One of the telemedicine technologies that is gaining considerable interest is telementoring, during which an expert medical professional (mentor) can provide real-time instructions to a novice (mentee) even if they are remotely located [1]. A growing application of telementoring technologies is seen in the field of surgery as a result of rising demands for specialized surgical expertise coupled with limited training opportunities [2]. Minimally invasive surgery (MIS) is one such specialized surgical field, which has emerged as the standard practice [3]. During MIS, which includes laparoscopic and robot-assisted procedures, surgeons use long, thin surgical instruments inserted through small incisions on the body. The operative view is captured through a camera scope and displayed on a screen. MIS is increasingly adopted due to the benefits it provides to patients, in terms of significant reduction in pain, surgical site infection, scarring, and recovery time as compared with the highly invasive open approach [4]. Despite these pronounced advantages, its widespread implementation is hindered due to the challenges in adequate skills acquisition. Surgeons performing MIS have an indirect view of the operative view through the screen projecting the live camera scope feed and experience a longer learning curve as compared with open surgery [5]. Hence, specialized training is needed to develop sufficient expertise. While basic MIS skills are learned outside the operating room, intraoperative training is vital for attaining proficiency [5]. Unlike open surgeries where the expert surgeon can easily point at critical structures with their hands over the operative view, mentors must rely on verbal instructions during MIS, which are prone to communication errors [6]. In some cases, the mentor must take complete control of the instruments from the mentee to show the required surgical technique. While the dual surgeon console in robot-assisted surgeries facilitates such takeovers by the mentor, the cost can be prohibitively high. Furthermore, many countries experience challenges in implementing MIS due to inadequate training programs and high costs associated with bringing in expert surgeons [7]. Telementoring technologies that allow an expert surgeon to efficiently guide the operating surgeon play a major role in addressing this gap [8].

Implementing telementoring technologies during MIS is relatively easier, as compared with open surgeries, since a live view of the operating field is already captured through the camera scopes [9,10]. During conventional telementoring, video captured from the operating room is transmitted in real time to an expert mentor at the remote location. Verbal instructions (audio cues) from the mentor are transferred to the mentee [11]. The mentor can also play surgical videos (prerecorded visual cues) to guide the mentee on a particular step [12]. However, such cues may not be sufficient for intraoperative instructions [6]. To overcome the communication challenges posed by verbal instructions, expert surgeons need a way to point at anatomical structures on the screen displaying the operative view to the mentee [13]. For this purpose, the mentor may use augmented reality (AR), which involves the use of computer-generated virtual objects that are superimposed on the mentee’s view of the operative field [14]. The AR cues may be rendered to the mentee on a visualization screen, or the surgeon’s console with stereoscopic vision during robot-assisted surgery. In some cases, external AR devices such as Microsoft HoloLens head-mounted display (HMD) or iPad tablets may also be used [15]. The cues could be in the form of 2-dimensional (2D) annotations (static AR cues) demonstrating an incision line or highlighting a specific anatomy [16,17]. Such forms of telementoring have been applied extensively for MIS demonstrating improved communication and performance [18]. Nonetheless, static AR cues are limited while compared with in-person guidance as complex tool-tissue interactions may not be accurately depicted [19,20].

Telementoring can be enhanced with the use of dynamic AR cues to effectively communicate surgical steps. As opposed to static AR cues, which remain stationary on the mentee’s field of view, dynamic AR cues move in real time under the mentor’s control [21]. These virtual cues can be in the form of hand gestures, pointers, or surgical tools. They are often 3-dimensional (3D) in nature, providing depth perception as the mentor guides the mentee through surgical sub steps. In cases where the mentor and mentee have similar macro skills, such as proficiency in critical anatomy and movement of surgical tools, these dynamic cues can aid the mentee’s development of micro skills related to a particular unfamiliar surgical technique [22]. Dynamic AR cues are not only useful for telementoring when the mentor is situated in a remote location, but can also be used to enhance in-person guidance during MIS where the mentor and mentee are in the same operating room [23].

