Eye tracking technology in endoscopy: Looking to the future

Eye movement patterns associated with colorectal adenoma detection: Post hoc analysis of randomized controlled trial

Background and study aims

The adenoma detection rate is higher among endoscopists who spend more time observing screen edges during colonoscopies. Nonetheless, eye movement parameters related to lesion detection remain unknown. This study aimed to determine the specific eye movement parameters related to colorectal adenoma detection, including the gaze rate in a particular area and eye movement speed.

Patients and methods

This study was a post hoc analysis of a randomized controlled trial investigating the effect of modifying eye movements of endoscopists on colorectal adenoma detection. Gaze rate at a specific area and eye movement speed were calculated based on endoscopist gaze coordinates in each examination. Time required for observation and treatment of polyps was excluded. The lower peripheral area was defined as the bottom row when the screen was divided into 6×6 sections. These parameters were compared between patients with and without adenomas.

Results

Five physicians performed 158 colonoscopies. The adenoma detection group exhibited a lower peripheral gaze rate (13.7% vs. 9.5%, P = 0.004) and smaller average eye movement distance (29.9 pixels/30 ms vs. 33.3 pixels/30 ms, P = 0.022). Logistic regression analysis showed that a lower peripheral gaze rate > 13.0% and an average eye movement distance <30 pixels/30 ms were increased independent predictors of adenoma detection ( P = 0.024, odds ratio [OR] 2.53, 95% confidence interval [CI] 1.71-3.28; P = 0.045, OR 4.57, 95% CI 1.03-20.2), whereas age, sex, and withdrawal time were not.

Conclusions

Lower peripheral gaze rate and slow eye movement are associated with colorectal adenoma detection.

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

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.

Eye-tracking dataset of endoscopist-AI teaming during colonoscopy: Retrospective and real-time acquisition

Recent studies have demonstrated that integrating AI into colonoscopy procedures significantly improves the adenoma detection rate (ADR) and reduces the adenoma miss rate (AMR). However, few studies address the critical issue of endoscopist-AI collaboration in real-world settings. Eye-tracking data collection is considered a promising approach to uncovering how endoscopists and AI interact and influence each other during colonoscopy procedures. A common limitation of existing studies is their reliance on retrospective video clips, which fail to capture the dynamic demands of real-time colonoscopy, where endoscopists must simultaneously navigate the colonoscope and identify lesions on the screen. To address this gap, we established a dataset to analyze changes in endoscopists' eye movements during the colonoscopy withdrawal phase. Eye-tracking data was collected from graduate students, nurses, senior endoscopists, and novice endoscopists while they reviewed retrospectively recorded colonoscopy withdrawal videos, both with and without computer-aided detection (CADe) assistance. Furthermore, 80 real-time video segments were prospectively collected during endoscopists' actual colonoscopy withdrawal procedures, comprising 43 segments with CADe assistance and 37 segments without assistance (normal control).

The Evidence-Based Medicine Management of Endometriosis Should Be Updated for the Limitations of Trial Evidence, the Multivariability of Decisions, Collective Experience, Heuristics, and Bayesian Thinking

Background/Objectives: The diagnosis and treatment of endometriosis should be based on the best available evidence. Emphasising the risk of bias, the pyramid of evidence has the double-blind, randomised controlled trial and its meta-analyses on top. After the grading of all evidence by a group of experts, clinical guidelines are formulated using well-defined rules. Unfortunately, the impact of evidence-based medicine (EBM) on the management of endometriosis has been limited and, possibly, occasionally harmful. Methods: For this research, the inherent problems of diagnosis and treatment were discussed by a working group of endometriosis and EBM specialists, and the relevant literature was reviewed. Results: Most clinical decisions are multivariable, but randomized controlled trials (RCTs) cannot handle multivariability because adopting a factorial design would require prohibitively large cohorts and create randomization problems. Single-factor RCTs represent a simplification of the clinical reality. Heuristics and intuition are both important for training and decision-making in surgery; experience, Bayesian thinking, and learning from the past are seldom considered. Black swan events or severe complications and accidents are marginally discussed in EBM since trial evidence is limited for rare medical events. Conclusions: The limitations of EBM for managing endometriosis and the complementarity of multivariability, heuristics, Bayesian thinking, and experience should be recognized. Especially in surgery, the value of training and heuristics, as well as the importance of documenting the collective experience and of the prevention of complications, are fundamental. These additions to EBM and guidelines will be useful in changing the Wild West mentality of surgery resulting from the limited scope of EBM data because of the inherent multivariability, combined with the low number of similar interventions.

