Adenoma detection rates decline with increasing procedural hours in an endoscopist's workload

Association of Intensive Endoscopic Burden with Esophageal Cancer Detection: A Nationwide Cohort Study

Background/Aims

Early diagnosis of esophageal cancer (EC) remains challenging despite the increasing frequency of endoscopic screenings globally. The rapidly increasing number of endoscopic screenings performed over a certain period might influence diagnostic performance. This study evaluated the association between the number of endoscopic screenings and EC detection rates in a nationwide cohort.

Methods

This retrospective population-based study used the Korean National Cancer Screening Program database, comprising 32,774,742 males and females aged ≥40 years between 2015 and 2019. Negative binomial regression model and least-squares mean evaluation were used to assess the association between month of the year and EC detection rates.

Results

This study enrolled 28,032,590 participants who underwent upper endoscopy. The number of participants in the fourth quarter (October to December: 10,923,142 [39.0%]) was 2.1 times higher than that in the first quarter (January to March: 5,085,087 [18.1%]); this trend continued for all 5 years. Contrarily, detection rates for EC in the fourth quarter (0.08/1,000 person) were half that in the first quarter (0.15/1,000 person). The odds of detecting EC were lowest in November; in 2015 the odds were 0.57 (95% confidence interval, 0.41 to 0.79; p=0.001) times lower and in 2016, they were 0.51 (95% confidence interval, 0.37 to 0.68; p<0.001) times lower compared to January. The predicted detection rates showed a decreasing trend toward the end of the year (p>0.05 for all).

Conclusions

The workload of endoscopists increased excessively with the rising number of endoscopies toward the end of the year, which was reflected by the decreased EC detection rates during this period.

Effectiveness of using a patient education mobile application to improve the quality of bowel preparation: a randomised controlled trial

AIMS

To determine the effectiveness of a mobile application (app) in improving the quality of bowel preparation for colonoscopy.

METHOD

An endoscopist-blinded randomised controlled trial enrolled patients who were undergoing a colonoscopy on the same day of bowel preparation. The intervention used a Vietnamese mobile app that provides instructions on bowel preparation while patients in the comparison group received conventional instructions. Outcomes included the Boston Bowel Preparation Scale (BBPS) to assess the quality of bowel preparation and the polyp detection rate (PDR) and adenoma detection rate (ADR).

RESULTS

The study recruited 515 patients (256 in the intervention group). The median age was 42 years, 50.9% were females, 69.1% high school graduates and higher, and 45.2% from urban area. Patients in the intervention group had higher adherence to instructions (60.9% vs 52.4%, p=0.05) and longer length of taking laxatives (mean difference 0.17 hours, 95% CI 0.06 to 0.27). The intervention did not reduce the risk of poor bowel cleansing (total BBPS<6) in both overall (7.4% vs 7.7%; risk ratio 0.96, 95% CI 0.53 to 1.76) and subgroup analysis. PDR and ADR were similar between the two groups.

CONCLUSIONS

The mobile app providing instructions on proper bowel preparation improved the practice during bowel preparation but did not improve the quality of bowel cleansing or PDR.

An Italian prospective multicenter study on colonoscopy practice and quality: What has changed in the last 10 years

BACKGROUND

A relevant number of adenomas can be missed during colonoscopy.

AIMS

Assess the current status of colonoscopy procedures in Italian centers.

METHODS

A prospective observational study involving 17 hospitals (34 endoscopists) included consecutive patients undergoing standard colonoscopy. In the first phase, endoscopists performed consecutive colonoscopies. In the second phase, retraining via an online learning platform was planned, while in the third phase data were collected analogously to phase 1.

RESULTS

A total of 3,504 patients were enrolled. Overall, a BBPS score ≥6 was obtained in 95.6% of cases (94.8% and 96.9% in the pre- and post-training phases, respectively). 88.4% of colonoscopies had a withdrawal time ≥6 min (88.2% and 88.7% in the pre- and post-training phases). Median adenoma detection rate (ADR) was 39.1%, with no significant differences between the pre- and post-training phases (40.1% vs 36.9%; P = 0.83). In total, 81% of endoscopists had a ADR performance above the 25% threshold.

CONCLUSION

High colonoscopy quality standards are achieved by the Italian hospitals involved. Quality improvement initiatives and repeated module-based colonoscopy-training have been promoted in Italy during the last decade, which appear to have had a significant impact on quality colonoscopy metrics together with the activation of colorectal cancer screening programs.

