Accuracy of endoscopists’ estimate of polyp size: A continuous dilemma

Measuring Size of Colorectal Polyps Using a Virtual Scale Endoscope or Visual Assessment: A Randomized Controlled Trial

INTRODUCTION

This study aimed to compare the accuracy of polyp size measurements using a virtual scale endoscope (VSE) with an integrated laser-based adaptive scale function and visual assessment (VA) during colonoscopies.

METHODS

We conducted a single-blinded, prospective randomized controlled trial. Eligible patients (aged 45-80 years) undergoing screening, surveillance, or diagnostic colonoscopies were randomly assigned (1:1) into 2 groups. In the intervention group, all detected polyps were measured for size using VSE; in the control group, all polyps were measured using VA. Size measurements were compared with a reference standard of digital caliper measurement immediately post polypectomy. The primary outcome was the relative accuracy of real-time VSE measurement compared with VA. Secondary outcomes included the mean differences and the correlations between VSE or VA sizes and the reference standard of measurement.

RESULTS

Overall, 230 patients were enrolled and randomized. The relative size measurement accuracy of VSE was 84% in 118 polyps, which was significantly higher than that of VA (105 polyps; 68.4%, P < 0.001). VSE resulted in a significantly higher percentage of size measurements within 25% of true size compared with VA (81.4% vs 41%, P < 0.001). VSE had a significantly lower percentage for >5-mm polyps incorrectly sized as 1-5 mm compared with VA (13.5% vs 57.1%; P < 0.001) and a significantly lower percentage for >3-mm polyps incorrectly sized as 1-3 mm compared with VA (11.3% vs 56.5%; P < 0.001).

DISCUSSION

VSE significantly improves the size measurement accuracy of colorectal polyps during colonoscopies compared with VA and results in fewer misclassifications at relevant decision-making size thresholds.

Leveraging a realistic synthetic database to learn Shape-from-Shading for estimating the colon depth in colonoscopy images

Colonoscopy is the choice procedure to diagnose, screening, and treat the colon and rectum cancer, from early detection of small precancerous lesions (polyps), to confirmation of malign masses. However, the high variability of the organ appearance and the complex shape of both the colon wall and structures of interest make this exploration difficult. Learned visuospatial and perceptual abilities mitigate technical limitations in clinical practice by proper estimation of the intestinal depth. This work introduces a novel methodology to estimate colon depth maps in single frames from monocular colonoscopy videos. The generated depth map is inferred from the shading variation of the colon wall with respect to the light source, as learned from a realistic synthetic database. Briefly, a classic convolutional neural network architecture is trained from scratch to estimate the depth map, improving sharp depth estimations in haustral folds and polyps by a custom loss function that minimizes the estimation error in edges and curvatures. The network was trained by a custom synthetic colonoscopy database herein constructed and released, composed of 248400 frames (47 videos), with depth annotations at the level of pixels. This collection comprehends 5 subsets of videos with progressively higher levels of visual complexity. Evaluation of the depth estimation with the synthetic database reached a threshold accuracy of 95.65%, and a mean-RMSE of 0.451cm, while a qualitative assessment with a real database showed consistent depth estimations, visually evaluated by the expert gastroenterologist coauthoring this paper. Finally, the method achieved competitive performance with respect to another state-of-the-art method using a public synthetic database and comparable results in a set of images with other five state-of-the-art methods. Additionally, three-dimensional reconstructions demonstrated useful approximations of the gastrointestinal tract geometry. Code for reproducing the reported results and the dataset are available at https://github.com/Cimalab-unal/ColonDepthEstimation.

Estimating the size of colorectal polyps endoscopically: random guess or systematic error?

BACKGROUND

Size of colorectal polyps reflects potential for malignancy and helps define advanced lesions. Studies measuring ability of endoscopists to estimate polyp size show significant variation. The aim of this study was to determine if there is a linear relationship between endoscopic and pathologic polyp size.

