Correction Control of Needle Deflection Based on Kernel Extreme Learning Machine

Percutaneous puncture surgery is one of the important methods for early diagnosis and treatment of liver tumors. However, due to the interaction between the needle and soft tissue, the tip of the needle receives unbalanced force, resulting in the needle body bending and deviating from the preoperative planned puncture path. We have developed an ultrasound guided puncture robot system and investigated a data -driven needle deflection correction control method. By using a kernel extreme learning machine to calculate the parameters of the needle deviation correction model in real-time, we can achieve accurate puncture deviation correction. Experimental results show that, compared to the no correction control strategy, our proposed needle defleciton correction control strategy has improved the punture accuracy in three aspects: with needle insertion depth 60mm, path following maximum error decreases from 7.193mm to 0.611mm, path following average error decreases from 3.521mm to 0.185mm, and target puncture error decreases from 7.193mm to 0.611mm, greatly improving the accuracy and safety of percutaneous puncture surgery.