Artificial intelligence for non-polypoid colorectal neoplasms

A deep neural network improves endoscopic detection of laterally spreading tumors

BACKGROUND

Colorectal cancer (CRC) is the malignant tumor of the digestive system with the highest incidence and mortality rate worldwide. Laterally spreading tumors (LSTs) of the large intestine have unique morphological characteristics, special growth patterns and higher malignant potential. Therefore, LSTs are a precancerous lesion of CRC that could be easily missed.

OBJECTIVE

The purpose of this study was to establish an LSTs lesion detection algorithm based on the YOLOv7 model and to evaluate the detection performance of the algorithm on LSTs.

METHOD

A total of 7985 LSTs images and 93,197 non-LSTs images were included in this study, and the training set, validation set, and 80% of the data in the dataset is used for training, 10% for validation, and 10% for testing. In detail, a total of 6261 LSTs images and 74,798 non-LSTs images were used as the training set to train the LSTs lesion detection algorithm to identify LSTs. A total of 743 LSTs images and 9486 non-LSTs images were used as validation set to evaluate the learning ability of the LSTs lesion detection algorithm. A total of 981 LSTs images and 8913 non-LSTs images were used as test set to evaluate the generalization ability of the LSTs lesion detection algorithm. To evaluate the diagnostic ability of the LSTs lesion detection algorithm for LSTs, we selected 3636 images (562 LSTs, 3074 non-LSTs) images from the test set as the subtest set. Finally, we compared the performance of the AI algorithm with endoscopist in the diagnosis of LSTs.

RESULT

The accuracy of LSTs lesion detection algorithm in identifying LSTs is 99.34%, sensitivity is 96.88%, specificity is 99.8%, positive predictive value is 98.94%, and negative predictive value is 99.41%.

CONCLUSION

Our model based on the YOLOv7 achieved high diagnostic accuracy in LSTs lesion, significantly better than that of novice and senior doctors, and reaching the same level as expert endoscopists.

Accuracy of Computer-aided Diagnosis in Colonoscopy Varies According to Polyp Location: A Systematic Review and Meta-analysis

BACKGROUND & AIMS

Computer-aided diagnosis (CADx) assists endoscopists in differentiating between neoplastic and non-neoplastic polyps during colonoscopy. This study aimed to evaluate the impact of polyp location (proximal vs. distal colon) on the diagnostic performance of CADx for ≤5 mm polyps.

METHODS

We searched for studies evaluating the performance of real-time CADx alone (ie, independently of endoscopist judgement) for predicting the histology of colorectal polyps ≤5 mm. The primary endpoints were CADx sensitivity and specificity in the proximal and distal colon. Secondary outcomes were the negative predictive value (NPV), positive predictive value (PPV), and the accuracy of the CADx alone. Distal colon was limited to the rectum and sigmoid.

RESULTS

We included 11 studies for analysis with a total of 7782 polyps ≤5 mm. CADx specificity was significantly lower in the proximal colon compared with the distal colon (62% vs 85%; risk ratio (RR), 0.74; 95% confidence interval [CI], 0.72-0.84). Conversely, sensitivity was similar (89% vs 87%); RR, 1.00; 95% CI, 0.97-1.03). The NPV (64% vs 93%; RR, 0.71; 95% CI, 0.64-0.79) and accuracy (81% vs 86%; RR, 0.95; 95% CI, 0.91-0.99) were significantly lower in the proximal than distal colon, whereas PPV was higher in the proximal colon (87% vs 76%; RR, 1.11; 95% CI, 1.06-1.17).

CONCLUSION

The diagnostic performance of CADx for polyps in the proximal colon is inadequate, exhibiting significantly lower specificity compared with its performance for distal polyps. Although current CADx systems are suitable for use in the distal colon, they should not be employed for proximal polyps until more performant systems are developed specifically for these lesions.

Role of Artificial Intelligence in Colonoscopy Detection of Advanced Neoplasias : A Randomized Trial

BACKGROUND

The role of computer-aided detection in identifying advanced colorectal neoplasia is unknown.

OBJECTIVE

To evaluate the contribution of computer-aided detection to colonoscopic detection of advanced colorectal neoplasias as well as adenomas, serrated polyps, and nonpolypoid and right-sided lesions.

DESIGN

Multicenter, parallel, randomized controlled trial. (ClinicalTrials.gov: NCT04673136).

