Computer-Aided Diagnosis for Leaving Colorectal Polyps In Situ : A Systematic Review and Meta-analysis

BACKGROUND

Computer-aided diagnosis (CADx) allows prediction of polyp histology during colonoscopy, which may reduce unnecessary removal of nonneoplastic polyps. However, the potential benefits and harms of CADx are still unclear.

PURPOSE

To quantify the benefit and harm of using CADx in colonoscopy for the optical diagnosis of small (≤5-mm) rectosigmoid polyps.

DATA SOURCES

Medline, Embase, and Scopus were searched for articles published before 22 December 2023.

STUDY SELECTION

Histologically verified diagnostic accuracy studies that evaluated the real-time performance of physicians in predicting neoplastic change of small rectosigmoid polyps without or with CADx assistance during colonoscopy.

DATA EXTRACTION

The clinical benefit and harm were estimated on the basis of accuracy values of the endoscopist before and after CADx assistance. The certainty of evidence was assessed using the GRADE (Grading of Recommendations Assessment, Development and Evaluation) framework. The outcome measure for benefit was the proportion of polyps predicted to be nonneoplastic that would avoid removal with the use of CADx. The outcome measure for harm was the proportion of neoplastic polyps that would be not resected and left in situ due to an incorrect diagnosis with the use of CADx. Histology served as the reference standard for both outcomes.

DATA SYNTHESIS

Ten studies, including 3620 patients with 4103 small rectosigmoid polyps, were analyzed. The studies that assessed the performance of CADx alone (9 studies; 3237 polyps) showed a sensitivity of 87.3% (95% CI, 79.2% to 92.5%) and specificity of 88.9% (CI, 81.7% to 93.5%) in predicting neoplastic change. In the studies that compared histology prediction performance before versus after CADx assistance (4 studies; 2503 polyps), there was no difference in the proportion of polyps predicted to be nonneoplastic that would avoid removal (55.4% vs. 58.4%; risk ratio [RR], 1.06 [CI, 0.96 to 1.17]; moderate-certainty evidence) or in the proportion of neoplastic polyps that would be erroneously left in situ (8.2% vs. 7.5%; RR, 0.95 [CI, 0.69 to 1.33]; moderate-certainty evidence).

LIMITATION

The application of optical diagnosis was only simulated, potentially altering the decision-making process of the operator.

CONCLUSION

Computer-aided diagnosis provided no incremental benefit or harm in the management of small rectosigmoid polyps during colonoscopy.

PRIMARY FUNDING SOURCE

European Commission. (PROSPERO: CRD42023402197).

Computer-aided diagnosis for optical diagnosis of diminutive colorectal polyps including sessile serrated lesions: a real-time comparison with screening endoscopists

BACKGROUND : We aimed to compare the accuracy of the optical diagnosis of diminutive colorectal polyps, including sessile serrated lesions (SSLs), between a computer-aided diagnosis (CADx) system and endoscopists during real-time colonoscopy. METHODS : We developed the POLyp Artificial Recognition (POLAR) system, which was capable of performing real-time characterization of diminutive colorectal polyps. For pretraining, the Microsoft-COCO dataset with over 300 000 nonpolyp object images was used. For training, eight hospitals prospectively collected 2637 annotated images from 1339 polyps (i. e. publicly available online POLAR database). For clinical validation, POLAR was tested during colonoscopy in patients with a positive fecal immunochemical test (FIT), and compared with the performance of 20 endoscopists from eight hospitals. Endoscopists were blinded to the POLAR output. Primary outcome was the comparison of accuracy of the optical diagnosis of diminutive colorectal polyps between POLAR and endoscopists (neoplastic [adenomas and SSLs] versus non-neoplastic [hyperplastic polyps]). Histopathology served as the reference standard. RESULTS : During clinical validation, 423 diminutive polyps detected in 194 FIT-positive individuals were included for analysis (300 adenomas, 41 SSLs, 82 hyperplastic polyps). POLAR distinguished neoplastic from non-neoplastic lesions with 79 % accuracy, 89 % sensitivity, and 38 % specificity. The endoscopists achieved 83 % accuracy, 92 % sensitivity, and 44 % specificity. The optical diagnosis accuracy between POLAR and endoscopists was not significantly different (P = 0.10). The proportion of polyps in which POLAR was able to provide an optical diagnosis was 98 % (i. e. success rate). CONCLUSIONS : We developed a CADx system that differentiated neoplastic from non-neoplastic diminutive polyps during endoscopy, with an accuracy comparable to that of screening endoscopists and near-perfect success rate.