Association Between Endoscopist Specialty and Colonoscopy Quality: A Systematic Review and Meta-analysis

BACKGROUND & AIMS

Colonoscopy quality indicators provide measurable assessments of performance, but significant provider-level variations exist. We performed a systematic review and meta-analysis to assess whether endoscopist specialty is associated with adenoma detection rate (ADR) - the primary outcome - or cecal intubation rate, adverse event rates, and post-colonoscopy colorectal cancer rates.

METHODS

We searched EMBASE, Google Scholar, MEDLINE, and the Cochrane Central Registry of Controlled Trials from inception to December 14, 2020. Two reviewers independently screened titles and abstracts. Citations underwent duplicate full-text review, with disagreements resolved by a third reviewer. Data were abstracted in duplicate. The DerSimonian and Laird random effects model was used to calculate pooled odds ratios (ORs) with respective 95% confidence intervals (CIs). Risk of bias was assessed using Risk of Bias in Non-randomised Studies of Interventions.

RESULTS

Of 11,314 citations, 36 studies representing 3,500,832 colonoscopies were included. Compared with colonoscopies performed by gastroenterologists, those by surgeons were associated with lower ADRs (OR, 0.81; 95% CI, 0.74-0.88) and lower cecal intubation rates (OR, 0.76; 95% CI, 0.63-0.92). Compared with colonoscopies performed by gastroenterologists, those by other (non-gastroenterologist, non-surgeon) endoscopists were associated with lower ADRs (OR, 0.91; 95% CI, 0.87-0.96), higher perforation rates (OR, 3.02; 95% CI, 1.65-5.51), and higher post-colonoscopy colorectal cancer rates (OR, 1.23; 95% CI, 1.14-1.33). Substantial to considerable heterogeneity existed for most analyses, and overall certainty in the evidence was low according to the Grading of Recommendations, Assessment, Development, and Evaluations framework.

CONCLUSION

Colonoscopies performed by surgeons or other endoscopists were associated with poorer quality metrics and outcomes compared with those performed by gastroenterologists. Targeted quality improvement efforts may be warranted.

Impact of the Volume and Distribution of Training Datasets in the Development of Deep-Learning Models for the Diagnosis of Colorectal Polyps in Endoscopy Images

Background: Establishment of an artificial intelligence model in gastrointestinal endoscopy has no standardized dataset. The optimal volume or class distribution of training datasets has not been evaluated. An artificial intelligence model was previously created by the authors to classify endoscopic images of colorectal polyps into four categories, including advanced colorectal cancer, early cancers/high-grade dysplasia, tubular adenoma, and nonneoplasm. The aim of this study was to evaluate the impact of the volume and distribution of training dataset classes in the development of deep-learning models for colorectal polyp histopathology prediction from endoscopic images. Methods: The same 3828 endoscopic images that were used to create earlier models were used. An additional 6838 images were used to find the optimal volume and class distribution for a deep-learning model. Various amounts of data volume and class distributions were tried to establish deep-learning models. The training of deep-learning models uniformly used no-code platform Neuro-T. Accuracy was the primary outcome on four-class prediction. Results: The highest internal-test classification accuracy in the original dataset, doubled dataset, and tripled dataset was commonly shown by doubling the proportion of data for fewer categories (2:2:1:1 for advanced colorectal cancer: early cancers/high-grade dysplasia: tubular adenoma: non-neoplasm). Doubling the proportion of data for fewer categories in the original dataset showed the highest accuracy (86.4%, 95% confidence interval: 85.0–97.8%) compared to that of the doubled or tripled dataset. The total required number of images in this performance was only 2418 images. Gradient-weighted class activation mapping confirmed that the part that the deep-learning model pays attention to coincides with the part that the endoscopist pays attention to. Conclusion: As a result of a data-volume-dependent performance plateau in the classification model of colonoscopy, a dataset that has been doubled or tripled is not always beneficial to training. Deep-learning models would be more accurate if the proportion of fewer category lesions was increased.

No-Code Platform-Based Deep-Learning Models for Prediction of Colorectal Polyp Histology from White-Light Endoscopy Images: Development and Performance Verification