Several reviews related to telementoring in health care have been previously published; however, they did not focus on applications for surgery in specific [15,24]. Other reviews covering surgical applications of telemedicine and telementoring have been presented [1,11,25-30]. These articles largely covered implementations of telementoring systems where video feed from the mentee is sent for audio feedback from the remote mentor. Some of them did not focus on MIS and explored applications in open surgery as well [11,29,30]. Additional review articles focused on specific use of telementoring for certain specialties such as head and neck [31], robotic [32], and urology [33]. With respect to MIS, Nickel et al [23] provided a concise clinical summary, whereas others reported on the use of audio and static AR cues [16,17,19,34,35]. While these reviews offer a comprehensive understanding of telementoring during MIS, they do not highlight the developments in the use of dynamic AR cues. To the best of our knowledge, there has not yet been a systematic examination specifically on the use of dynamic AR cues used by remote mentors during telementoring in MIS. Such a synthesis providing an overview of the different types of dynamic AR cues as well as the surgical procedures, clinical settings, and factors used for testing is needed. This will help guide future research and developments on telementoring in MIS and ensure that they are relevant and practical.

Through this scoping review we aim to provide a comprehensive summary of dynamic AR cues used for telementoring during MIS. A survey of the various types of surgeries, participants, dynamic AR cues, and evaluation of the telementoring technology is presented to guide future developments on telementoring during MIS. We aim to address the following research questions (RQs), which are yet to be answered by review articles published in the field:

One of the aims of the scoping review was to present a comprehensive summary of the use of dynamic AR cues for telementoring during MIS. Through systematic searches of the scientific literature, it was observed that dynamic AR cues for telementoring were more frequently implemented for laparoscopic surgeries (as compared with robot-assisted surgeries). It was mainly used for teaching suturing skills on synthetic phantoms and remote mentoring during simulated cholecystectomy procedures. The mentors primarily consisted of expert surgeons, while medical students and surgical trainees dominated the mentee group. The use of hand gestures and surgical tools as dynamic AR cues were more frequent as compared with virtual pointers. Hand gestures and surgical tools allowed the mentor to convey information such as position, orientation, and tool tip motions to the mentee. To display the dynamic AR cues to the mentee, a small proportion of the studies used iPad tablets (n=1, 5%) Microsoft HoloLens (n=3, 14%), and the surgeon’s console with stereoscopic vision for robot-assisted surgery (n=4, 19%). However, most of the articles used a standard visualization screen (n=13, 62%) through which the mentee was able to see the AR cues. Among the included articles, at least 14 (67%) of them had a study design where the use of dynamic AR cue was compared with different modes of instruction with the most common comparators being audio cues and in-person guidance. The key factor used for evaluation was skills. The most frequently reported outcome measures were time taken for task completion and number of errors.

In general, the results of the included papers suggest that participants subjected to telementoring with dynamic AR cues experienced considerable beneficial effects when compared with groups receiving no cue, prerecorded visual cue, in-person guidance, audio cue, or static AR cue. More specifically, dynamic AR cues resulted in significant improvement of FLS score when compared with no cue [44] and lowered error count (P<.05) when compared with prerecorded visual cues [43]. While comparing against in-person guidance, some studies reported longer operating time [42], increased anesthesia use, and cauterization [46] with the use of dynamic AR cues. At the same time, few others noted a reduced complication percentage [20], lower operating time (P=.01), steeper learning curve, and fewer failed attempts [54]. Furthermore, several other studies found no significant difference in performance [42,44], number of errors [54], operating time [43,46], and clinical outcomes [46]. This indicates that telementoring using dynamic AR cues was similar to in-person guidance. While comparing with audio cues, some studies reported no significant difference in CVS, GOALS score [40], number of errors [50], and operating time [6,40,50]. However, the majority of studies demonstrated significant advantages of dynamic AR cues over audio cues. These were in terms of subjective performance [40,47,50], user ratings [40], NASA-TLX score (P<.022) [13], economy of movement [47,50], error rates [13,18,47], gaze behavior (P<.01) [18], GOALS score [6,13], operating time (P<.001) [13], OSATS score (P<.01) [6,13,18], turn frequency (P<.0001) [49], and complications [6,13]. These results show that dynamic AR cues were comparable, and significantly superior to telementoring with audio cues alone. In addition, a study by Jarc et al [53] noted that users preferred dynamic AR cues over static 2D telestrations. Overall, dynamic AR cues resulted in superior outcomes, and in some cases, they were considered on par with conventional methods such as in-person guidance and audio cues.

The extent to which the above-mentioned findings may be used to determine the benefits of the technology is limited. This is because majority of the analytical studies with a comparator were conducted on synthetic phantoms, mostly assessing suturing [42-44,54]. Only 2 of the analytical studies were conducted in real patient settings comparing dynamic AR cue with audio during cholecystectomy [40] and in-person guidance during aquablation [46]. In addition, the outcome measures used to assess the different dynamic AR cues were not consistently used across the studies. The effectiveness of hand gestures was primarily assessed through OSATS, pointers through subjective performance, and surgical tools through latency and operating time. As a result, it is difficult to objectively compare between the different dynamic AR cues and determine the optimal one. Nevertheless, the articles published in the past few years reveal an upward trend in the use of hand gestures and surgical tools (Figure 5). This strongly indicates a rising interest in the use of dynamic AR cues for telementoring in MIS as well as the relevance of using hand gestures and surgical tools.