Delphi Initiative for Early-Onset Colorectal Cancer (DIRECt) International Management Guidelines

BACKGROUND & AIMS

Patients with early-onset colorectal cancer (eoCRC) are managed according to guidelines that are not age-specific. A multidisciplinary international group (DIRECt), composed of 69 experts, was convened to develop the first evidence-based consensus recommendations for eoCRC.

METHODS

After reviewing the published literature, a Delphi methodology was used to draft and respond to clinically relevant questions. Each statement underwent 3 rounds of voting and reached a consensus level of agreement of ≥80%.

RESULTS

The DIRECt group produced 31 statements in 7 areas of interest: diagnosis, risk factors, genetics, pathology-oncology, endoscopy, therapy, and supportive care. There was strong consensus that all individuals younger than 50 should undergo CRC risk stratification and prompt symptom assessment. All newly diagnosed eoCRC patients should receive germline genetic testing, ideally before surgery. On the basis of current evidence, endoscopic, surgical, and oncologic treatment of eoCRC should not differ from later-onset CRC, except for individuals with pathogenic or likely pathogenic germline variants. The evidence on chemotherapy is not sufficient to recommend changes to established therapeutic protocols. Fertility preservation and sexual health are important to address in eoCRC survivors. The DIRECt group highlighted areas with knowledge gaps that should be prioritized in future research efforts, including age at first screening for the general population, use of fecal immunochemical tests, chemotherapy, endoscopic therapy, and post-treatment surveillance for eoCRC patients.

CONCLUSIONS

The DIRECt group produced the first consensus recommendations on eoCRC. All statements should be considered together with the accompanying comments and literature reviews. We highlighted areas where research should be prioritized. These guidelines represent a useful tool for clinicians caring for patients with eoCRC.

Comprehensive review of publicly available colonoscopic imaging databases for artificial intelligence research: availability, accessibility, and usability

BACKGROUND AND AIMS

Publicly available databases containing colonoscopic imaging data are valuable resources for artificial intelligence (AI) research. Currently, little is known regarding the available number and content of these databases. This review aimed to describe the availability, accessibility, and usability of publicly available colonoscopic imaging databases, focusing on polyp detection, polyp characterization, and quality of colonoscopy.

METHODS

A systematic literature search was performed in MEDLINE and Embase to identify AI studies describing publicly available colonoscopic imaging databases published after 2010. Second, a targeted search using Google's Dataset Search, Google Search, GitHub, and Figshare was done to identify databases directly. Databases were included if they contained data about polyp detection, polyp characterization, or quality of colonoscopy. To assess accessibility of databases, the following categories were defined: open access, open access with barriers, and regulated access. To assess the potential usability of the included databases, essential details of each database were extracted using a checklist derived from the Checklist for Artificial Intelligence in Medical Imaging.

RESULTS

We identified 22 databases with open access, 3 databases with open access with barriers, and 15 databases with regulated access. The 22 open access databases contained 19,463 images and 952 videos. Nineteen of these databases focused on polyp detection, localization, and/or segmentation; 6 on polyp characterization, and 3 on quality of colonoscopy. Only half of these databases have been used by other researcher to develop, train, or benchmark their AI system. Although technical details were in general well reported, important details such as polyp and patient demographics and the annotation process were under-reported in almost all databases.

CONCLUSIONS

This review provides greater insight on public availability of colonoscopic imaging databases for AI research. Incomplete reporting of important details limits the ability of researchers to assess the usability of current databases.

The effect of the endoscopic device Endocuff™/Endocuff vision™ on quality standards in colonoscopy: A systematic review and meta-analysis of randomized trials

AIM

The adenoma detection rate (ADR) is an important quality measure, with a high ADR reflecting high-quality colonoscopy. This systematic review and meta-analysis aimed to assess the effects of Endocuff™/Endocuff Vision™-assisted colonoscopy (EAC) versus standard colonoscopy (SC) on ADR and other clinical, patient and resource-use outcomes.

