Research On Robot-assisted Spinal Ultrasound Automatic Scanning System

Scoliosis is a three-dimensional(3-D)spinal deformity disease with great harm.Doctors usually evaluate scoliosis with the help of medical imaging technology in diagnosis,regular examination,and preoperative planning.At present,X-ray is the most common way to observe and detect the severity of scoliosis.However,exposure to radiation of X-ray can increase the risk of cancer.Compared with X-ray,ultrasound imaging has the advantages of non-radiation,low cost,high popularity and simple operation.3-D ultrasound imaging broadens the vision of traditional 2-D ultrasound imaging,which has the ability of providing doctors with the 3-D spatial structure of the scanning tissue,thus better assist doctors in diagnosis and treatment.It is reasonable to believe that 3-D ultrasound imaging is an effective and feasible way to image the spine.At present,3-D ultrasound imaging is mainly obtained by freehand scanning.However,human intervention lead to not only poor repeatability of scanning process and results,but also low level of automation and intelligence.The heavy scanning workload could also bring doctors musculoskeletal disorders.To develop the application of non-radiation ultrasound imaging technology in spinal clinical scanning,a robotic-assisted ultrasound scanning and 3-D imaging system for human spine is proposed and established in this paper.The main contributions of this study are as follow:1.A robotic arm system is proposed to replace the doctor to scan the spine.The shortcomings of freehand scanning are solved well by robotic-assisted scanning.The automatic robot assisted scanning mode can easily obtain the scanning process and results with excellent repeatability,and greatly reduce the workload of doctors.2.An image segmentation method based on deep learning is proposed to realize the automatic recognition of spine region.In order to establish an intelligent system with the ability to recognize the spine area and pre-plan the scan path automatically,a multimodal feature fusion fully convolution neural network combined with spatial information is built to automatically recognize and segment the spine region of human back.The multi-modal feature fusion method fuses the spatial features and RGB features of the spine area,integrating the two features to guide the network to complete higher precision segmentation and improve the recognition of the changeable spine area.Then a pre-planning path is obtained based on the segmentation result.3.A robotic-assisted ultrasound scanning and 3-D imaging system for human spine is established.The system can automatically recognize and scan the spine region,and realize the 3-D reconstruction of spine and the measurement of Cobb angle after scanning.Two control strategies of robotic arm are adopted to determine and adjust the pose of the robotic arm.More specifically,during the scanning,the ultrasonic probe and human skin should keep closely and constantly coupled to ensure the effective transmission of ultrasonic signal,hence obtain high-quality ultrasound data for the subsequent 3-D reconstruction and spinal morphology evaluation.The robotic control strategies based on the normal vector and real-time force feedback are adopted to keep the probe and spinal region closely contact in the whole scanning.As mentioned above,the quantitative evaluation of scoliosis is currently carried out on 2-D X-ray films.Since the spine is a 3-D structure,the 2-D imaging results ignore the morphological and structural characteristics of the spine in 3-D space,which may introduce deviation in evaluation results.After the scan,the 3-D reconstruction and visualization is completed,and an iterative optimization algorithm based on 3-D reconstruction body is adopted to calculate Cobb angle to realize the quantitative evaluation of scoliosis.4.The system is verified by phantom experiment and in vivo experiment.To verify the performance of the system,phantom experiments and human experiments are designed and conducted.The experimental results demonstrate that the system can complete the automatic scanning of the spine area by the robot equipped with the ultrasonic probe,and ensure that the ultrasonic probe is closely coupled with the skin with the correct posture and strength during the whole scanning.Finally,the system achieve high-precision 3-D reconstruction of the spine and small measurement error of scoliosis.