We designed and constructed a surgical assistance system that can receive an arthroscopic screen output and reflect it in real time. Arthroscopy is a minimally invasive surgical procedure performed on a joint by making multiple small incisions. The surgeon inserts medical instruments and an arthroscope into the portals through incisions to perform a surgery. Since the surgeon controls arthroscope and operates the surgical tool simultaneously, the surgery is difficult and the surgeon is likely to become fatigued. Hence, a support system such as the aforementioned surgical assistance system is expected to increase the quality of surgery.
It is necessary to know the position of the surgical tool to obtain optimal visual feedback. For this,
methods such as the attachment of a marker to the surgical tool are suggested by preceding studies, but they are difficult to employ owing to the resulting cost and complexity of the procedure. Image processing that can be performed without physical change is suitable.
This study employed machine-learning-based object detection to detect surgical instruments and facilitate their tracking by the arthroscope.