SETTING

Spanish colorectal cancer screening program.

PARTICIPANTS

3213 persons with a positive fecal immunochemical test.

INTERVENTION

Enrollees were randomly assigned to colonoscopy with or without computer-aided detection.

MEASUREMENTS

Advanced colorectal neoplasia was defined as advanced adenoma and/or advanced serrated polyp.

RESULTS

The 2 comparison groups showed no significant difference in advanced colorectal neoplasia detection rate (34.8% with intervention vs. 34.6% for controls; adjusted risk ratio [aRR], 1.01 [95% CI, 0.92 to 1.10]) or the mean number of advanced colorectal neoplasias detected per colonoscopy (0.54 [SD, 0.95] with intervention vs. 0.52 [SD, 0.95] for controls; adjusted rate ratio, 1.04 [99.9% CI, 0.88 to 1.22]). Adenoma detection rate also did not differ (64.2% with intervention vs. 62.0% for controls; aRR, 1.06 [99.9% CI, 0.91 to 1.23]). Computer-aided detection increased the mean number of nonpolypoid lesions (0.56 [SD, 1.25] vs. 0.47 [SD, 1.18] for controls; adjusted rate ratio, 1.19 [99.9% CI, 1.01 to 1.41]), proximal adenomas (0.94 [SD, 1.62] vs. 0.81 [SD, 1.52] for controls; adjusted rate ratio, 1.17 [99.9% CI, 1.03 to 1.33]), and lesions of 5 mm or smaller (polyps in general and adenomas and serrated lesions in particular) detected per colonoscopy.

LIMITATIONS

The high adenoma detection rate in the control group may limit the generalizability of the findings to endoscopists with low detection rates.

CONCLUSION

Computer-aided detection did not improve colonoscopic identification of advanced colorectal neoplasias.

PRIMARY FUNDING SOURCE

Medtronic.

Usefulness of a novel computer-aided detection system for colorectal neoplasia: a randomized controlled trial

BACKGROUND AND AIMS

Artificial intelligence-based computer-aid detection (CADe) devices have been recently tested in colonoscopies, increasing the adenoma detection rate (ADR), mainly in Asian populations. However, evidence for the benefit of these devices in the occidental population is still low. We tested a new CADe device, namely, ENDO-AID (OIP-1) (Olympus, Tokyo, Japan), in clinical practice.

METHODS

This randomized controlled trial included 370 consecutive patients who were randomized 1:1 to CADe (n = 185) versus standard exploration (n = 185) from November 2021 to January 2022. The primary endpoint was the ADR. Advanced adenoma was defined as ≥10 mm, harboring high-grade dysplasia, or with a villous pattern. Otherwise, the adenoma was nonadvanced. ADR was assessed in both groups stratified by endoscopist ADR and colon cleansing.

RESULTS

In the intention-to-treat analysis, the ADR was 55.1% (102/185) in the CADe group and 43.8% (81/185) in the control group (P = .029). Nonadvanced ADRs (54.8% vs 40.8%, P = .01) and flat ADRs (39.4 vs 24.8, P = .006), polyp detection rate (67.1% vs 51%; P = .004), and number of adenomas per colonoscopy were significantly higher in the CADe group than in the control group (median [25th-75th percentile], 1 [0-2] vs 0 [0-1.5], respectively; P = .014). No significant differences were found in serrated ADR. After stratification by endoscopist and bowel cleansing, no statistically significant differences in ADR were found.

CONCLUSIONS

Colonoscopy assisted by ENDO-AID (OIP-1) increases ADR and number of adenomas per colonoscopy, suggesting it may aid in the detection of colorectal neoplastic lesions, especially because of its detection of diminutive and flat adenomas. (Clinical trial registration number: NCT04945044.).

Efficacy of international web-based educational intervention in the detection of high-risk flat and depressed colorectal lesions higher (CATCH project) with a video: Randomized trial

OBJECTIVES

Three subcategories of high-risk flat and depressed lesions (FDLs), laterally spreading tumors non-granular type (LST-NG), depressed lesions, and large sessile serrated lesions (SSLs), are highly attributable to post-colonoscopy colorectal cancer (CRC). Efficient and organized educational programs on detecting high-risk FDLs are lacking. We aimed to explore whether a web-based educational intervention with training on FIND clues (fold deformation, intensive stool/mucus attachment, no vessel visibility, and demarcated reddish area) may improve the ability to detect high-risk FDLs.