METHODS

Data for adenomas removed completely by snare, in one piece, were retrieved from a prospectively recorded polyp database. Endoscopic estimate of maximum diameter was compared to that on the pathology report by linear regression analysis.

RESULTS

There were 126 polyps in 126 patients, 85 men and 41 women. Mean age was 63.2 ± 12.9 years. Mean endoscopic polyp size was 12.2 ± 9.3 mm and mean pathology size was 9.3 ± 6.9 mm. Endoscopically, 16 polyps were ≤ 5 mm, 62 were from 6 to 10 mm, 21 were from 11 to 15 mm, and 27 were from 16 to 55 mm. Twenty-nine polyps were right sided, 86 were left and 11 were rectal. Regression of endoscopic size against pathology size yielded a significant r² of 0.761. Using the regression formula of endoscopic size = 0.7 + 1.175× pathology size an endoscopic estimate of 10 mm (= advanced adenoma) means a pathologic size of 8 mm. For a pathologic size of 10 mm, an endoscopic estimate of 12 mm is needed. A large polyp is ≥20 mm; for this endoscopist a 20 mm polyp is really 16.4 mm.

CONCLUSIONS

The relationship between endoscopic and directly measured size is linear over all polyp diameters, and likely represents a systematic error.

Comparing size measurement of colorectal polyps using a novel virtual scale endoscope, endoscopic ruler or forceps: A preclinical randomized trial

Background and study aims  Accurate polyp size measurement is important for guideline conforming choice of polypectomy techniques and subsequent surveillance interval assignments. Some endoscopic tools (biopsy forceps [BF] or endoscopic rulers [ER]) exist to help with visual size estimation. A virtual scale endoscope (VSE) has been developed that allows superimposing a virtual measurement scale during live endoscopies. Our aim was to evaluate the performance of VSE when compared to ER and BF-based measurement. Methods  We conducted a preclinical randomized trial to evaluate the relative accuracy of size measurement of simulated colorectal polyps when using: VSE, ER, and BF. Six endoscopists performed 60 measurements randomized at a 1:1:1 ratio using each method. Primary outcome was relative accuracy in polyp size measurement. Secondary outcomes included misclassification of sizes at the 5-, 10-, and 20-mm thresholds. Results  A total of 360 measurements were performed. The relative accuracy of BF, ER, and VSE was 78.9 % (95 %CI = 76.2-81.5), 78.4 % (95 %CI = 76.0-80.8), and 82.7 % (95 %CI = 80.8-84.8). VSE had significantly higher accuracy compared to BF ( P  = 0.02) and ER ( P  = 0.006). VSE misclassified a lower percentage of polyps > 5 mm as ≤ 5 mm (9.4 %) compared to BF (15.7 %) and ER (20.9 %). VSE misclassified a lower percentage of ≥ 20 mm polyps as < 20 mm (8.3 %) compared with BF (66.7 %) and ER (75.0 %). Of polyps ≥10mm, 25.6 %, 25.5 %, and 22.5 % were misclassified as <10 mm with ER, BF, and VSE, respectively. Conclusions  VSE had significantly higher relative accuracy in measuring polyps compared to ER or BF assisted measurement. VSE improves correct classification of polyps at clinically important size thresholds.

Artificial intelligence-based measurement outperforms current methods for colorectal polyp size measurement

OBJECTIVES

An accurate polyp size estimation during colonoscopy is crucial to determine the surveillance interval and predict the risk of malignant progression. However, there is a high degree of subjectivity in estimating polyp size among endoscopists in clinical practice. We aimed to assess the efficacy of a novel method that uses artificial intelligence (AI) to measure the size of colon polyps and compare it with current approaches.

METHODS

Using the W-Net model for vessel segmentation and based on retinal image datasets (DRIVE, STARE, CHASE-DB, and HRF) and colonoscopy images, we developed the bifurcation-to-bifurcation (BtoB) distance measuring method and applied it to endoscopic images. Measurements were compared with those obtained by eight endoscopists (four expert and four trainees). Diagnostic ability and reliability were evaluated using Lin's concordance correlation coefficients (CCCs) and Bland-Altman analyses.

