1. Navigation-assisted System For Suture Anchor Insertion
A navigation-assisted system for rotator cuff repair was developed to guide the anchor insertion at an angle close to the preset target angle (Fig. 1-A). This system provided both arthroscopic and navigation views. Unlike typical arthroscopic view, the position of a surgical tool in the navigation-assisted arthroscopic view is marked by an arrow, which turns red and shows a virtual instrument when a surgical tool is not seen on the arthroscopic image, or green when the surgical tool is visible on the image. This augmented reality-based technique allows the operator to keep track of the direction of surgical tools even when they are invisible on the arthroscopy screen, thereby improving the hand-eye coordination of the operator during surgery. To implement this notification function, first optical tracking markers (Passive Sphere Markers, Northern Digital Inc., Waterloo, Canada) were attached to the arthroscope and surgical tools to obtain their location and direction information followed by calibrations of the arthroscope camera, hand-eye, and pivot .
The navigation view developed displays a 3D shoulder model and surgical tools from three perspectives, that improve 3D spatial recognition of the relative positions of the patient shoulder and the tool, while supplementing the limited depth recognition of 2D arthroscopic image. The anchor-insertion angle, defined by vertical and horizontal angles, is calculated as the angle between the humerus and the surgical tool, with each vertical and horizontal baseline as a reference, respectively (Fig. 1-B) and displayed real time for operators. The vertical baseline is a straight line that connects the greater tuberosity and the surgical neck, while the horizontal baseline is perpendicular to the vertical baseline, bisecting the lesser tuberosity on a plane that faces the glenoid in front. Defining the ideal target insertion angle against each baseline before anchor insertion, enables the operators to insert anchors at an insertion angle close to the target using the proposed navigation.
To implement the navigation view, location-tracking markers were attached to a shoulder phantom or cadaver, and cadaver-to-image registration  was conducted to reproduce the relative positions of the shoulder and the surgical tools in a 3D virtual space. Accuracy of anchor insertion using the navigation system is determined by patient-to-image registration. Fiducial registration error and target registration error were 1.68 mm and 1.34 mm in phantom model, and 3.76 mm and 3.91 mm in cadaveric model, respectively.
2. Experimental Setup
2.1 Surgical performance analysis in phantom model
The surgical performance using the conventional arthroscopic system and the proposed navigation-assisted system was evaluated by motion analysis (Prime 41; Natural Point, Inc., Corvallis, OR, USA) (Fig. 2-A). The conventional arthroscopic system (IM4000, IM4120; ConMed Linvatec, Utica, NY, USA) was a commercial 4-mm diameter 30arthroscope with a viewing angle of 105, whereas the proposed arthroscopic system with navigation technology was a 7-mm diameter 0 arthroscope with a wide viewing angle of 150 (MGB Endoscopy, Seoul, Korea). Both arthroscopic systems have an image resolution of 19201080 pixels, and the acquired images were displayed for the participating operators on a full-HD monitor on the experiment table (Fig. 2-B).
Phantom models of the shoulder joint (Arthrex, Naples, FL, USA) were used in this study. All phantom models had same sized rotator cuff tears. All conditions and environment surrounding the participants were applied consistently. Participants included three experts who were shoulder and elbow fellowship trained orthopedic surgeons and two novices who were orthopedic residents with no experience of shoulder arthroscopy. All five operators were right-handed and handled the proposed system first and then the conventional arthroscopic system. Prior to the experiment, all participants were given instructions regarding the arthroscopic tasks and preset portals. Operators conducted a given task once with each of the two arthroscopic systems.
To ensure that the experimental environment for all participants was identical, the positions of the arthroscope, surgical tool, shoulder phantom, and both their hands were preset using an arrangement tool (Fig. 2 -C). All participants began their tests by placing their hands on the palm contour of the arrangement tool and finished the tests by laying the arthroscope and surgical tool, as determined in advance. Two anchor insertion spots were marked, and an anchor was pre-inserted at one of the spots in the phantom model (Fig. 2 -D).
Two reflective markers were attached on the dorsal side of the third metacarpal of each operator’s hands. Four motion analysis camera were set and calibrated prior to the experiment. The motion analysis system was capable of storing 3D (x, y, z) location data with a resolution of up to 0.01 cm. Data obtained from the system were analyzed using Matlab (R2012b; MathWorks, Torrance, CA, USA). Surgical performance metric used to analyze the surgical skill of the operators included the total path length (mm) of the arthroscope and surgical tool, the count of movements, and the time taken (sec). The total path length refers to the sum of the distances of all 3D movements of the operator’s hands during surgery. The count of movements was defined as the number of occasions during which the instantaneous velocity exceeded the average velocity . The time taken was counted from the moment when the operator inserted an arthroscope into the portal to the moment when the operator placed it on the experiment table after the completion of the given experimental task.
The experimental tasks comprised anchor insertion and anchor suture retrieval. In the former, the operator observes arthroscopy images to find the ideal anchor insertion angle and places surgical tools accordingly. For phantom models, the ideal anchor insertion angle was preset to a single value, i.e., 135 in a vertical direction. This task was conducted twice, each at the two anchor insertion spots. The anchor suture retrieval task required the operator to pull the four sutures of the inserted anchor out of the portal one by one using an arthroscopic retriever.
In the cadaveric experiment, the conventional and the proposed navigation assisted arthroscopic systems were compared for variation in the anchor insertion angles (Fig. 3-A), (B)]. The anchor insertion was conducted three times using each system. The conventional and proposed navigation assisted arthroscopic systems used in the cadaveric experiment were the same as those with the phantom model. Three optical tracking markers were used — a patient reference marker was fixed to the humerus using Steinman pins, and the other two markers were attached to the arthroscope and surgical tools, respectively. To implement the position and orientation between the surgical tool and the humerus in a virtual space, a patient-to-image registration was performed using anatomic landmarks as in the phantom setting test.
For the cadaveric experiment, three participants from experts group first conducted anchor insertion using the navigation system and then the conventional system. Anchor insertion was conducted three times each at supraspinatus (SSP) and infraspinatus (ISP) footprints on the greater tuberosity (Fig. 3-C). The ideal anchor insertion angle was set to within a range between 45° and 90° from the footprint cortex by an expert surgeon with shoulder arthroscopy experience [2, 1]. Using our vertical and horizontal references, the angles in the vertical and horizontal directions for the first anchor were 140 and 90respectively, and 144 and 102 respectively, for the second anchor. The participants performed the anchor insertion task with the goal of achieving the predetermined target angles.
As in the motion analysis experiments, the operators attempted to find the ideal anchor insertion angle and place surgical tools accordingly. Once the operator determined an anchor insertion angle, the angle data were recorded.
3. Statistical analysis
Statistical analysis of the surgical performance and anchoring angle was performed using OriginPro software (ver. 9b; OriginLab Corp., Northampton, MA, USA). Data homogeneity of the three surgical performance metrics (total path length, count of movements, and time taken) and anchoring angle errors was evaluated by the Shapiro-Wilk normality test. Since all data were not normally distributed, the paired-sample Wilcoxon signed rank test was used to determine the statistical significance of each measurement (total path length, count of movements, and time taken) using the proposed navigation system, compared to the conventional arthroscopic system. Additionally, Mann-Whitney U test was used to confirm significant differences between the expert and novice groups. The level of significance was set at P < 0.05, * for P < 0.05 and ** for P < 0.01. Lastly, a post-experiment power analysis was performed at a 0.05-significance level to determine the power to detect a significant difference between the two systems.