METHOD

MEDLINE, EMBASE, Web of Science, Scopus and the Cochrane Central Register of Controlled Trials were searched for full papers reporting randomized studies comparing EAC with SC. The primary outcome was ADR. Secondary outcomes comprised key polyp/adenoma detection, procedure-related, patient-related and health economic measures. Random effects meta-analyses provided pooled estimates of outcomes [risk ratio (RR) or mean difference (MD), with 95% confidence intervals (CI)].

RESULTS

Twelve parallel-group randomized controlled trials (RCTs) and three crossover RCTs with data on 9140 patients were included. EAC significantly increased the ADR (RR 1.18, 95% CI 1.09-1.29), mean adenomas per procedure (MAP) (MD 0.19, 95% CI 0.06-0.33), polyp detection rate (PDR) (RR 1.20, 95% CI 1.10-1.30) and mean polyps per procedure (MPP) (MD 0.39, 95% CI 0.14-0.63) versus SC. EAC significantly increased segmental PDR versus SC in the sigmoid (RR 2.02, 95% CI 1.64-2.49), transverse (RR 1.63, 95% CI 1.09-2.42), ascending (RR 1.74, 95% CI 1.26-2.41) and caecal segments (RR 1.91, 95% CI 1.29-2.82). Procedure-related variables did not differ between arms. There were insufficient data for meta-analysis of health economic or patient-centred outcomes.

CONCLUSIONS

EAC increased ADR, MAP, PDR and MPP versus SC without detrimental effects on procedure measures. Cost-effectiveness and patient experience data are lacking and would be valuable to inform practice recommendations.

Computer-aided classification of colorectal segments during colonoscopy: a deep learning approach based on images of a magnetic endoscopic positioning device

OBJECTIVE

Assessment of the anatomical colorectal segment of polyps during colonoscopy is important for treatment and follow-up strategies, but is largely operator dependent. This feasibility study aimed to assess whether, using images of a magnetic endoscope imaging (MEI) positioning device, a deep learning approach can be useful to objectively divide the colorectum into anatomical segments.

METHODS

Models based on the VGG-16 based convolutional neural network architecture were developed to classify the colorectum into anatomical segments. These models were pre-trained on ImageNet data and further trained using prospectively collected data of the POLAR study in which endoscopists were using MEI (3930 still images and 90,151 video frames). Five-fold cross validation with multiple runs was used to evaluate the overall diagnostic accuracies of the models for colorectal segment classification (divided into a 5-class and 2-class colorectal segment division). The colorectal segment assignment by endoscopists was used as the reference standard.

RESULTS

For the 5-class colorectal segment division, the best performing model correctly classified the colorectal segment in 753 of the 1196 polyps, corresponding to an overall accuracy of 63%, sensitivity of 63%, specificity of 89% and kappa of 0.47. For the 2-class colorectal segment division, 1112 of the 1196 polyps were correctly classified, corresponding to an accuracy of 93%, sensitivity of 93%, specificity of 90% and kappa of 0.82.

CONCLUSION

The diagnostic performance of a deep learning approach for colorectal segment classification based on images of a MEI device is yet suboptimal (clinicaltrials.gov: NCT03822390).

Linked Colour imaging for the detection of polyps in patients with Lynch syndrome: a multicentre, parallel randomised controlled trial

OBJECTIVE

Despite regular colonoscopy surveillance, colorectal cancers still occur in patients with Lynch syndrome. Thus, detection of all relevant precancerous lesions remains very important. The present study investigates Linked Colour imaging (LCI), an image-enhancing technique, as compared with high-definition white light endoscopy (HD-WLE) for the detection of polyps in this patient group.

DESIGN

This prospective, randomised controlled trial was performed by 22 experienced endoscopists from eight centres in six countries. Consecutive Lynch syndrome patients ≥18 years undergoing surveillance colonoscopy were randomised (1:1) and stratified by centre for inspection with either LCI or HD-WLE. Primary outcome was the polyp detection rate (PDR).