METHODS

This was an international web-based randomized control trial that enrolled non-expert endoscopists in 13 Asian countries. The participants were randomized into either education or non-education group. All participants took the pre-test and post-test to read 60 endoscopic images (40 high-risk FDLs, five polypoid, 15 no lesions) and answered whether there was a lesion. Only the education group received a self-education program (video and training questions and answers) between the tests. The primary outcome was a detection rate of high-risk FDLs.

RESULTS

In total, 284 participants were randomized. After excluding non-responders, the final data analyses were based on 139 participants in the education group and 130 in the non-education group. The detection rate of high-risk FDLs in the education group significantly improved by 14.7% (66.6-81.3%) compared with -0.8% (70.8-70.0%) in the non-education group. Similarly, the detection rate of LST-NG, depressed lesions, and large SSLs significantly increased only in the education group by 12.7%, 12.0%, and 21.6%, respectively.

CONCLUSION

Short self-education focusing on detecting high-risk FDLs was effective for Asian non-expert endoscopists. (UMIN000042348).

Performance of artificial intelligence for detection of subtle and advanced colorectal neoplasia

OBJECTIVES

There is uncertainty regarding the efficacy of artificial intelligence (AI) software to detect advanced subtle neoplasia, particularly flat lesions and sessile serrated lesions (SSLs), due to low prevalence in testing datasets and prospective trials. This has been highlighted as a top research priority for the field.

METHODS

An AI algorithm was evaluated on four video test datasets containing 173 polyps (35,114 polyp-positive frames and 634,988 polyp-negative frames) specifically enriched with flat lesions and SSLs, including a challenging dataset containing subtle advanced neoplasia. The challenging dataset was also evaluated by eight endoscopists (four independent, four trainees, according to the Joint Advisory Group on gastrointestinal endoscopy [JAG] standards in the UK).

RESULTS

In the first two video datasets, the algorithm achieved per-polyp sensitivities of 100% and 98.9%. Per-frame sensitivities were 84.1% and 85.2%. In the subtle dataset, the algorithm detected a significantly higher number of polyps (P < 0.0001), compared to JAG-independent and trainee endoscopists, achieving per-polyp sensitivities of 79.5%, 37.2% and 11.5%, respectively. Furthermore, when considering subtle polyps detected by both the algorithm and at least one endoscopist, the AI detected polyps significantly faster on average.

CONCLUSIONS

The AI based algorithm achieved high per-polyp sensitivities for advanced colorectal neoplasia, including flat lesions and SSLs, outperforming both JAG independent and trainees on a very challenging dataset containing subtle lesions that could have been overlooked easily and contribute to interval colorectal cancer. Further prospective trials should evaluate AI to detect subtle advanced neoplasia in higher risk populations for colorectal cancer.

Colonoscopic image synthesis with generative adversarial network for enhanced detection of sessile serrated lesions using convolutional neural network

Computer-aided detection (CADe) systems have been actively researched for polyp detection in colonoscopy. To be an effective system, it is important to detect additional polyps that may be easily missed by endoscopists. Sessile serrated lesions (SSLs) are a precursor to colorectal cancer with a relatively higher miss rate, owing to their flat and subtle morphology. Colonoscopy CADe systems could help endoscopists; however, the current systems exhibit a very low performance for detecting SSLs. We propose a polyp detection system that reflects the morphological characteristics of SSLs to detect unrecognized or easily missed polyps. To develop a well-trained system with imbalanced polyp data, a generative adversarial network (GAN) was used to synthesize high-resolution whole endoscopic images, including SSL. Quantitative and qualitative evaluations on GAN-synthesized images ensure that synthetic images are realistic and include SSL endoscopic features. Moreover, traditional augmentation methods were used to compare the efficacy of the GAN augmentation method. The CADe system augmented with GAN synthesized images showed a 17.5% improvement in sensitivity on SSLs. Consequently, we verified the potential of the GAN to synthesize high-resolution images with endoscopic features and the proposed system was found to be effective in detecting easily missed polyps during a colonoscopy.

