Development of a Needle Deflection Detection System for a CT Guided Robot

The uncertainty and unpredictability regarding the occurrence of needle deflection during percutaneous puncture, especially when using very fine needles, can greatly complexify surgical tasks such as needle insertion in the lower abdomen. To avoid the increased risks induced by prolonged CT scan radiation exposure, this paper offers an alternative to the retrieval of needle tip position from CT scan images. In this method, the deflection of the needle is detected and reported in accordance with insertion force data as the needle is inserted into the bowel. This method relies on the use of a Gated Recurrent Unit based neural network to predict the occurrence and type of deflection met during the procedure depending on the intended path and tissue type to be punctured in order to reach the target (cancer tumor). This system accounts for the original angle of insertion of the needle. Results of final experiments returned a 100% true positive rate, signifying that in the eventuality of needle deflection, it would systematically have been predicted by the neural network.