RESULTS

Between January 2018 and March 2020, 357 patients were randomised and 332 patients analysed (160 LCI, 172 HD-WLE; 6 excluded due to incomplete colonoscopies and 19 due to insufficient bowel cleanliness). No significant difference was observed in PDR with LCI (44.4%; 95% CI 36.5% to 52.4%) compared with HD-WLE (36.0%; 95% CI 28.9% to 43.7%) (p=0.12). Of the secondary outcome parameters, more adenomas were found on a patient (adenoma detection rate 36.3%; vs 25.6%; p=0.04) and a colonoscopy basis (mean adenomas per colonoscopy 0.65 vs 0.42; p=0.04). The median withdrawal time was not statistically different between LCI and HD-WLE (12 vs 11 min; p=0.16).

CONCLUSION

LCI did not improve the PDR compared with HD-WLE in patients with Lynch syndrome undergoing surveillance. The relevance of findings more adenomas by LCI has to be examined further.

TRIAL REGISTRATION NUMBER

NCT03344289.

Definition of competence standards for optical diagnosis of diminutive colorectal polyps: European Society of Gastrointestinal Endoscopy (ESGE) Position Statement

BACKGROUND : The European Society of Gastrointestinal Endoscopy (ESGE) has developed a core curriculum for high quality optical diagnosis training for practice across Europe. The development of easy-to-measure competence standards for optical diagnosis can optimize clinical decision-making in endoscopy. This manuscript represents an official Position Statement of the ESGE aiming to define simple, safe, and easy-to-measure competence standards for endoscopists and artificial intelligence systems performing optical diagnosis of diminutive colorectal polyps (1 - 5 mm). METHODS : A panel of European experts in optical diagnosis participated in a modified Delphi process to reach consensus on Simple Optical Diagnosis Accuracy (SODA) competence standards for implementation of the optical diagnosis strategy for diminutive colorectal polyps. In order to assess the clinical benefits and harms of implementing optical diagnosis with different competence standards, a systematic literature search was performed. This was complemented with the results from a recently performed simulation study that provides guidance for setting alternative competence standards for optical diagnosis. Proposed competence standards were based on literature search and simulation study results. Competence standards were accepted if at least 80 % agreement was reached after a maximum of three voting rounds. RECOMMENDATION 1:  In order to implement the leave-in-situ strategy for diminutive colorectal lesions (1-5 mm), it is clinically acceptable if, during real-time colonoscopy, at least 90 % sensitivity and 80 % specificity is achieved for high confidence endoscopic characterization of colorectal neoplasia of 1-5 mm in the rectosigmoid. Histopathology is used as the gold standard.Level of agreement 95 %. RECOMMENDATION 2:  In order to implement the resect-and-discard strategy for diminutive colorectal lesions (1-5 mm), it is clinically acceptable if, during real-time colonoscopy, at least 80 % sensitivity and 80 % specificity is achieved for high confidence endoscopic characterization of colorectal neoplasia of 1-5 mm. Histopathology is used as the gold standard.Level of agreement 100 %. CONCLUSION : The developed SODA competence standards define diagnostic performance thresholds in relation to clinical consequences, for training and for use when auditing the optical diagnosis of diminutive colorectal polyps.

Guidance for setting easy-to-adopt competence criteria for optical diagnosis of diminutive colorectal polyps: a simulation approach

BACKGROUND AND AIMS

One reason the optical diagnosis strategy for diminutive colorectal polyps has not yet been implemented is that the current competence criteria (Preservation and Incorporation of Valuable Endoscopic Innovation [PIVI] initiative) are difficult to use in daily practice. To provide guidance for setting alternative easy-to-adopt competence criteria, we determined the lowest proportion of diminutive polyps that should have a correct optical diagnosis to meet the PIVI.

METHODS

For this simulation study, we used datasets from 2 prospectively collected cohorts of patients who underwent colonoscopy in either a primary colonoscopy or fecal immunochemical test (FIT) screening setting. In the simulation approach, virtual endoscopists or computer-aided diagnosis systems performed optical diagnosis of diminutive polyps with a fixed diagnostic performance level (strategy) on all individuals in the cohort who had ≥1 diminutive polyp. Strategies were defined by systematically varying the proportion of correct optical diagnoses for each polyp subtype (ie, adenomas, hyperplastic polyps, sessile serrated lesions). For each strategy, we determined whether PIVI-1 (≥90% agreement with U.S. or European Society for Gastrointestinal Endoscopy [ESGE] surveillance guidelines) and PIVI-2 (≥90% negative predictive value [NPV] for neoplastic lesions in the rectosigmoid) were met using Monte Carlo sampling with 1000 repetitions, with histology as reference.

