Effects of a training system that tracks the operator’s gaze pattern during endoscopic submucosal dissection on hemostasis

Abstract

BACKGROUND

The early acquisition of skills required to perform hemostasis during endoscopy may be hindered by the lack of tools that allow assessments of the operator’s viewpoint. Understanding the operator’s viewpoint may facilitate the skills.

AIM

To evaluate the effects of a training system using operator gaze patterns during gastric endoscopic submucosal dissection (ESD) on hemostasis.

METHODS

An eye-tracking system was developed to record the operator’s viewpoints during gastric ESD, displaying the viewpoint as a circle. In phase 1, videos of three trainees’ viewpoints were recorded. After reviewing these, trainees were recorded again in phase 2. The videos from both phases were retrospectively reviewed, and short clips were created to evaluate the hemostasis skills. Outcome measures included the time to recognize the bleeding point, the time to complete hemostasis, and the number of coagulation attempts.

RESULTS

Eight cases treated with ESD were reviewed, and 10 video clips of hemostasis were created. The time required to recognize the bleeding point during phase 2 was significantly shorter than that during phase 1 (8.3 ± 4.1 seconds vs 23.1 ± 19.2 seconds; P = 0.049). The time required to complete hemostasis during phase 1 and that during phase 2 were not significantly different (15.4 ± 6.8 seconds vs 31.9 ± 21.7 seconds; P = 0.056). Significantly fewer coagulation attempts were performed during phase 2 (1.8 ± 0.7 vs 3.2 ± 1.0; P = 0.004).

CONCLUSION

Short-term training did not reduce hemostasis completion time but significantly improved bleeding point recognition and reduced coagulation attempts. Learning from the operator’s viewpoint can facilitate acquiring hemostasis skills during ESD.

Key Words: Eye tracking; Hemostasis; Endoscopic submucosal dissection; Gastric cancer; Training

Core Tip: The early acquisition of endoscopic hemostasis skills may be hindered by the lack of tools for assessing the operator’s viewpoint. We validated the effects of a training system that allowed an understanding of the operator’s gaze pattern during gastric endoscopic submucosal dissection on hemostasis. The time required to recognize the bleeding point was shorter after gaze pattern education using the training system. The number of coagulation attempts was fewer after the education. Learning from the operator’s viewpoint can facilitate acquiring the skills required to achieve hemostasis during endoscopic submucosal dissection.

INTRODUCTION

Endoscopic submucosal dissection (ESD) was developed in the 1990s as a minimally invasive treatment technique for early gastric cancer[1,2]. Recent improvements in dedicated electrosurgical knives and high-frequency electrosurgical units have contributed to the use of ESD worldwide[3,4]. Nevertheless, operators who perform ESD must be sufficiently skilled to prevent intraoperative and postoperative complications[5]. Controlling intraoperative bleeding is essential not only for reducing the need for blood transfusions but also for maintaining a clear view during the procedure to prevent more severe complications such as perforation[6,7]. However, a training method that equips trainees with the necessary hemostasis skills for ESD has yet to be established. The traditional ESD training method follows a hierarchical approach, including the accumulation of sufficient ESD knowledge, assistance from an ESD trainer, practice with an ESD simulator, and performing ESD under supervision[8]. While this structured approach is widely used, the learning curve varies among trainees. The difficulty of sharing trainees’ viewpoints has been the main limitation to the development of relevant training methods. Eye-tracking technology enables the real-time capture of the operator’s viewpoint during procedures. Analyzing the differences in visual gaze patterns of trainees and experts can elucidate the optimal training methods applicable to various medical fields[9,10]. Moreover, using eye-tracking devices during surgical procedures can enhance the ability of trainees to acquire specific skills[11]. Additionally, in the field of gastrointestinal endoscopy, differences in viewpoints and clinical outcomes, such as colorectal adenoma detection, are especially relevant[12-14]. Therefore, eye-tracking technology could be used to enhance gastrointestinal ESD training; however, no such reports have been published. We previously developed a recording system based on eye-tracking technology that can visualize and record the viewpoint of the operator during endoscopy[15]. By reviewing the recorded videos, trainees can gain an understanding of their viewpoints during endoscopy and improve their eye movements. Additionally, trainers can guide trainees by referring to their viewpoints so that modifications can be performed to achieve an optimal gaze pattern. This study evaluated the effects of gaze patterns on hemostasis during gastric ESD.

MATERIALS AND METHODS

Study design

This retrospective study was approved by the institutional review board of the International University of Health and Welfare Ichikawa Hospital, approval No.23-Ic-008 on February 27, 2024. The study was divided into the following two phases: Before learning (phase 1) and after learning (phase 2) (Figure 1). The requirement for written informed consent was waived because of the retrospective nature of this study. The study protocol was published on the hospital’s website. Patients who did not agree to participate in this study were excluded by using the opt-out approach. This study was conducted according to the guidelines of the Declaration of Helsinki.

