A CNN-based Method for Guidewire Tip Collisions Detection in Vascular Interventional Surgery

摘要

In vascular interventional surgery, the tips of the guidewire or catheter are easy to collide the vascular wall and cause rupture or harm. These risky actions are difficult to track and can only be memorized by the surgeons. In this paper, we designed a convolutional neural network (CNN) model to identify whether the tips of guidewire collide the vessel wall. Finally, through the training and testing, our model got 95.9% accuracy in simulated vascular model. In addition, we also found some samples which the guidewire closed to the vascular wall were misdiagnosed by our models. If the images preprocessing accuracy can be improved, the results of the model will be increased in the future.