RESULTS

The level of overall diagnostic accuracy to achieve the PIVI differed significantly depending on the clinical setting and guidelines used. In the colonoscopy screening setting, all diagnostic strategies in which 92% of all diminutive polyps (regardless of histology) were diagnosed correctly led to 90% or more agreement with U.S. surveillance intervals (ie, PIVI-1). For all diagnostic strategies in which ≥89% of all diminutive polyps were correctly diagnosed, at least 90% NPV was achieved (ie, PIVI-2). For the FIT screening setting, values were respectively ≥77% and ≥94%. When using ESGE guidelines, PIVI-1 was in both settings already met when 40% of all diminutive polyps were diagnosed correctly.

CONCLUSIONS

In contrast to the fixed PIVI criteria, our simulation study shows that different thresholds for the proportion of correctly diagnosed diminutive polyps lead to different clinical consequences depending on guidelines and clinical setting. However, this target proportion of diminutive colorectal polyps correctly diagnosed with optical diagnosis represents easier-to-adopt competence criteria.

[Electronic chromoendoscopy: optical diagnosis strategy for colorectal polyps not yet implemented]

During colonoscopy, all detected polyps are resected and sent for histopathological assessment by the pathologist to determine the interval for a surveillance colonoscopy. Diminutive polyps (1-5 mm), which constitute up to 60% of all polyps, are rarely malignant (0-0.1%). If it would be possible to predict the histology of these diminutive polyps during colonoscopy, histopathological examination could be omitted. This 'optical diagnosis strategy' by endoscopists could lead to significant cost savings. For safe implementation of this strategy in daily practice, its accuracy must meet minimum thresholds. The development of 'electronic chromoendoscopy' has led to more accurate optical diagnosis. However, research shows that some endoscopists are yet not able to meet the required accuracy thresholds. Due to this variability in accuracy and the absence of an accreditation system implementation of the optical diagnosis strategy in daily practice is still unsuccessful.

Curriculum for optical diagnosis training in Europe: European Society of Gastrointestinal Endoscopy (ESGE) Position Statement

This manuscript represents an official Position Statement of the European Society of Gastrointestinal Endoscopy (ESGE) aiming to guide general gastroenterologists to develop and maintain skills in optical diagnosis during endoscopy. In general, this requires additional training beyond the core curriculum currently provided in each country. In this context, ESGE have developed a European core curriculum for optical diagnosis practice across Europe for high quality optical diagnosis training. 1:  ESGE suggests that every endoscopist should have achieved general competence in upper and/or lower gastrointestinal (UGI/LGI) endoscopy before commencing training in optical diagnosis of the UGI/LGI tract, meaning personal experience of at least 300 UGI and/or 300 LGI endoscopies and meeting the ESGE quality measures for UGI/LGI endoscopy. ESGE suggests that every endoscopist should be able and competent to perform UGI/LGI endoscopy with high definition white light combined with virtual and/or dye-based chromoendoscopy before commencing training in optical diagnosis. 2:  ESGE suggests competency in optical diagnosis can be learned by attending a validated optical diagnosis training course based on a validated classification, and self-learning with a minimum number of lesions. If no validated training course is available, optical diagnosis can only be learned by attending a non-validated onsite training course and self-learning with a minimum number of lesions. 3:  ESGE suggests endoscopists are competent in optical diagnosis after meeting the pre-adoption and learning criteria, and meeting competence thresholds by assessing a minimum number of lesions prospectively during real-time endoscopy. ESGE suggests ongoing in vivo practice by endoscopists to maintain competence in optical diagnosis. If a competent endoscopist does not perform in vivo optical diagnosis on a regular basis, ESGE suggests repeating the learning and competence phases to maintain competence.Key areas of interest were optical diagnosis training in Barrett's esophagus, esophageal squamous cell carcinoma, early gastric cancer, diminutive colorectal lesions, early colorectal cancer, and neoplasia in inflammatory bowel disease. Condition-specific recommendations are provided in the main document.