Resistance Recognition of Moving Guidewire in Vascular Interventional Operation

Vascular interventional surgery is prevalent in treating cardiovascular and cerebrovascular diseases because of its low-invasive and rapid recovery. Surgeons cannot observe the vascular contour in real-time during the operation. So it is a technical task. In this paper, we try to research the problem by using the object detection and recognition model. We divide the guidewire moving in the 2D image into two states: resistance deformation and nonresistance deformation. Then we use the Faster-RCNN to classify and track the guidewires in phantom and synthetic surgical videos. The operation datasets in the phantom come from the PCI trainer. After extracting the mask of the guidewire as the ground truth, we randomly add the factual X-ray image information of the patient in the background. Finally, we compare the recognition results of the two cases. The model converges on the validation datasets, which proves that our method is effective. In vascular interventional surgery, the AI model can learn the expert experience, which may render an idea for the development of surgical robots.