Background: The authors previously developed deep-learning models for the prediction of colorectal polyp histology (advanced colorectal cancer, early cancer/high-grade dysplasia, tubular adenoma with or without low-grade dysplasia, or non-neoplasm) from endoscopic images. While the model achieved 67.3% internal-test accuracy and 79.2% external-test accuracy, model development was labour-intensive and required specialised programming expertise. Moreover, the 240-image external-test dataset included only three advanced and eight early cancers, so it was difficult to generalise model performance. These limitations may be mitigated by deep-learning models developed using no-code platforms. Objective: To establish no-code platform-based deep-learning models for the prediction of colorectal polyp histology from white-light endoscopy images and compare their diagnostic performance with traditional models. Methods: The same 3828 endoscopic images used to establish previous models were used to establish new models based on no-code platforms Neuro-T, VLAD, and Create ML-Image Classifier. A prospective multicentre validation study was then conducted using 3818 novel images. The primary outcome was the accuracy of four-category prediction. Results: The model established using Neuro-T achieved the highest internal-test accuracy (75.3%, 95% confidence interval: 71.0–79.6%) and external-test accuracy (80.2%, 76.9–83.5%) but required the longest training time. In contrast, the model established using Create ML-Image Classifier required only 3 min for training and still achieved 72.7% (70.8–74.6%) external-test accuracy. Attention map analysis revealed that the imaging features used by the no-code deep-learning models were similar to those used by endoscopists during visual inspection. Conclusion: No-code deep-learning tools allow for the rapid development of models with high accuracy for predicting colorectal polyp histology.

The Feasibility of Applying Artificial Intelligence to Gastrointestinal Endoscopy to Improve the Detection Rate of Early Gastric Cancer Screening

Convolutional neural networks in the field of artificial intelligence show great potential in image recognition. It assisted endoscopy to improve the detection rate of early gastric cancer. The 5-year survival rate for advanced gastric cancer is less than 30%, while the 5-year survival rate for early gastric cancer is more than 90%. Therefore, earlier screening for gastric cancer can lead to a better prognosis. However, the detection rate of early gastric cancer in China has been extremely low due to many factors, such as the presence of gastric cancer without obvious symptoms, difficulty identifying lesions by the naked eye, and a lack of experience among endoscopists. The introduction of artificial intelligence can help mitigate these shortcomings and greatly improve the accuracy of screening. According to relevant reports, the sensitivity and accuracy of artificial intelligence trained on deep cirrocumulus neural networks are better than those of endoscopists, and evaluations also take less time, which can greatly reduce the burden on endoscopists. In addition, artificial intelligence can also perform real-time detection and feedback on the inspection process of the endoscopist to standardize the operation of the endoscopist. AI has also shown great potential in training novice endoscopists. With the maturity of AI technology, AI has the ability to improve the detection rate of early gastric cancer in China and reduce the death rate of gastric cancer related diseases in China.

A Noninvasive Risk Stratification Tool Build Using an Artificial Intelligence Approach for Colorectal Polyps Based on Annual Checkup Data

Colorectal cancer is the leading cause of cancer-related deaths worldwide, and early detection has proven to be an effective method for reducing mortality. The machine learning method can be implemented to build a noninvasive stratifying tool that helps identify patients with potential colorectal precancerous lesions (polyps). This study aimed to develop a noninvasive risk-stratified tool for colorectal polyps in asymptomatic, healthy participants. A total of 20,129 consecutive asymptomatic patients who underwent a health checkup between January 2005 and August 2007 were recruited. Positive relationships between noninvasive risk factors, such as age, Helicobacter pylori infection, hypertension, gallbladder polyps/stone, and BMI and colorectal polyps were observed (p < 0.0001), regardless of sex, whereas significant findings were noted in men with tooth disease (p = 0.0053). A risk stratification tool was developed, for colorectal polyps, that considers annual checkup results from noninvasive examinations. For the noninvasive stratified tool, the area under the receiver operating characteristic curve (AUC) of obese females (males) aged <50 years was 91% (83%). In elderly patients (>50 years old), the AUCs of the stratifying tools were >85%. Our results indicate that the risk stratification tool can be built by using random forest and serve as an efficient noninvasive tool to identify patients requiring colonoscopy.

Time-of-day variation in the diagnostic quality of screening colonoscopies: a registry-based study

Background

The diagnostic quality of screening colonoscopies has been found to differ between morning and afternoon. Specifically, the adenoma detection rate (ADR) is higher in the morning. Our aim was to assess if time-of-day dependent differences in colonoscopy quality exist in a Danish screening setting. Following national screening guidelines, an individual will be exempt from screening invitations for 8 years if the colonoscopy is without pathology. Therefore, it is of utmost importance to identify factors systematically affecting the detection of lesions.

Methods

This was a single-center study of screening colonoscopies performed between 2014 and 2018. Records were retrieved from the Danish Colorectal Cancer Screening Database and coupled with local data. The ADR and the cecal intubation rate were compared between morning (8-12 a.m.) and afternoon (12-4 p.m.) colonoscopies. Multivariate logistic regression analysis was performed.

Results

A total of 3659 screening colonoscopies were included. The ADR was 51% in the morning and 58% in the afternoon. Multivariate analysis found this statistically significant, with the "afternoon vs. morning" odds ratio for adenoma detection being 1.4 (95% confidence interval 1.17-1.68; P<0.001). The cecal intubation rate was 95.6% in the morning and 94.7%, a non-significant difference.

