Penetration Event Identification Based on Neural Network for Needle Tip Location in Robot Assisted Lumbar Puncture Surgery

Abstract

Accurate determination of the needle tip location is critical not only to the success of the lumbar puncture procedure, but also to the reduction of complication rates. The interaction force at the needle tip contains important information for the needle tip location estimation. However, precise identification of puncture forces to key tissue layer is a great challenge to achieve only by traditional blind manipulation. This letter proposes an approach based on a multilayer backpropagation neural network to determine the needle tip location during needle penetration into multi-layered lumbar tissues with a robot-assisted needle insertion system. A new neural network model is used to predict the penetration event in real time only based on the monitoring of the needle–tissue interaction forces. A protection lock algorithm is proposed to prevent the forward movement of the needle expeditiously when the prediction algorithm of the neural network is triggered. The experimental results demonstrate higher accuracy and real-time performance compared to those of the conventional methods.