Accuracy of artificial intelligence on histology prediction and detection of colorectal polyps: a systematic review and meta-analysis

doi:10.1016/j.gie.2020.02.033

Gastrointestinal Endoscopy

Volume 92, Issue 1, July 2020, Pages 11-22.e6

https://doi.org/10.1016/j.gie.2020.02.033 Get rights and content

Background and Aims

We performed a meta-analysis of all published studies to determine the diagnostic accuracy of artificial intelligence (AI) on histology prediction and detection of colorectal polyps.

Method

We searched Embase, PubMed, Medline, Web of Science, and Cochrane library databases to identify studies using AI for colorectal polyp histology prediction and detection. The quality of included studies was measured by the Quality Assessment of Diagnostic Accuracy Studies tool. We used a bivariate meta-analysis following a random-effects model to summarize the data and plotted hierarchical summary receiver operating characteristic curves. The area under the hierarchical summary receiver operating characteristic curve (AUC) served as an indicator of the diagnostic accuracy and during head-to-head comparisons.

Results

A total of 7680 images of colorectal polyps from 18 studies were included in the analysis of histology prediction. The accuracy of the AI (AUC) was .96 (95% confidence interval [CI], .95-.98), with a corresponding pooled sensitivity of 92.3% (95% CI, 88.8%-94.9%) and specificity of 89.8% (95% CI, 85.3%-93.0%). The AUC of AI using narrow-band imaging (NBI) was significantly higher than the AUC using non-NBI (.98 vs .84, P < .01). The performance of AI was superior to nonexpert endoscopists (.97 vs .90, P < .01). For characterization of diminutive polyps using a deep learning model with nonmagnifying NBI, the pooled negative predictive value was 95.1% (95% CI, 87.7%-98.1%). For polyp detection, the pooled AUC was .90 (95% CI, .67-1.00) with a sensitivity of 95.0% (95% CI, 91.0%-97.0%) and a specificity of 88.0% (95% CI, 58.0%-99.0%).

Conclusions

AI was accurate in histology prediction and detection of colorectal polyps, including diminutive polyps. The performance of AI was better under NBI and was superior to nonexpert endoscopists. Despite the difference in AI models and study designs, AI performances are rather consistent, which could serve as a reference for future AI studies.

Section snippets

Definition

In this study, colorectal polyp was defined as any endoscopic lesion that had been removed and examined histologically. We defined adenomatous polyp as a polyp with an adenomas component that was histologically proven. Nonadenomatous polyps were polyps without an adenomatous component, including hyperplastic polyps and serrated lesions. Serrated polyps, when reported separately, were defined as polyps with a serrated component, including traditional serrated adenomas and sessile serrated

Literature search and bias assessment

Among the 8004 studies retrieved, 6763 were selected after removal of duplicates. Eighteen studies were selected for inclusion in the meta-analysis of polyp histology prediction and 6 studies for polyp detection analysis according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses flowchart (Supplementary Fig. 1, available online at www.giejournal.org). The characteristics of these included studies are presented in Tables 1 and 2.

The quality of included studies and the

Discussion

Although a number of studies had reported the performance of AI on colorectal polyp detection and characterization, this is the first systematic review and meta-analysis on the performance of AI on histology prediction and detection of colorectal polyps. Herein, we showed that AI was accurate on histology prediction, with a pooled sensitivity of 92.3%, specificity 89.8%, and AUC .96. For polyp detection, the respective pooled sensitivity and specificity of AI was 95% and 88%, with AUC of .90.

References (46)