Conclusions

The ADR of screening colonoscopies was higher in the afternoon. Our study highlights the need for local/regional evaluation of factors affecting colonoscopy quality to address such issues. A clean colonoscopy exempts the patient from subsequent screening invitations for 8 years. Therefore, any observed systematic differences in quality must be addressed and eliminated.

Automated detection of cecal intubation with variable bowel preparation using a deep convolutional neural network

Background and study aims  Colonoscopy completion reduces post-colonoscopy colorectal cancer. As a result, there have been attempts at implementing artificial intelligence to automate the detection of the appendiceal orifice (AO) for quality assurance. However, the utilization of these algorithms has not been demonstrated in suboptimal conditions, including variable bowel preparation. We present an automated computer-assisted method using a deep convolutional neural network to detect the AO irrespective of bowel preparation. Methods  A total of 13,222 images (6,663 AO and 1,322 non-AO) were extracted from 35 colonoscopy videos recorded between 2015 and 2018. The images were labelled with Boston Bowel Preparation Scale scores. A total of 11,900 images were used for training/validation and 1,322 for testing. We developed a convolutional neural network (CNN) with a DenseNet architecture pre-trained on ImageNet as a feature extractor on our data and trained a classifier uniquely tailored for identification of AO and non-AO images using binary cross entropy loss. Results  The deep convolutional neural network was able to correctly classify the AO and non-AO images with an accuracy of 94 %. The area under the receiver operating curve of this neural network was 0.98. The sensitivity, specificity, positive predictive value, and negative predictive value of the algorithm were 0.96, 0.92, 0.92 and 0.96, respectively. AO detection was > 95 % regardless of BBPS scores, while non-AO detection improved from BBPS 1 score (83.95 %) to BBPS 3 score (98.28 %). Conclusions  A deep convolutional neural network was created demonstrating excellent discrimination between AO from non-AO images despite variable bowel preparation. This algorithm will require further testing to ascertain its effectiveness in real-time colonoscopy.

Applications of Artificial Intelligence for the Diagnosis of Gastrointestinal Diseases

The development of convolutional neural networks has achieved impressive advances of machine learning in recent years, leading to an increasing use of artificial intelligence (AI) in the field of gastrointestinal (GI) diseases. AI networks have been trained to differentiate benign from malignant lesions, analyze endoscopic and radiological GI images, and assess histological diagnoses, obtaining excellent results and high overall diagnostic accuracy. Nevertheless, there data are lacking on side effects of AI in the gastroenterology field, and high-quality studies comparing the performance of AI networks to health care professionals are still limited. Thus, large, controlled trials in real-time clinical settings are warranted to assess the role of AI in daily clinical practice. This narrative review gives an overview of some of the most relevant potential applications of AI for gastrointestinal diseases, highlighting advantages and main limitations and providing considerations for future development.

The impact of reader fatigue on the accuracy of capsule endoscopy interpretation

BACKGROUND AND AIMS

Capsule endoscopy (CE) interpretation requires the review of many thousands of images, with lesions often limited to just a few frames. In this study we aim to determine whether lesion detection declines according to the number of capsule videos read.

METHODS

32 participants, 16 of which were novices (NR) and 16 experienced (ER) capsule readers took part in this prospective evaluation study. Participants read six capsule cases with a variety of lesions, in a randomly assigned order during a single sitting. Psychomotor Vigilance Tests and Fatigue Scores were recorded prior to commencing and then after every two capsules read. Changes in lesion detection and measures of fatigue were assessed across the duration of the study.

RESULTS

Mean agreement with the predefined lesions was 48.3% (SD:16.1), and 21.3% (SD:15.1) for the experienced and novice readers respectively. Lesion detection declined amongst experienced reader after the first study (p = 0.01), but remained stable after subsequent capsules read, while NR accuracy was unaffected by capsule numbers read. Objective measures of fatigue did not correlate with reading accuracy.

CONCLUSION

This study demonstrates that reader accuracy declines after reading just one capsule study. Subjective and objective measures of fatigue were not sufficient to predict the onset of the effects of fatigue.