RESULTS

For both experts and trainees, visually estimated sizes of the same polyp were significantly inconsistent depending on the camera view used (P < 0.001). Bland-Altman analyses showed that there was a trend toward underestimation of the sizes of the polyps in both groups, especially for polyps larger than 10 mm. The new technique was highly accurate and reliable in measuring the size of colon polyp (CCC, 0.961; confidence interval 0.926-0.979), clearly outperforming the visual estimation and open biopsy forceps methods.

CONCLUSION

The new AI measurement method improved the accuracy and reliability of polyp size measurements in colonoscopy images. Incorporating AI might be particularly important to improve the efficiency of trainees at estimating polyp size during colonoscopy.

Prevalence of Forceps Polypectomy of Nondiminutive Polyps Is Substantial But Modifiable

BACKGROUND AND AIMS

The use of forceps for removal of nondiminutive polyps is associated with incomplete resection compared with snare polypectomy. However, few studies have characterized the frequency of forceps polypectomy for nondiminutive polyps or identified strategies to improve this practice. To address this gap, we estimated the prevalence and predictors of forceps polypectomy in clinical practice and examined the effectiveness of a multicomponent intervention to reduce inappropriate forceps polypectomy.

METHODS

We retrospectively reviewed all colonoscopies with polypectomies performed at 2 U.S. health systems between October 1, 2017, and September 30, 2019. We used a mixed-effects logistic regression model to examine the effect of a multicomponent intervention, including provider education and a financial incentive, to reduce inappropriate forceps polypectomy, defined as use of forceps polypectomy for polyps ≥5 mm.

RESULTS

A total of 9968 colonoscopies with 25,534 polypectomies were performed by 42 gastroenterologists during the study period. Overall, 8.5% (n = 2176) of polyps were removed with inappropriate forceps polypectomy. Inappropriate forceps polypectomy significantly decreased after the intervention (odds ratio [OR], 0.34, 95% confidence interval [CI], 0.30-0.39), from 11.4% (n = 1539) to 5.3% (n = 637). Predictors of inappropriate forceps polypectomy included inadequate bowel prep (OR, 1.25; 95% CI, 1.06-1.47), polyps in the right colon (vs left: OR, 1.29; 95% CI, 1.09-1.51), and number of polyps removed (OR, 0.96; 95% CI, 0.94-0.97). Inappropriate forceps polypectomy also varied by gastroenterologist (median OR, 3.43). In a post hoc analysis, the proportion of polyps >2 mm removed with forceps decreased from 50.0% before the intervention to 43.0% after it (OR, 0.62; 95% CI, 0.58-0.68).

CONCLUSIONS

Inappropriate forceps polypectomy is common but modifiable. The proportion of nondiminutive polyps removed with forceps polypectomy should be considered as a quality measure.

Localization of Small Neoplastic Lesions in Colonoscopy by Estimating Edge, Texture and Motion Saliency

Early screening in Colorectal Cancer consists in finding and removing small precancerous masses or neoplastic lesions developed from the mucosa, usually lesions smaller than 10 mm. Localization of small neoplastic lesions is a very challenging task since colon exploration is highly dependent on the expert training and colon preparation. Several strategies have attempted to locate neoplasias, but usually they are huge lesions that a trained gastroenterologist could hardly miss. This work presents a saliency-based strategy to localize polypoid and non-polypoid neoplastic lesions smaller than 10 mm in colonoscopy videos by combining spatio-temporal descriptors. For doing so, a per-frame-multi-scale representation is computed and edge, texture and motion features are extracted. Each of these features is used to construct a primary saliency map which are then combined to obtain a coarse saliency map. Finally, the neoplasia is localized as the bounding box of a circular region, approximated by the Hough transform, with the largest salience. The proposed approach was evaluated in 8 short colonoscopy videos obtaining an average of Annotated Area Covered of 0.75 and a precision of 0.82.