Figure 1 Study flow diagram. The study was divided into two phases: Before learning (phase 1) and after learning (phase 2). Trainees performed endoscopic submucosal dissection during phase 1 and phase 2 The viewpoints during phase 1 and phase 2 were reviewed and compared. ESD: Endoscopic submucosal dissection.

Participants

All physicians, including three trainees and two experts at the International University of Health and Welfare Ichikawa Hospital, participated in this study. The trainees were defined as those who had performed fewer than 10 gastric ESD procedures. In contrast, the experts had performed more than 100 gastric ESD procedures and were board-certified members of the Japan Gastroenterological Endoscopy Society. Gastric ESD procedures performed by three trainees and two experts between October 1, 2023, and January 31, 2024, were retrospectively collected. All gastric ESD procedures were performed for early gastric cancers that were eligible for endoscopic treatment.

Procedures

Endoscopy was performed using EZ-760R and EZ-860T with the ELUXEO 7000 SYSTEM (Fujifilm Co., Tokyo, Japan). Additionally, an electrosurgical knife (ORISE ProKnife 2.0 mm; Boston Scientific Co., Boston, MA, United states), hemostatic forceps (RAICHO2; Kaneka Medix Co., Osaka, Japan), and a high-frequency electrosurgical unit were used (VIO300D; Erbe Elektromedizin GmbH, Marietta, GA, United states). The trainees performed ESD under the supervision of one of the two experts. When bleeding occurred during ESD, hemostasis was attempted with an electrosurgical knife or hemostatic forceps, depending on the degree of bleeding. The eye-tracking technology-based recording system is routinely used at our institution to help experts recognize trainees’ viewpoints during endoscopy. Therefore, this system was used to record all ESD procedures during the study period. The system comprises an eye tracker, an analyzing PC, and connecting cables. The screen-based eye tracker (Tobii Pro Spark; Tobii, Stockholm, Sweden) was attached to a 27-inch endoscopy monitor. The endoscopists stood 1.5 m from the front of the monitor so their eye movements could be captured effectively. The eye tracker sent coordinate information regarding the endoscopist’s viewpoint captured on the monitor at 33 Hz to the analyzing computer. The analyzing PC integrated the coordinate information and endoscopy data within 35 milliseconds and created the video for the PC. The video indicated the operator’s viewpoint using a green circle with a diameter of 30 pixels. Endoscopists calibrated their eye gaze position before starting ESD procedures to ensure accurate tracking of eye movements. The eye tracker used demonstrated extremely high accuracy, with a 0.26 route mean square error in capturing eye position.

Training using the recorded videos

During phase 1 (before learning), the trainees performed gastric ESD under the supervision of the experts (Figure 1). The system tracked the viewpoint of the operator during all ESD procedures. After phase 1 ended, the trainees watched the full-length ESD videos with their tracked viewpoints and those of the experts. By watching the experts’ videos, the trainees learned the appropriate visual gaze pattern during hemostasis. Additionally, by reviewing the trainees’ videos, they learned whether the visual gaze pattern during hemostasis was correct and what required correction. During phase 2 (after learning), the trainees performed gastric ESD while their viewpoints were tracked. However, the trainees did not review the videos recorded after each ESD during phase 2.

Processing and analysis of videos

One clinical investigator (Fumiaki Ishibashi) reviewed the full-length videos of ESD recorded by the system during both phases. Clips of the bleeding scenes, which included the time from the beginning of bleeding until the end of bleeding, were created. If a single ESD video contained multiple bleeding scenes, then each bleeding scene was prepared as an independent video clip. Two clinical investigators (Takao Tonishi and Fumiaki Ishibashi) independently reviewed the processed videos to assess the outcomes. The values measured by these two reviewers were averaged and analyzed. Additionally, the videos of ESD performed by the experts during the study period were analyzed in the same manner and compared with those of the trainees. The data of the experts were considered proficiency goals for the trainees.

Study outcomes

The outcomes were the time required to recognize the bleeding point, the time required to complete hemostasis, and the number of coagulation attempts. The outcomes during phase 1 were compared with those during phase 2. Additionally, the outcomes of the experts were compared with those of the trainees. The time required to recognize the bleeding point was defined as the time from the start of bleeding until the operator’s viewpoint recognized the correct bleeding point. If the gaze position, which was indicated by a green circle, coincided with the bleeding point for more than 1 s, then it was determined that the operator correctly identified that point. The time required to complete hemostasis was defined as the time from the start of bleeding until the completion of hemostasis. A coagulation attempt was considered when the electrosurgical knife or coagulation forceps was energized or when the tissue was denatured.