Use of ambient lighting during colonoscopy and its effect on adenoma detection rate and eye fatigue: results of a pilot study

Background and study aims  Adenoma detection rate (ADR) appears to decrease as the number of consecutive hours performing procedures increases, and eye strain may be a contributing factor. Ambient light may improve symptoms of eye strain, but its effects have yet to be explored in the field of gastroenterology. We aim to determine if using ambient lighting during screening colonoscopy will maintain ADRs and improve eye strain symptoms compared with low lighting. Methods  At a single center, retrospective data were collected on colonoscopies performed under low lighting and compared to prospective data collected on colonoscopies with ambient lighting. Eye fatigue surveys were completed by gastroenterologists. Satisfaction surveys were completed by physicians and staff. Results  Of 498 low light and 611 ambient light cases, 172 and 220 adenomas were detected, respectively ( P  = 0.611). Under low lighting, the ADR decreased 5.6 % from first to last case of the day ( P  = 0.2658). With ambient lighting, the ADR increased by 2.80 % ( P  = 0.5445). The difference in the overall change in ADR between first and last cases with ambient light versus low light was statistically significant (8.40 % total unit change, P  = 0.01). The average eye strain scores were 8.12 with low light, and 5.63 with ambient light ( P  = 0.3341). Conclusions  Performing screening colonoscopies with ambient light may improve the differential change in ADR that occurs from the beginning to the end of the day. This improvement in ADR may be related to improvement in operator fatigue. The effect of ambient light on eye strain is unclear. Further investigation is warranted on the impact of ambient light on symptoms of eye strain and ADR.

Association of Intensive Endoscopic Screening Burden With Gastric Cancer Detection

IMPORTANCE

The rapidly increasing number of gastric cancer examinations performed over a short period might influence screening performance. Accessing the association between calendar month and gastric cancer detection rates might improve policy and guide institutional support.

OBJECTIVE

To evaluate the association between the increased number of examinations over a certain period and gastric cancer detection rates among a large population included in the Korean National Cancer Screening Program (KNCSP).

DESIGN, SETTING, AND PARTICIPANTS

This retrospective, population-based cohort study used data from the KNCSP comprising 26 765 665 men and women aged 40 years or older who participated in the screening program between January 1, 2013, and December 31, 2016. Data were analyzed from November 1, 2019, to March 31, 2020.

EXPOSURES

Gastric cancer screening with endoscopy.

MAIN OUTCOMES AND MEASURES

The primary outcome was monthly gastric cancer detection rates in the KNCSP. A negative binomial regression model was used to evaluate the association between the screening month and detection rates.

RESULTS

In total, 21 535 222 individuals underwent endoscopy (mean [SD] age, 55.61 [10.61] years; 11 761 709 women [54.62%]). The quarterly number of participants was the highest in the last quarter of the study period (2013-2014: 4 094 951 [41.39%], 2015-2016: 4 911 629 [42.19%]); this proportion was 2.48 to 2.84 times greater than that of the first quarter. Cancer detection rates were the lowest in December (2013-2014: 0.22; 95% CI, 0.22-0.23; 2015-2016: 0.21; 95% CI, 0.21-0.22); this was approximately a 40.0% to 45.0% reduction compared with the rates in January. The age group was the significant factor for monthly detection rates. After adjustment for the age group and taking account of the number of screenings, the estimated coefficient range for the screening month was negative and the detection rate in December was significantly different than in January for both the consequent cycles (2013-2014: -0.05 to -0.18; P < .001; and 2015-2016: -0.06 to -0.19; P < .001). In the multivariable logistic model, the association of calendar month with detected cancer remained after adjusting for other confounding factors (December, 2013-2014: odds ratio, 0.82; 95% CI, 0.76-0.87; P < .001; 2015-2016: odds ratio, 0.83; 95% CI, 0.79-0.89; P < .001).

CONCLUSIONS AND RELEVANCE

The findings of this cohort study suggest that the workload of endoscopists increased with the increasing number of examinations toward the end of the year, as demonstrated by the decreased cancer detection rates. These findings may help to improve gastric cancer detection rates of screening programs by controlling the monthly screening number and policy modifications.

Effect of successive endoscopic procedures in polyp and adenoma detection rates: Too early is not always too good

BACKGROUND AND AIMS: There is conflicting evidence regarding the impact of hypothetical cumulative fatigue after performing too many endoscopic procedures on both polyp and adenoma detection rates (PDR, and ADR, respectively). The aim of this study is to evaluate the effect of successive endoscopic procedures on PDR and ADR.

METHODS

A retrospective cross-sectional study was undertaken among consecutive patients on whom colonoscopy and/or esophagogastroduodenoscopy were performed between January 2012 and August 2014. Data regarding polyp and adenoma detection, cecal intubation, and bowel cleansing quality as well as demographical data of subjects were extracted. Endoscopic procedures were classified according to the time slots of the procedures throughout the endoscopy session in three groups: from the 1st to 4th endoscopy study (round 1), from the 5th to the 8th study (round 2), above the 9th study (round 3). We compared PDR and ADR among rounds.

