This study proposes the development of a computer vision-based system for the automated detection of surgical sutures, employing YOLOv8 architecture. For the training and validation process, a collection of free-access images in real-world scenarios was collected, preprocessed, and labeled. The tests covered in this study demonstrated a competent capability of detection and classification, supported by metric values such as mAP = 92.8%, accuracy = 95% and recall = 95% (at the standard of IoU=50%), with metric values rounding the 90% at higher values of IoU (70%). Performance of this system depicts a robust solution in the detection of sutures in non-ideal images, laying the groundwork for future developments in automated feedback within training environments.