When the epidural needle is punctured into human body during epidural anesthesia surgery, the location of the needle tip is of great importance. In our study, we developed an OCT endoscopic system to help locate the needle tip in real time. Backbones from pigs were utilized to test our system. According to the tissue types that epidural needle punctures through, we imaged five different tissues (fat, ligament, flavum, epidural space and spinal cord). Furthermore, deep-learning methods were used to automatically distinguish the tissue types and predict the distance between the needle tip and the spinal cord. We achieved an average prediction accuracy of 96.65% in tissue classification, and an absolute percentage error at 3.05%±0.55% in distance measurement.
|