RESULTS

Overall, 3388 patients were enrolled. Median age was 50 years (range 18-95) and 52.39% were female. There was a significant difference in terms of PDR among rounds (36.83%, 41.24%, and 43.38%, respectively, p = 0.007) and a non-significant numerical difference when ADR was compared (23.2%, 25.71%, and 26.78%, p = 0.07). On multivariate analysis, ADR was significantly associated with age (odds ratio [OR] 1.02 [1.01-1.03]), and male sex (OR 1.64 [1.38-1.94]).

CONCLUSION

Theoretical endoscopist's fatigue due to cumulative performance of endoscopies does not diminish colonoscopy quality. Both PDR and ADR seem to improve after endoscopist's cumulative rounds of performed endoscopies. This could be due to a "warm-up" effect.

Morning versus afternoon adenoma detection rate: a systematic review and meta-analysis

Adenoma detection rate (ADR) is a quality marker of colonoscopy and operator performance. Prior studies evaluating the effect of an extended workday on the ADR reported variable outcomes that remain controversial. Given the variable results of prior studies and the potential legal implications of reduced ADR in the afternoon, we aimed to further evaluate this parameter and its effect on ADR. We performed a systematic review of the PubMed, CINAHL and Scopus electronic databases. Studies were included if they reported ADR in patients undergoing colonoscopy in the morning session and the afternoon session. Afternoon sessions included both sessions following a morning shift and half-day block shifts. Subgroup analyses were performed for ADR comparing morning and afternoon colonoscopies in a continuous workday, advanced ADRs (AADRs) and polyp detection rates (PDRs) were also compared. Thirteen articles with 17 341 (61.2%) performed in the morning session and 10 994 (38.8%) performed in the afternoon session were included in this study. There was no statistical significance in the ADR or AADR between morning and afternoon sessions, respectively [relative risk (RR) 1.06, 95% confidence interval (CI) 0.99-1.14] and (RR 1.19, 95% CI 0.95-1.5). Afternoon procedures had a significantly higher PDR than morning procedures (RR 0.93, 95% CI 0.88-0.98). ADR was not significantly influenced in the afternoon session when operators continued to perform procedures throughout the day or on a half-day block schedule.

Automated endoscopic detection and classification of colorectal polyps using convolutional neural networks

BACKGROUND

Recently the American Society for Gastrointestinal Endoscopy addressed the 'resect and discard' strategy, determining that accurate in vivo differentiation of colorectal polyps (CP) is necessary. Previous studies have suggested a promising application of artificial intelligence (AI), using deep learning in object recognition. Therefore, we aimed to construct an AI system that can accurately detect and classify CP using stored still images during colonoscopy.

METHODS

We used a deep convolutional neural network (CNN) architecture called Single Shot MultiBox Detector. We trained the CNN using 16,418 images from 4752 CPs and 4013 images of normal colorectums, and subsequently validated the performance of the trained CNN in 7077 colonoscopy images, including 1172 CP images from 309 various types of CP. Diagnostic speed and yields for the detection and classification of CP were evaluated as a measure of performance of the trained CNN.

RESULTS

The processing time of the CNN was 20 ms per frame. The trained CNN detected 1246 CP with a sensitivity of 92% and a positive predictive value (PPV) of 86%. The sensitivity and PPV were 90% and 83%, respectively, for the white light images, and 97% and 98% for the narrow band images. Among the correctly detected polyps, 83% of the CP were accurately classified through images. Furthermore, 97% of adenomas were precisely identified under the white light imaging.

CONCLUSIONS

Our CNN showed promise in being able to detect and classify CP through endoscopic images, highlighting its high potential for future application as an AI-based CP diagnosis support system for colonoscopy.

Study on the influence of assistant experience on the quality of colonoscopy: A pilot single-center study

BACKGROUND AND OBJECTIVE

Colonoscopy is the most important method for the diagnosis and treatment of intestinal diseases, and there are many factors affecting the quality of examination. Although the assistant is one of the factors influencing the quality of colonoscopy, there are few studies on the effect of different assistants with different experiences on the quality of colonoscopy. Therefore, the study was aimed to research the correlation between different assistants with different experiences and the quality of water-injection colonoscopy.

METHOD

In this study, a single-center randomized controlled trial was conducted to analyze the key quality indicators (the rate to arrive cecum, time to arrive cecum, total operation time, detection rate of polyps, detection rate of adenoma, pain score, operation satisfaction, and the pressure on abdomen) of patients who underwent water-injection colonoscopy under non-sedation from January 2018 to June 2018 in the center. Patients were randomly assigned to different assistant groups based on the actual working period of 6 months (0∼6 months inexperienced assistant group and assistant group with more than 6 months of experience). Through fitting the bivariate and multivariate logistic regression models, the differences between the two groups and the effects on the key quality indicators of colon examination were analyzed.