The implementation of telementoring using dynamic AR cues requires consideration of several technical factors. The Society of American Gastrointestinal and Endoscopic Surgeons (SAGES) issued a set of guidelines that telementoring systems are recommended to comply with. SAGES recommends that a minimum bandwidth of 40 megabits per second must be maintained. In addition, the latency (delay in sending mentee’s operative view to the remote mentor) must fall below 450 milliseconds [57]. One of the methods used to reduce latency is by encoding the video frames at the mentee’s location and decoding it at the mentor’s side. However, during this process, there is a risk of lowered video quality experienced by the mentor. To measure the extent of visual degradation experienced by the mentor, metrics such as structural similarity index measure (SSIM), peak signal-to-noise ratio (PSNR), and mean square error (MSE) are often used [58,59]. SSIM is considered more effective since it is more closely related to the visual quality perceived by the human eye, as compared with PSNR and MSE [22]. The delay experienced by the mentee in receiving mentor’s instructions can be reduced when point cloud data (depicting hand gestures or surgical tool) is used instead of video data. Harnessing 5G and future 6G capabilities that provide extremely low latencies can ensure little to no disruption [23]. While several studies included in the review reported information latency between mentor and mentee, they were mostly conducted on synthetic phantoms [22,41,48]. Further research presenting technical results during actual live MIS is imperative.

Ethical, safety, and legal concerns must be addressed while implementing telementoring systems. Informed patient consent must be ensured [23]. Every effort should be made to minimize cybersecurity risks and ensure compliance with guidelines on telemedicine issued by HIPAA (Health Insurance Portability and Accountability Act) [31]. The computer systems at both the mentor and mentee locations must be secure with monitoring schemes in place to detect and prevent potential breaches of sensitive patient data. To protect the data while being transferred between the mentor and mentee, appropriate data encryption and virtual private networks must be used [57]. In addition, a clearly outlined process for licensure and credentialing of surgeons is also required for remotely located mentors [8]. Generally, the mentee (who will be a competent operating surgeon) holds primary medical responsibility and is legally liable. Nonetheless, the SAGES states that during telementoring, the remote mentor retains shared responsibility for the patient [60]. Consequently, the mentor must hold the necessary medical license and surgical privileges in the region where the surgery is taking place. This becomes a barrier to adopting telementoring nationally in places like the United States and Canada as licensing is usually at the state or provincial level. However, this would not be a problem in the European Union (EU) where surgeons licensed under one EU member state can practice in other EU states as well [22]. Furthermore, licensing requirements would not be an issue when dynamic AR cues are used to enhance in-person guidance.

Telementoring using dynamic AR cues addresses the need for efficient ways to remotely guide an operating surgeon. This can be potentially replaced with telesurgery where the remote surgeon performs the surgery through robot-assisted surgical systems. While successful telesurgery cases were reported since the early 2000s, significant progress was made only in the recent years owing to the advances in surgical robot systems as well as telecommunication technologies [61]. However, several factors prevent widespread adoption of telesurgery. Although 5G networks are available, 3G or 4G is the standard in most countries leading to major concerns about latency. In addition, more research is needed to assess the cost-effectiveness of telesurgical services. A long-term sustainable approach must be explored for implementation and maintenance. Nonetheless, several cases of telesurgery for surgical specialties such as general, urology, and neurosurgery have been reported [62]. This provides hope for using remote technologies to reduce the global burden of death and disease that can be prevented or treated through surgical care.

This review is limited by the inclusion of articles published in English only, making it less comprehensive. A risk of bias assessment of the included studies was not carried out. A comparative analysis between the different dynamic AR cues was not conducted due to the variety in outcome measures reported. Instead, overall summary potentially guiding future research on dynamic AR cues used for MIS telementoring is presented.

Although further research is needed to establish its benefits over conventional mentoring, the use of dynamic AR cues in MIS telementoring proves to be promising. Future developments may focus on evaluating the technology through multi-institutional randomized controlled trials. In addition to measures of perceptions and skills, objective clinical outcomes that reliably assess the effectiveness of telementoring should also be evaluated.