Statistical analysis

All statistical analyses were performed using R software (version 4.0.4). The dataset for continuous variables was assessed for normality using the Shapiro-Wilk test. For variables that followed a normal distribution, comparisons were made using Student’s t-test. The before-learning, after-learning, and trainer groups comprised independent datasets, with each comparison aimed at different objectives; therefore, no corrections for multiple comparisons were applied. The χ² or Fisher’s exact test was performed to compare categorical variables. P < 0.05 was considered statistically significant.

RESULTS

Characteristics of patients and lesions treated with ESD

Five gastric ESD procedures performed by trainees and four gastric ESD procedures performed by the experts during phase 1 were collected. Nine videos with tracked viewpoints were used for training. Furthermore, five gastric ESD procedures performed by the trainees and two gastric ESD procedures performed by the experts during phase 2 were collected. Using the videos of these procedures, 10 video clips during phase 1, eight video clips during phase 2, and eight video clips from the videos of ESD performed by the experts were created and analyzed. Significant differences in the characteristics of the patients and lesions treated with ESD during phase 1 and phase 2 were not observed (Table 1). A representative video clip obtained during phase 1 is shown in Video 1.

Table 1 Patient and lesion characteristics.

Characteristics	Phase 1 before learning (n = 5)	Phase 2 after learning (n = 5)	P value (before learning vs after learning)
Patient characteristics
Sex, male, n (%)	4 (80.0)	4 (80.0)	1.000
Age, (years), mean ± SD	71.3 ± 5.2	65.4 ± 14.0	0.324
Use of anticoagulant agents, n (%)	1 (20.0)	2 (40.0)	1.000
Lesion characteristics
Tumor size (mm), mean ± SD	9.4 ± 3.8	9.4 ± 3.7	0.990
Tumor location, U:M:L	0:1:4	0:3:2	0.372
Pathology, pT1a: pT1b	5:0	5:0	0.182

P < 0.05 was considered statistically significant. Age and tumor size are expressed as the mean ± SD. L: Lower; M: Middle; U: Upper; pT1a: Pathological T1a; pT1b: Pathological T1b.

Time required to recognize the bleeding point

The time required to recognize the bleeding point during phase 2 was significantly shorter than that during phase 1 (8.3 ± 4.2 seconds vs 23.1 ± 19.2 seconds, P = 0.049) (Figure 2A). A comparison of the time required to recognize the bleeding point by the experts and that during phase 1 and phase 2 indicated no significant differences (5.6 ± 2.9 seconds vs 23.1 ± 19.2 seconds, P = 0.022; 5.6 ± 2.9 seconds vs 8.3 ± 4.2 seconds, P = 0.141).

Figure 2 Comparison of the study outcomes. A student t-test was used for the comparison. A P-value less than 0.05 was considered significant. A: The time required to recognize a bleeding point. The time required to recognize a bleeding point during phase 2 was significantly shorter (after learning) than that during phase 1 (before learning) (8.3 ± 4.2 seconds vs 23.1 ± 19.2 seconds, P = 0.049); B: The time required to complete hemostasis. The time required to complete hemostasis during phase 1 (before learning) and that during phase 2 (after learning) were not significantly different (15.4 ± 6.8 seconds vs 31.9 ± 21.7 seconds, P = 0.056); C: The number of coagulation attempts. A comparison of the number of coagulation attempts performed during phase 1 (before learning) and phase 2 (after learning) indicated that significantly fewer coagulation attempts were performed during phase 2 (1.8 ± 0.7 vs 3.2 ± 1.0, P = 0.004).

Time required to complete hemostasis

The time required to complete hemostasis during phase 1 and that during phase 2 were not significantly different (15.4 ± 6.8 seconds vs 31.9 ± 21.7 seconds, P = 0.056) (Figure 2B). In contrast, the time required to complete hemostasis by the experts was significantly shorter than that during phase 1 and phase 2 (9.3 ± 2.8 seconds vs 31.9 ± 21.7 seconds, P = 0.010; 9.3 ± 2.8 seconds vs 15.4 ± 6.8 seconds, P = 0.034).

Coagulation attempts

Compared to the number of coagulation attempts performed during phase 1, significantly fewer coagulation attempts were performed during phase 2 (1.8 ± 0.7 vs 3.2 ± 1.0, P = 0.004) (Figure 2C). Similar to the time required to recognize the bleeding point, a comparison of the number of bleeding points indicated that the experts encountered significantly fewer bleeding points than those encountered by the trainees during phase 1; however, when the number of bleeding points encountered by the experts was compared with that encountered by the trainees during phase 2, the difference was not significant (1.4 ± 0.5 vs 3.2 ± 1.0, P < 0.001; 1.4 ± 0.5 vs 1.8 ± 0.7, P = 0.236).