RESULTS

A total of 331 patients who were eligible for non-sedation colonoscopy were randomly assigned to the experienced assistant group (n = 179) and the inexperienced assistant group (n = 152). Among them, 103 cases of polyp and 70 cases of adenoma were detected. The rate to arrive cecum, polyp detection rate and adenoma detection rate were compared between the two groups during operation (P > 0.05). However, there were significant differences in the time to arrive cecum, patients' satisfaction with operation, pain score and abdominal pressure (P < .05). In the inexperienced assistant group, 20% of the operation time was one standard deviation higher than the mean value, while the experienced assistant group was 12% (339 s vs 405s, OR 0.541, 95% 0.295-0.990). Compared with the inexperienced assistant group, patients in the experienced assistant group had higher operational satisfaction (98.32% vs 92.11%, OR 0.199, 95% 0.055-0.718) and lower pain score (0.3 vs 0.49, OR 1.993, 95% 1.52-3.775). All relations remained unchanged after adjusting for potential confounders.

CONCLUSION

The assistant is a key factor in the quality of colonoscopy, especially in the case of non-sedating colonoscopy. The experience of the assistant is related to the time to arrive cecum, the degree of pain and the overall satisfaction of patient with the operation. The assistant should be subject to the quality supervision of the endoscopic inspector. Proof of human Clinical Trial Registration: The institutional review board of Fifth Affiliated Hospital of Wenzhou Medical College, Zhejiang Province, China approved the study. The study is registered on. Chinese Clinical Trial Registry (ChiCTR1800015650).

Effect of time of day and specialty on polyp detection rates in Australia

BACKGROUND AND AIM

Adenoma detection rate (ADR) is an important quality metric in colonoscopy. However, there is conflicting evidence around factors that influence ADR. This study aims to investigate the effect of time of day and endoscopist background on ADR and sessile serrated adenoma/polyp detection rate (SSA/P-DR) for screening colonoscopies.

METHODS

Consecutive patients undergoing colonoscopy in 2016 were retrospectively evaluated. Primary outcome was the effect of time of day and endoscopist specialty on screening ADR. Secondary outcomes included evaluation of the same factors on SSA/P-DR and other metrics and collinearity of ADR and SSA/P-DR. Linear regression models were used for association between ADR, time of day, and endoscopist background. Bowel preparation, endoscopist, session, patient age, and gender were adjusted for. Linear regression model was also used for comparing ADR and SSA/P-DR. Chi-square was used for difference of proportions.

RESULTS

Two thousand six hundred fifty-seven colonoscopies, of which 558 were screening colonoscopies, were performed. The adjusted mean ADR (screening) was 36.8% in the morning compared with 30.5% in the afternoon (P < 0.0001) and was 36.8% for gastroenterologists compared with 30.4% for surgeons (P < 0.0001). For every 1-h delay in commencing the procedure, there was a reduction in mean ADR by 3.4%. Using a linear regression model, a statistically significant positive association was found between ADR and SSA/P-DR (P < 0.0001).

CONCLUSIONS

Morning and afternoon sessions and gastroenterologists and surgeons achieved the minimum standards recommended for ADR. Afternoon lists and surgeons were associated with a lower ADR compared with morning and gastroenterologists, respectively. Additionally, SSA/P-DR showed collinearity with ADR.

Spotting malignancies from gastric endoscopic images using deep learning

BACKGROUND

Gastric cancer is a common kind of malignancies, with yearly occurrences exceeding one million worldwide in 2017. Typically, ulcerous and cancerous tissues develop abnormal morphologies through courses of progression. Endoscopy is a routinely adopted means for examination of gastrointestinal tract for malignancy. Early and timely detection of malignancy closely correlate with good prognosis. Repeated presentation of similar frames from gastrointestinal tract endoscopy often weakens attention for practitioners to result in true patients missed out to incur higher medical cost and unnecessary morbidity. Highly needed is an automatic means for spotting visual abnormality and prompts for attention for medical staff for more thorough examination.

METHODS

We conduct classification of benign ulcer and cancer for gastrointestinal endoscopic color images using deep neural network and transfer-learning approach. Using clinical data gathered from Gil Hospital, we built a dataset comprised of 200 normal, 367 cancer, and 220 ulcer cases, and applied the inception, ResNet, and VGGNet models pretrained on ImageNet. Three classes were defined-normal, benign ulcer, and cancer, and three separate binary classifiers were built-those for normal vs cancer, normal vs ulcer, and cancer vs ulcer for the corresponding classification tasks. For each task, considering inherent randomness entailed in the deep learning process, we performed data partitioning and model building experiments 100 times and averaged the performance values.