Serrated polyp detection rate in colonoscopies performed by gastrointestinal fellows

Background:

Clinically significant serrated polyp detection rate (CSSDR) and proximal serrated polyp detection rate (PSDR) have been suggested as the potential quality benchmarks for colonoscopy (CSSDR = 7% and PSDR = 11%) in comparison to the established benchmark adenoma detection rate (ADR). Another emerging milestone is the detection rate of lateral spreading lesions (LSLs).

Objectives:

This study aimed to evaluate CSSDR, PSDR, ADR, and LSL detection rates among gastrointestinal (GI) fellows performing a colonoscopy. A secondary aim was to evaluate patient factors associated with the detection rates of these lesions.

Design and Methods:

A retrospective review of 799 colonoscopy reports was performed. GI fellow details, demographic data, and pathology found on colonoscopy were collected. Multiple logistic regression analysis was performed to identify the factors associated with CSSDR, PSDR, ADR, and LSL detection rates. A p value < 0.05 was considered statistically significant.

Results:

For our patient population, the median age was 58 years; 396 (49.8%) were male and 386 (48.6%) were African American. The 15 GI fellows ranged from first (F1), second (F2), or third (F3) year of training. We found an overall CSSDR of 4.4%, PSDR of 10.5%, ADR of 42.1%, and LSL detection rate of 3.2%. Female gender was associated with CSSDR, while only age was associated with PSDR. GI fellow level of training was associated with LSL detection rate, with the odds of detecting them expected to be four times higher in F2/F3s than F1s.

Conclusion:

Although GI fellows demonstrated an above-recommended ADR and nearly reached target PSDR, they failed to achieve target CSSDR. Future studies investigating a benchmark for LSL detection rate are needed to quantify if GI fellows are detecting these lesions at adequate rates.

Establishing key research questions for the implementation of artificial intelligence in colonoscopy: a modified Delphi method

BACKGROUND : Artificial intelligence (AI) research in colonoscopy is progressing rapidly but widespread clinical implementation is not yet a reality. We aimed to identify the top implementation research priorities. METHODS : An established modified Delphi approach for research priority setting was used. Fifteen international experts, including endoscopists and translational computer scientists/engineers, from nine countries participated in an online survey over 9 months. Questions related to AI implementation in colonoscopy were generated as a long-list in the first round, and then scored in two subsequent rounds to identify the top 10 research questions. RESULTS : The top 10 ranked questions were categorized into five themes. Theme 1: clinical trial design/end points (4 questions), related to optimum trial designs for polyp detection and characterization, determining the optimal end points for evaluation of AI, and demonstrating impact on interval cancer rates. Theme 2: technological developments (3 questions), including improving detection of more challenging and advanced lesions, reduction of false-positive rates, and minimizing latency. Theme 3: clinical adoption/integration (1 question), concerning the effective combination of detection and characterization into one workflow. Theme 4: data access/annotation (1 question), concerning more efficient or automated data annotation methods to reduce the burden on human experts. Theme 5: regulatory approval (1 question), related to making regulatory approval processes more efficient. CONCLUSIONS : This is the first reported international research priority setting exercise for AI in colonoscopy. The study findings should be used as a framework to guide future research with key stakeholders to accelerate the clinical implementation of AI in endoscopy.

Artificial intelligence-aided colonoscopy: Recent developments and future perspectives

Artificial intelligence (AI) systems, especially after the successful application of Convolutional Neural Networks, are revolutionizing modern medicine. Gastrointestinal Endoscopy has shown to be a fertile terrain for the development of AI systems aiming to aid endoscopists in various aspects of their daily activity. Lesion detection can be one of the two main aspects in which AI can increase diagnostic yield and abilities of endoscopists. In colonoscopy, it is well known that a substantial rate of missed neoplasia is still present, representing the major cause of interval cancer. In addition, an extremely high variability in adenoma detection rate, the main key quality indicator in colonoscopy, has been extensively reported. The other domain in which AI is believed to have a considerable impact on everyday clinical practice is lesion characterization and aid in “optical diagnosis”. By predicting in vivo histology, such pathology costs may be averted by the implementation of two separate but synergistic strategies, namely the “leave-in-situ” strategy for < 5 mm hyperplastic lesions in the rectosigmoid tract, and “resect and discard” for the other diminutive lesions. In this opinion review we present current available evidence regarding the role of AI in improving lesions’ detection and characterization during colonoscopy.