Sizing of Polyp Illustrations Differs by Endoscopists’ Gender and Improves With a Measurement Reference

Colorectal cancer surveillance intervals by colonoscopy are based on the size and number of polyps removed. Evidence suggests endoscopists’ estimation of polyp size is often inaccurate, but the differences by endoscopists’ characteristics have not been reported. This study assesses endoscopists’ accuracy of measuring polyp illustrations, the effect of endoscopists’ characteristics, and the impact of having a measurement reference. Endoscopists in a community-based, gastroenterology practice estimated the size of several illustrations in a booklet. One month later, they estimated the size of illustrations with a provided measurement reference. Accuracy was defined as no difference between estimated and actual value. Endoscopists were accurate in sizing only 15% of the time, with a tendency toward undersizing. Female endoscopists, those with less than 10 years in practice and those with lower adenoma detection rates, were more likely to undersize polyps. Accuracy of measuring the polyp illustrations increased to 50% ( p < .01) with the measurement reference. The improvement in accuracy was seen across endoscopists’ demographic groups. Endoscopists had poor accuracy of measuring polyp illustrations. Almost universally, endoscopists tended to undersize the polyp illustrations. Accuracy improved significantly with the use of a polyp-measuring guide, particularly when considering important surveillance thresholds of 5 and 10 mm.

Variation between Pathological Measurement and Endoscopically Estimated Size of Colonic Polyps

Background and Aims

Accurate determination of colonic polyp size is vital to an appropriate surveillance. The main aim of this study was to evaluate variation between the polyp size reported by the endoscopist and its pathological measurement.

Methods

A retrospective analysis of all colonic adenomatous polyps resected in a 12-month period was performed at our center. Endoscopic and pathological size for each polyp were compared, and overestimation rates, underestimation rates, and endoscopic-pathological variation (EPV) were calculated.

Results

Among the 573 polyps that were included, the mean endoscopic and pathological sizes were 8.00 and 6.66 mm, respectively. The most frequent error, in 62.1%, was overestimation by the colonoscopist. Overestimation and EPV were associated with resection technique (higher in endoscopic mucosal resection and smaller with biopsy forceps) and colonoscopist. They were not associated with years of experience in colonoscopy. Overestimation was more frequent in larger polyps.

Conclusions

Our study shows significant discordance between endoscopic and pathological size of colonic polyps with a clear tendency for endoscopic overestimation. Larger polyps are more difficult to accurately assess than smaller ones. This propensity for error was not related to colonoscopist's years of experience and seems to be an individual tendency.

A structured light laser probe for gastrointestinal polyp size measurement: a preliminary comparative study

BACKGROUND AND STUDY AIMS

 Polyp size measurement is an important diagnostic step during gastrointestinal endoscopy, and is mainly performed by visual inspection. However, lack of depth perception and objective reference points are acknowledged factors contributing to measurement errors in polyp size. In this paper, we describe the proof-of-concept of a polyp measurement device based on structured light technology for future endoscopes.

PATIENTS AND METHODS

 Measurement accuracy, time, user confidence, and satisfaction were evaluated for polyp size assessment by (a) visual inspection, (b) open biopsy forceps of known size, (c) ruled snare, and (d) structured light probe, for a total of 392 independent polyp measurements in ex vivo porcine stomachs.

RESULTS

 Visual assessment resulted in a median estimation error of 2.2 mm, IQR = 2.6 mm. The proposed probe can reduce the error to 1.5 mm, IQR = 1.67 mm ( P  = 0.002, 95 %CI) and its performance was found to be statistically similar to using forceps for reference ( P  = 0.81, 95 %CI) or ruled snare ( P  = 0.99, 95 %CI), while not occluding the tool channel. Timing performance with the probe was measured to be on average 54.75 seconds per polyp. This was significantly slower than visual assessment (20.7 seconds per polyp, P  = 0.005, 95 %CI) but not significantly different from using a snare (68.5 seconds per polyp, P  = 0.73, 95 %CI). However, the probe's timing performance was partly due to lens cleaning problems in our preliminary design. Reported average satisfaction on a 0 - 10 range was highest for the proposed probe (7.92), visual assessment (7.01), and reference forceps (7.82), while significantly lower for snare users with a score of 4.42 ( P  = 0.035, 95 %CI).

