Deep learning aided epidural needle guidance using a forward-view polarization-sensitive optical coherence tomography probe

Accurate guidance of the epidural needle is important for ensuring the safety and efficacy of epidural anesthesia. Within this study, we proposed an endoscopic system built on polarization-sensitive optical coherence tomography (PS-OCT). To evaluate its viability, we performed experiments on ex-vivo human epidural specimens. Throughout the experimental process, we captured and analyzed various layers of spinal tissue that the epidural needle goes through during the surgery, including subcutaneous fat, supraspinous ligament, interspinous ligament, ligamentum flavum, epidural space, dura, and the spinal cord. Each of these tissue layers had distinctive OCT imaging patterns. Furthermore, we employed deep learning techniques for automated tissue recognition.