Objective Evaluation of Gaze Location Patterns Using Eye Tracking During Cystoscopy and Artificial Intelligence-Assisted Lesion Detection

Background: The diagnostic accuracy of cystoscopy varies according to the knowledge and experience of the performing physician. In this study, we evaluated the difference in cystoscopic gaze location patterns between medical students and urologists and assessed the differences in their eye movements when simultaneously observing conventional cystoscopic images and images with lesions detected by artificial intelligence (AI). Methodology: Eye-tracking measurements were performed, and observation patterns of participants (24 medical students and 10 urologists) viewing images from routine cystoscopic videos were analyzed. The cystoscopic video was captured preoperatively in a case of initial-onset noninvasive bladder cancer with three low-lying papillary tumors in the posterior, anterior, and neck areas (urothelial carcinoma, high grade, and pTa). The viewpoint coordinates and stop times during observation were obtained using a noncontact type of gaze tracking and gaze measurement system for screen-based gaze tracking. In addition, observation patterns of medical students and urologists during parallel observation of conventional cystoscopic videos and AI-assisted lesion detection videos were compared. Results: Compared with medical students, urologists exhibited a significantly higher degree of stationary gaze entropy when viewing cystoscopic images (p < 0.05), suggesting that urologists with expertise in identifying lesions efficiently observed a broader range of bladder mucosal surfaces on the screen, presumably with the conscious intent of identifying pathologic changes. When the participants observed conventional and AI-assisted lesion detection images side by side, contrary to urologists, medical students showed a higher proportion of attention directed toward AI-detected lesion images. Conclusion: Eye-tracking measurements during cystoscopic image assessment revealed that experienced specialists efficiently observed a wide range of video screens during cystoscopy. In addition, this study revealed how lesion images detected by AI are viewed. Observation patterns of observers' gaze may have implications for assessing and improving proficiency and serving educational purposes. To the best of our knowledge, this is the first study to utilize eye tracking in cystoscopy. University of Tsukuba Hospital, clinical research reference number R02-122.

Wearable Near-Eye Tracking Technologies for Health: A Review

With the rapid advancement of computer vision, machine learning, and consumer electronics, eye tracking has emerged as a topic of increasing interest in recent years. It plays a key role across diverse domains including human-computer interaction, virtual reality, and clinical and healthcare applications. Near-eye tracking (NET) has recently been developed to possess encouraging features such as wearability, affordability, and interactivity. These features have drawn considerable attention in the health domain, as NET provides accessible solutions for long-term and continuous health monitoring and a comfortable and interactive user interface. Herein, this work offers an inaugural concise review of NET for health, encompassing approximately 70 related articles published over the past two decades and supplemented by an in-depth examination of 30 literatures from the preceding five years. This paper provides a concise analysis of health-related NET technologies from aspects of technical specifications, data processing workflows, and the practical advantages and limitations. In addition, the specific applications of NET are introduced and compared, revealing that NET is fairly influencing our lives and providing significant convenience in daily routines. Lastly, we summarize the current outcomes of NET and highlight the limitations.

Eye-controlled endoscopy - a benchtop trial of a novel robotic steering platform - iGAZE2

The endoscopic control system has remained similar in design for many decades The remit of advanced therapeutic endoscopy continues to expand requiring precision control and high cognitive workloads. Robotic systems are emerging, but all still require bimanual control and expensive and large new systems. Eye tracking is an exciting area that can be used as an endoscope control system. This is a study to establish the feasibility of an eye-controlled endoscope and compare its performance and cognitive demand to use of a conventional endoscope. An eye gaze-control system consisting of eye-tracking glasses, customised software and a small motor unit was built and attached to a conventional endoscope. Twelve non-endoscopists used both the eye gaze system and a conventional endoscope to complete a benchtop task in a simulated oesophagus and stomach. Completion of tasks was timed. Subjective feedback was collected from each participant on task load using the NASA Task Load Index. Participants were significantly quicker completing the task using iGAZE2 vs a conventional endoscope (65.02 ± 16.34s vs 104.21 ± 51.31s, p = 0.013) Participants were also significantly quicker completing retroflexion using iGAZE2 vs a conventional endoscope (8.48 ± 3.08 vs 11.38 ± 5.36s, p = 0.036). Participants reported a significantly lower workload (raw NASA-TLX score) when using iGAZE2 vs the conventional endoscope (152.1 ± 63.4 vs 319.6 ± 81.6, p = 0.0001) (Fig. 7). Users found iGAZE2 to have a significantly lower temporal demand, mental demand, effort, mental demand, physical demand, and frustration level. The eye gaze system is an exciting, small, and retrofittable system to any endoscope. The system shows exciting potential as a novel endoscopic control system with a significantly lower workload and better performance in novices suggesting a more intuitive control system.

Human-artificial intelligence interaction in gastrointestinal endoscopy

The number and variety of applications of artificial intelligence (AI) in gastrointestinal (GI) endoscopy is growing rapidly. New technologies based on machine learning (ML) and convolutional neural networks (CNNs) are at various stages of development and deployment to assist patients and endoscopists in preparing for endoscopic procedures, in detection, diagnosis and classification of pathology during endoscopy and in confirmation of key performance indicators. Platforms based on ML and CNNs require regulatory approval as medical devices. Interactions between humans and the technologies we use are complex and are influenced by design, behavioural and psychological elements. Due to the substantial differences between AI and prior technologies, important differences may be expected in how we interact with advice from AI technologies. Human–AI interaction (HAII) may be optimised by developing AI algorithms to minimise false positives and designing platform interfaces to maximise usability. Human factors influencing HAII may include automation bias, alarm fatigue, algorithm aversion, learning effect and deskilling. Each of these areas merits further study in the specific setting of AI applications in GI endoscopy and professional societies should engage to ensure that sufficient emphasis is placed on human-centred design in development of new AI technologies.