CONCLUSIONS

 The common practice of visual assessment of polyp size was found to be significantly less accurate than tool-based assessment, but easy to carry out. The proposed technology offers an accuracy on par with using a reference tool or ruled snare with the same satisfaction levels of visual assessment and without occluding the tool channel. Further study will improve the design to reduce the operating time by integrating the probe within the scope tip.

Correlation between adenoma detection rate and novel quality indicators for screening colonoscopy. A proposal for quality measures tool kit

BACKGROUND AND AIMS

Adenoma detection rate (ADR) is the most validated quality indicator for colonoscopy. Calculating ADR may not yield information in regards to advanced adenoma detection rate (advanced ADR). In addition, calculating ADR for individual endoscopists cannot distinguish between those who find only one versus more than one adenoma per colonoscopy. Several novel quality indictors were recently proposed to ensure adequate assessment of quality during colonoscopy. Our study aims to determine the correlation between ADR and novel quality indicators.

METHODS

A retrospective cohort study of patients undergoing screening colonoscopy in a university hospital setting. Patient characteristics and colonoscopy findings were combined and analyzed to calculate the correlation of ADR with novel quality indicators using Spearman's rank-order correlation were used.

RESULTS

A total of 1433 patients out of 2116 patients met the inclusion criteria. There was a significant positive correlation between ADR correlated with [advanced-ADR-2, nonadvanced-ADR, adenoma per colonoscopy, Multiplicity detection rate and ADR-Plus] r = (0.82, 0.99, 0.99, 0.07 and 0.85), respectively. However, ADR did not correlate with advanced-ADR and adenomas per positive participant.

CONCLUSION

Adding advanced-ADR and adenomas per positive participant to ADR may create a more comprehensive quality indicators tool kit, which is sensitive and difficult to game. Future studies are needed to investigate the impact of the tool kit on the interval cancers and adenoma missing rate.

Measurement Bias of Polyp Size at Colonoscopy

BACKGROUND

The success of current and proposed strategies to reduce colorectal cancer (CRC) incidence and mortality rates are fundamentally based on measurement accuracy.

OBJECTIVE

The aim of this study was to evaluate the densities of colorectal polyps individually measured at colonoscopy and whether measurement bias is a systemic phenomenon among colonoscopists.

DESIGN

A population-wide, observational study.

SETTING

All hospitals of the government-funded health system in Brisbane, Australia.

PATIENTS

Our study investigated measurement bias at colonoscopy through systematic analysis of 8,591 individual polyp measurements recorded from 12,597 colonoscopies. All colonoscopies performed over a 12-month period between December 1, 2014, and November 30, 2015, were included.

RESULTS

A total of 12,597 electronic colonoscopy reports were individually reviewed, hospital-by-hospital, and 8,591 individual size measurements from 18,276 detected polyps (47%) were obtained.

LIMITATIONS

Our study is limited because the true size of unresected polyps was unknown. We chose not to compare pathologic and histologic sizes as resection specimens sent to pathologists are morphologically different and are measured differently to the pre-resection polyp images seen by endoscopists.

CONCLUSIONS

Colonoscopists may be inaccurate in the measurement of polyp size and appear biased towards and against certain size measurements. These findings cast doubt over the validity of international post-polypectomy surveillance guidelines and the safety of optical diagnosis as a potential management paradigm for diminutive colorectal polyps. They also question the historical accuracy of polyp size data and risk estimates upon which these strategies were based.