An Image Information-based Classification Method for Vascular Interventional Surgery Operating Skills

The success rate of the vascular interventional surgery (VIS) depends largely on the skill level of the surgeon. Surgeons with different skill levels will have differences in generating movement trajectory inside blood vessels. The operation skills and skill levels of surgeons during VIS can be evaluated through the images that include the movement trajectory of the distal part of the catheter. Thus, it is very meaningful to propose a method to correctly distinguish the operations of experienced surgeons from the operations of inexperienced surgeons. This paper presents a method to differentiate surgical skills of surgeons in vascular interventional surgery. In our study, the movement trajectory of the guidewire in the images based on the two-dimensional vascular models was firstly collected. Then, these images were manually annotated and the Elan software was used to annotate the operation time. In addition, whether the guidewire deformed when it collided with the vascular wall during the operation was obtained indicate the significant differences between the two groups. Corner detection algorithm was used to obtain the motion coordinates of the distal part of the guidewire in each operation. The coordinates of the distal part were drawn on a picture, that is, the distal end trajectory in an operation is generated. The above method was used to obtain all the movement trajectories of experienced and inexperienced operations. Finally, the VGG network was used to classify them and the results were obtained. Finally, the classification accuracy of the proposed method can reach 97.4% from the experimental results, which proved that the proposed method was effective and feasible.