A Wearable Assistant System And Its Quantitative Diagnosis And Image Segmentation Algorithm For CHD Surgery

The congenital heart disease(CHD)surgery is very difficult.It is mainly manifested in small lesion structure,small surgical field and the difficult identification of intraoperative anatomy.The preoperative clear image data is not fully utilized,and can not be well combined with what the doctors see during the surgery.After full discussion with the doctor,the research work of this article was carried out.In this paper,a wearable surgical assistant system was constructed.The system included the hardware framework and software development of the system,the diagnostic algorithm for the quantitative assessment of aortic coarctation by CT images in the system,and the multi-scale Hessian matrix based cardiac catheter inspection image segmentation algorithm.Firstly,the framework of the wearable surgical assistant system mainly included two components of a wearable device and a image processing platform.This article has carried out development based on 3DMed and 3DSlicer respectively,including modifying the functions and interfaces,adding new features and plugins.In wearable devices,this article used Google Glass,Microsoft Holo Lens,and self-constructed wearable helmets to achieve back-end and front-end connectivity,and integrated 3D image viewing applications and remote experts guidance function on wearable devices.In the system,a diagnostic algorithm for quantitative assessment of coarctation of the aorta by CT images was performed,and a large number of patients with congenital heart disease in the database of the Guangdong General Hospital were included in the database.The author collected CT images,ultrasound reports and the course records.And then defined eight major features and used machine learning methods to train the model.After feature selection and machine learning training,we can finally assess the severity of coarctation of the aorta by CT images.Thirdly,the multi-scale Hessian matrix based cardiovascular angiography images segmentation algorithm in the system was also based on the cardiovascular angiography images of congenital heart disease patients collected from the database of the Guangdong General Hospital.Based on the classic Hessian matrix identification algorithm for tubular structures,this paper used eight-scale Hessian matrix identification algorithm to segment the cardiogram images.Under the guidance of the principle of not missing the small blood vessels as much as possible,the maximum solution of the segmentation results was taken.The system has undergone multiple clinical trials in the Department of Cardiology Surgery of Guangdong General Hospital.The wearable surgical assistant system provided the surgeon with a very comprehensive cardiogram and preoperative diagnostic information.The specificity of the diagnostic algorithm for quantitative assessment of coarctation of the aorta was low,but the AUC value and sensitivity were high.The image segmentation algorithm based on the multi-scale Hessian matrix had a good segmentation effect for catheterization angiographic images,and almost no small blood vessel was missed.After clinical trial,the system has basically reached to the design aim.