Camera-Based Surgical Navigation System: Evaluation of Classification and Object Detection CNN Models for X-markers Detection

In hip and spine surgery, accurate tracking of the medical instruments allows the surgeon to navigate the instruments accurately and safely within the operating field. X-markers based optical tracking are now applied in various fields owing to their simple shape and robustness against contamination and coverage. Nevertheless, they are not unique and easily distinguishable compared to other visual markers, e.g. ArUco, April Tags, QR codes, etc. Incorrect localization and classification of the X-markers hinder the pose estimation process. Recent studies have shown the great potential of deep learning approaches in solving visual marker detection problems often faced in conventional approaches. In this paper, we evaluate the performance of classification and object detection models for X-markers detection. To evaluate the models, we present our unique X-markers annotated dataset. The evaluation results show the high performance of classification and object detection models for X-markers detection, thereby outperforming the conventional methods. The models, however, have a higher computational cost in contrast to conventional methods.