Real-time Spinal Needle Position Detection Using YOLOv8

• Published in: Proceedings of ... International Conference on Instrumentation Control and Automation pp. 131 - 136
• Main Authors: Haq, Agung Izzul, Susanti, Hesty, Barri, Muhammad Hablul, Zafira Putri, Aurelia, Fillah, Fajrul Falah
• Format: Conference Proceeding
• Language: English
• Published: IEEE 09.08.2023
• Subjects: Acoustic properties; Cameras; Deep learning; Needle positioning; Needles; Object detection; Real-time detection; Real-time systems; Spinal needles; Spinal procedures; Testing device; Training; Ultrasonic imaging; Ultrasound guidance; Visualization; YOLOv8 model
• ISSN: 2639-5045
• DOI: 10.1109/ICA58538.2023.10273077

The use of ultrasound guidance for spinal procedures is common practice, but the visibility of spinal needles can be inconsistent. To address this issue, a needle testing device should be developed to determine the acoustic properties of spinal needles. The device should be capable of accurately and precisely positioning the needle according to the instructed coordinates. In addition, it is important to have visual feedback for the device to ensure proper needle placement. To provide this, a camera will be placed in front of the device to provide real-time visualization of the needle as it is moved. The camera will read the location or coordinates of the spinal needle, providing feedback for the device. This visual feedback allows the user to monitor the needle's position and make any necessary adjustments for proper placement. To accomplish this, a needle detection system needs to be created as a subsystem for the device using deep learning and object detection techniques with the YOLOv8 model. The system achieves Precision-Recall of 79% in training with 95% accuracy in testing on detecting spinal needles in real-time, with the potential to improve the accuracy of the spinal testing device by providing reliable needle location.