Puncture Status Perception Method for a Venipuncture Robot Based on Force Prediction

Research on venipuncture robots has excellent potential for reducing workload and addressing healthcare professional shortages while advancing medical automation and intelligent healthcare. This article proposes a method for improving the perception ability of venipuncture robots by predicting the puncture force to efficiently determine whether the puncture needle has entered the vein, thus increasing the safety and success rate of the robot. Initially, the characteristics of the puncture force are analyzed, and a force-time-series-prediction model is established via a bidirectional long short-term memory (BiLSTM) network and Kalman filter error compensation. A time-series dataset is constructed using puncture force data collected under different conditions. The experimental results indicate that the model’s prediction error is less than 0.01 N, and the calculation time is less than 0.02 s. Based on this prediction model, a puncture status perception method is proposed, which considers the trend and regularity of puncture force changes during venipuncture to determine whether the needle has entered the vein. Finally, experiments are conducted via this method in a venipuncture phantom and compared with those of other methods. The findings indicate a substantial improvement in puncture status perception efficacy of 82% and a corresponding 11% increase in the success rate of robot-assisted venipuncture. These results demonstrate the effectiveness of this method in enhancing the performance and safety of venipuncture robots and highlight its potential for practical application.