Real-time Lung Markings Tracking for Percutaneous Puncture Robots

Accurate lung markings tracking in real time is a cornerstone to percutaneous puncture robots. We present a method of lung markings tracking used in fluoroscopic video in this paper, which employs correlation filters combined with shallow and deep features. After that, we propose some improvements aiming to make it real-time. We evaluate our method on fluoroscopic video of dog and orthogonal digitally reconstructed radiographs generated by four-dimensional computed tomography and wish to implement it to aid percutaneous puncture robots soon. Results demonstrate great real-time performance and high accuracy when tracking lung markings and tumors. As far as we know, this is the first time such method for real-time lung markings tracking has been proposed which is applicable for various types of target.