RESULTS

Areas under curves of respective receiver operating characteristics were 0.95, 0.97, and 0.85 for the three classifiers. The ResNet showed the highest level of performance. The cases involving normal, i.e., normal vs ulcer and normal vs cancer resulted in accuracies above 90%. The case of ulcer vs cancer classification resulted in a lower accuracy of 77.1%, possibly due to smaller difference in appearance than those cases involving normal.

CONCLUSIONS

The overall level of performance of the proposed method was very promising to encourage applications in clinical environments. Automatic classification using deep learning technique as proposed can be used to complement manual inspection efforts for practitioners to minimize dangers of missed out positives resulting from repetitive sequence of endoscopic frames and weakening attentions.

Adenoma Detection Rate Falls at the End of the Day in a Large Multi-site Sample

BACKGROUND

There is concern that mental and physical fatigue among endoscopists over the course of the day will lead to lower adenoma detection rate (ADR). There are mixed findings in the prior literature on whether such an association exists.

AIMS

The aim of this study was to measure the association between the number of colonoscopies performed in a day and ADR and withdrawal time.

METHODS

We analyzed 86,624 colonoscopy and associated pathology reports between October 2013 and September 2015 from 131 physicians at two medical centers. A previously validated natural language processing program was used to abstract relevant data. We identified the order of colonoscopies performed in the physicians' schedule and calculated the ADR and withdrawal time for each colonoscopy position.

RESULTS

The ADR for our overall sample was 29.9 (CI 29.6-30.2). The ADR for colonoscopies performed at the 9th + position was significantly lower than those at the 1st-4th or 5th-8th position, 27.2 (CI 25.8-28.6) versus 29.9 (CI 29.5-30.3), 30.2 (CI 29.6-30.9), respectively. Withdrawal time steadily decreased by colonoscopy position going from 11.6 (CI 11.4-11.9) min for the 1st colonoscopy to 9.6 (8.9-10.3) min for the 9th colonoscopy.

CONCLUSION

In our study population, ADR and withdrawal time decrease by roughly 7 and 20%, respectively, by the end of the day. Our results imply that rather than mental or physical fatigue, lower ADR at the end of the day might be driven by endoscopists rushing.

Application of Convolutional Neural Networks in the Diagnosis of Helicobacter pylori Infection Based on Endoscopic Images

BACKGROUND AND AIMS

The role of artificial intelligence in the diagnosis of Helicobacter pylori gastritis based on endoscopic images has not been evaluated. We constructed a convolutional neural network (CNN), and evaluated its ability to diagnose H. pylori infection.

METHODS

A 22-layer, deep CNN was pre-trained and fine-tuned on a dataset of 32,208 images either positive or negative for H. pylori (first CNN). Another CNN was trained using images classified according to 8 anatomical locations (secondary CNN). A separate test data set (11,481 images from 397 patients) was evaluated by the CNN, and 23 endoscopists, independently.

RESULTS

The sensitivity, specificity, accuracy, and diagnostic time were 81.9%, 83.4%, 83.1%, and 198s, respectively, for the first CNN, and 88.9%, 87.4%, 87.7%, and 194s, respectively, for the secondary CNN. These values for the 23 endoscopists were 79.0%, 83.2%, 82.4%, and 230±65min (85.2%, 89.3%, 88.6%, and 253±92min by 6 board-certified endoscopists), respectively. The secondary CNN had a significantly higher accuracy than endoscopists (by 5.3%; 95% CI, 0.3-10.2).

CONCLUSION

H. pylori gastritis could be diagnosed based on endoscopic images using CNN with higher accuracy and in a considerably shorter time compared to manual diagnosis by endoscopists.

What Can We Do to Optimize Colonoscopy and How Effective Can We Be?

In the USA, colorectal cancer is the third most common cancer and third leading cause of cancer death among both men and women. Declining rates of colon cancer in the past decade have been attributed in part to screening and removal of precancerous polyps via colonoscopy. Recent emphasis has been placed on measures to increase the quality and effectiveness of colonoscopy. These have been divided into pre-procedure quality metrics (bowel preparation), procedural quality metrics (cecal intubation, withdrawal time, and adenoma detection rate), post-procedure metrics (surveillance interval), and other quality metrics (patient satisfaction and willingness to repeat the procedure). The purpose of this article is to review the data and controversies surrounding each of these and identify ways to optimize the performance of colonoscopy.