Abstract

One of the most common forms of heart arrhythmia is atrial fibrillation, a disorder caused by the uncoordinated electrical distribution of heart muscle. Catheter ablation is an effective treatment of choice for this heart disorder that destroys the arrhythmogenic spots in the heart muscle to recover the normal beating of the heart. To perform this minimally invasive surgery, the surgeon inserts a flexible and steerable wire-like tool, called ablation catheter into the cardiovascular system of the body. After reaching the target spot inside the heart chamber, the surgeon or robotic system pushes the catheter's tip against the heart tissue. Afterward, the heat generated by the embedded electrode at the tip of the catheter burns the target spot. A properly controlled catheter-tissue contact force (CF) results not only in a safer procedure, but also considerably increases the success of treatment. To this end, the main motivation of the present thesis is to propose an accurate learning-based CF estimator and then use it in a control loop. This research is a proof of concept for estimating and controlling the CF. An experimental setup was designed in an attempt to mimic the ablation surgery. Firstly, a vision-based method using machine learning model is proposed in order to estimate the CF without using a sensor at the tip of the catheter. This model requires the image of bending section of catheter to output the CF. The real image should come from the X-ray machine in the operating room but to simulate the X-ray imaging system, in this research, a camera was implemented. This real-time and sensor-free technique provides acceptable accuracy when estimating the CF of standard catheters. A catheter manipulator mechanism is implemented to control the CF at the desired value. The feedback element of the control loop is the mentioned vision-based force estimator. In addition, an adaptive imaging system including a robotic arm and camera is implemented to track the catheter for force control.