DISCUSSION

Visualizing the viewpoints of trainees during ESD has been difficult, and this difficulty has resulted in major limitations to ESD training. A better understanding of what the trainee sees during ESD contributes to better training. Therefore, knowledge of the viewpoints of the experts during ESD, particularly during hemostasis, during phase 1 and phase 2, may help trainees acquire skills more quickly. The application of eye-tracking technology to medical training, including surgical and endoscopic screening training, has been attempted to allow analyses of the differences in viewpoints among operators. However, previous studies have analyzed the differences in viewpoints of trainees and those of experts or explored the visual gaze patterns associated with screening quality[14,16,17]. Thus, no study has applied this technology to endoscopy training programs. During this study, we improved our previous system used to record videos of the operator’s viewpoint during endoscopy to quantitatively evaluate hemostatic techniques during ESD. This is the first study to report endoscopy training using eye-tracking technology. Another study showed that the ESD procedure time can be significantly shortened by gaining more experience by performing more ESD procedures[18]. The time required for hemostasis comprises a large part of the trainees’ procedure times during ESD[19]. Thus, a shorter time required to complete hemostasis during ESD may be an indirect indicator of ESD skills. During this study, the time required to complete hemostasis during phase 1 and that during phase 2 were not significantly different. However, the time required to recognize the bleeding point and the number of coagulation attempts were significantly improved during phase 2. These results suggest that intensive training programs involving eye-tracking technology may facilitate the acquisition of hemostasis skills. Further randomized controlled trials should be conducted to verify this supposition. The human visual system is highly regulated by fast information processing, which requires only 150 to 200 milliseconds for object recognition, with little variation among individuals[20]. Therefore, the difference in the time required to recognize a bleeding point among operators reflects their ability to correctly identify that point. However, the time required to recognize a bleeding point is not dependent on the difference in the visual function of individuals. During this study, the time required to recognize a bleeding point was significantly shorter during phase 2, indicating that an appropriate training intervention could modify visual perception. During phase 1, the trainees vaguely recognized the bleeding point and could not coagulate the correct bleeding point (Video 1). Several studies have reported significantly different visual gaze patterns of trainees and experts during endoscopy[14,16,17]. Our study revealed that the time required to recognize a bleeding point significantly differed between the trainees and the experts during phase 1. Previously, the only hemostatic technique training methods involved the expert indicating the bleeding point in real time during treatment or the trainees reviewing and reflecting on recorded videos. However, sharing viewpoints between the trainee and the expert, as performed during the training method used for this study, enabled more efficient learning.

Previous studies have shown that at least 30 ESD procedures must be performed to acquire sufficient skills[21,22]. During our study, all trainees had performed fewer than 30 ESD procedures by the end of the study period. Nevertheless, their ability to recognize the bleeding point was improved during phase 2. Hemostasis requires accurate recognition of the bleeding point and skillful manipulation of the scopes and devices[23]. Because of the slow progression of manipulation skills necessary to utilize the scopes and devices, the time to complete hemostasis during phase 2 may not have been significantly reduced. ESD comprises various procedures, including mucosal incision, submucosal dissection, and hemostasis. Previous studies utilizing eye-tracking technology across various surgical operations have successfully identified distinct visual gaze patterns between trainee and trainer[24-26]. These findings suggest that modifying visual gaze patterns could enhance operational skills. Consistent with this evidence, our analyses revealed that visual gaze patterns during ESD vary significantly based on the physician’s skill level. This variability suggests that understanding one’s own visual gaze patterns could facilitate learning. During this study, we used eye-tracking technology to create an objective evaluation system to show the operator’s viewpoint to improve learning and allow modifications of visual gaze patterns during hemostasis. This tool could be applied to improve other skills, such as those required to perform mucosal incisions and submucosal dissection, during ESD. Comprehensive training using this system may contribute to the acquisition of ESD skills. This study had some limitations. First, this was not a randomized controlled trial, so it remains unconfirmed whether training with eye-tracking technology improves outcomes. Second, the small number of enrolled endoscopists may have led to overestimations or underestimations of the results. Third, because the learning methods associated with the recorded videos were not rigorously defined, their effects could have varied among trainees. Fourth, since only five ESD procedures were performed by trainees in each phase, it was not possible to compare parameters of ESD skill acquisition, such as dissection speed. Therefore, large-scale randomized controlled trials should be conducted to overcome these limitations.

CONCLUSION

We successfully measured the time required to recognize a bleeding point using a newly developed video recording system based on eye-tracking technology. Teaching visual gaze patterns using this system could improve endoscopic proficiency.