Evaluating a Novel 3D Stereoscopic Visual Display for Transanal Endoscopic Surgery

A Randomized Controlled Crossover Study

Aimee N. Di Marco, MA, MBBS, MRCS; Jenifa Jeyakumar, BSc; Philip J. Pratt, PhD; Guang-Zhong Yang, FREng; Ara W. Darzi, MD, FRS, FMedSci, FRCS, FACS


Annals of Surgery. 2016;263(1):36-42. 

In This Article


This study was informed by its pilot, with participants carrying out the ex vivo TES task in 4 different visual modalities (Wolf stereoendoscope, novel box viewer, 2D HD, and 3D HD monitor with polarizing glasses), acting as their own controls and crossing over to each modality in a randomly allocated to minimize any learning effect. As a randomized trial, the study was registered with the Consolidated Standards of Reporting Trials (CONSORT).

Study Participants

Participants were recruited from the staff of a 1500-bedded group of 3 acute, teaching hospitals in London, United Kingdom, and medical students from that institution's Bachelor of Science program. Men and women ranging in experience from novices to consultant colorectal surgeons were included. All subjects were screened for stereoscopic vision[27] and were excluded if the test was failed.


The first operative step of a TES procedure was simulated using ex vivo porcine colon, containing a synthetic polyp consisting of a plastic material interposed between a dual layer of porcine bowel. These "polyps" were of a standardized size (median 35 mm, interquartile range [IQR] 34–37 mm) and, to guide the subjects to the intended surface boundary of the lesion and allow judgment of accuracy, the surface of the porcine bowel was marked with 12 circumferential "tattoos" in permanent marker ink (Fig. 2A).

Figure 2.

A, Porcine colon containing synthetic polyp and with circumferential ''tattoo'' markings. B, A participant with the simulated patient in the lithotomy position.

This porcine colon was placed within a phantom pelvis with lower limbs in the lithotomy position, as if excision of a posterior rectal tumor were to be done (the most common location for lesions managed with TES). Participants were given an angled, toothed grasper and pedal-operated diathermy wand (TEO, KARL STORZ Endoscopy (UK) Ltd., Slough, UK), on the tip of which was mounted an electromagnetic tracker (Aurora, Northern Digital International, Waterloo, Canada). They were asked to score around the lesion as in the first step of a TES procedure. The set-up, including endoscope position, illumination, distance between specimen and endoscope, and between viewing device and subject remained constant throughout the study, with participants being able to adjust the variables that would usually be dictated by the surgeon: the height of the operating table and chair, interocular distance of the stereoendoscope, and placement of the instruments in left versus right hands. Four different visual displays were utilized: an LG HD 2D 1920 × 1080 monitor, LG 3D HD 1920 × 1080 with passive polarizing glasses, Wolf 50° 3D stereoendoscope, and the novel stereoscopic box viewer.


Figure 3 shows the pathway of each participant through the study: written consent to participate was gained, a short, standardized explanation of TES and the task was given, and the stereopsis-screening tool was administered.[27] No participants failed this test. All then completed a pretest questionnaire to determine level of experience, visual acuity, and presence of color vision. Participants were then allowed to practice the task until a predetermined level of proficiency was achieved. Once proficiency was reached, participants completed the task 3 times using each visual display modality. The order of the visual displays was randomized to negate any learning effect.[28] After completing the task 3 times on a particular display, participants answered a NASA-TLX questionnaire[29] and moved on to the next display. After a total of 12 tasks, a posttest questionnaire was completed to collect feedback on the subjects' opinions of the visual displays and to assess face validity of the simulation.

Figure 3.

Flowchart showing progress of recruits through the study.

Primary Outcome Measures

Primary outcomes were path length, accuracy, time taken, and task workload, compared between the 4 visual displays and for subgroups based on the experience level of the participants (assessed on the pretest questionnaire). The path length of the diathermy wand was deduced from 3 vector coordinate data generated by the Aurora electromagnetic tracker. Accuracy was assessed by end-product analysis of each specimen, conducted by 2 blinded assessors (AdM and JJ), measuring distance between diathermy mark and target tattoo to the nearest half millimeter. Time taken for each task was obtained from the duration of each video file to centisecond accuracy. The perceived workload of the 3 tasks conducted with each modality was assessed by the participant using the NASA-TLX questionnaire,[29] which uses pair-wise comparisons to generate a weighted score between 0 and 100 (where 100 indicates a higher workload) based on the contribution of mental, physical, and temporal demands and performance, effort, and frustration to the task. The pilot study of 10 participants was used to inform a power calculation, which dictated a minimum sample size of 25 participants. Although the assessors were blinded to the modality, it was not possible to conceal the type of display system being used from the participant.

Secondary Outcome Measures

Participant opinion regarding each of the imaging modalities was assessed using a custom-designed questionnaire, in which each modality was ranked across 5 domains: image quality, visual discomfort, depth perception, ergonomics, and overall preference, using a Likert scale of 1–10 and an opportunity for free text. The questionnaire also assessed face validity of the simulation using the same Likert scale across 7 domains; realism of the overall set-up, the tissue, task, instruments, clinical scenario, perception of whether the simulation would improve performance, and interest in using the simulator in the future.

Errors during tasks, defined as inadvertent diathermy swipes or failing to mark at a tattoo, were recorded. However, the trial was not powered to analyze such errors.

Statistical Analyses

Primary outcome measures were assessed using median values and IQRs calculated using the SPSS statistical package for Windows, version 20.0.[30] Pair-wise comparisons between the 4 modalities and within the experience subgroups were calculated using the Wilcoxon matched-pair signed-rank test for nonparametric data. Inter-rater reliability of the measuring accuracy of assessors was determined using the same methodology. Secondary outcome measures were also expressed as median and IQRs and, where relevant, comparisons were made using the Wilcoxon matched-pair test. The significance level was set at P < 0.05 for all outcomes.