D.K. Rex et al.
The American Society for Gastrointestinal Endoscopy PIVI (Preservation and Incorporation of Valuable Endoscopic Innovations) on real-time endoscopic assessment of the histology of diminutive colorectal polyps
Gastrointest Endosc
(2011)
B.K. Abu Dayyeh et al.
ASGE technology committee systematic review and meta-analysis assessing the ASGE PIVI thresholds for adopting real-time endoscopic assessment of the histology of diminutive colorectal polyps
Gastrointest Endosc
(2015)
N. Gupta et al.
AGA white paper: training and implementation of endoscopic image enhancement technologies
Clin Gastroenterol Hepatol
(2017)
S. Gross et al.
Computer-based classification of small colorectal polyps by using narrow-band imaging with optical magnification
Gastrointest Endosc
(2011)
Y. Takemura et al.
Computer-aided system for predicting the histology of colorectal tumors by using narrow-band imaging magnifying colonoscopy (with video)
Gastrointest Endosc
(2012)
Y. Mori et al.
Novel computer-aided diagnostic system for colorectal lesions by using endocytoscopy (with videos)
Gastrointest Endosc
(2015)
Y. Kominami et al.
Computer-aided diagnosis of colorectal polyp histology by using a real-time image recognition system and narrow-band imaging magnifying colonoscopy
Gastrointest Endosc
(2016)
P.J. Chen et al.
Accurate classification of diminutive colorectal polyps using computer-aided analysis
Gastroenterology
(2018)
X. Zhu et al.
Detection and diagnosis of sessile serrated adenoma/polyps using convolutional neural network (artificial intelligence)
Gastrointest Endosc
(2018)
M. Misawa et al.
Artificial intelligence- assisted polyp detection system for colonoscopy, based on the largest available collection of clinical video data for machine learning [abstract]
Gastrointest Endosc
(2019)

T. Ka-Luen Lui et al.

Use of artificial intelligence image classifier for real-time detection of colonic polyps

Gastrointest Endosc

(2019)

C. Rombaoa et al.

Automated insertion time, cecal intubation and withdrawal time during live colonoscopy using convolutional neural networks—a video validation study [abstract]

Gastrointest Endosc

(2019)

S.K. McGill et al.

Narrow band imaging to differentiate neoplastic and non-neoplastic colorectal polyps in real time: a meta-analysis of diagnostic operating characteristics

Gut

(2013)

I. Hirata et al.

Usefulness of magnifying narrow-band imaging endoscopy for the diagnosis of gastric and colorectal lesions

Digestion

(2012)

Y. Takeuchi et al.

Proposal of a new “resect and discard” strategy using magnifying narrow band imaging: pilot study of diagnostic accuracy

Dig Endosc

(2014)

D.G. Hewett et al.

Validation of a simple classification system for endoscopic diagnosis of small colorectal polyps using narrow-band imaging

Gastroenterology

(2012)

C. Hassan et al.

A resect and discard strategy would improve cost-effectiveness of colorectal cancer screening

Clin Gastroenterol Hepatol

(2010)

J.E. East et al.

Advanced endoscopic imaging: European Society of Gastrointestinal Endoscopy (ESGE) technology review

Endoscopy

(2016)

J. Tischendorf et al.

Computer-aided classification of colorectal polyps based on vascular patterns: a pilot study

Endoscopy

(2010)

B. André et al.

Software for automated classification of probe-based confocal laser endomicroscopy videos of colorectal polyps

World J Gastroenterol

(2012)

T. Rath et al.

In vivo real-time assessment of colorectal polyp histology using an optical biopsy forceps system based on laser-induced fluorescence spectroscopy

Endoscopy

(2016)

Y. Mori et al.

Impact of an automated system for endocytoscopic diagnosis of small colorectal lesions: an international web-based study

Endoscopy

(2016)

M.F. Byrne et al.

Real-time differentiation of adenomatous and hyperplastic diminutive colorectal polyps during analysis of unaltered videos of standard colonoscopy using a deep learning model

Gut

(2019)

Cited by (82)

Application of artificial intelligence in modern medicine
2023, Clinical eHealth
Over the last decade, artificial intelligence in medicine has attracted much attention and interest for its robust automation and efficiency in disease diagnosis, treatment and prognosis. It has shown great potential in bringing the revolutionary change in clinical practice. This review aimed to summarize the progress AI has made recently in several fields of medicine, including patient screening, diagnosis and treatment of cancerous and non-cancerous diseases, pharmaceutical development, clinical trial management and basic science. Meanwhile, we will discuss its future direction and important considerations, focusing on its utility, usability, validation, safety and ethnic problem.
Computer-assisted detection versus conventional colonoscopy for proximal colonic lesions: a multicenter, randomized, tandem-colonoscopy study
2023, Gastrointestinal Endoscopy
Computer-assisted detection (CADe) is a promising technologic advance that enhances adenoma detection during colonoscopy. However, the role of CADe in reducing missed colonic lesions is uncertain. The aim of this study was to determine the miss rates of proximal colonic lesions by CADe and conventional colonoscopy.
This was a prospective, multicenter, randomized, tandem-colonoscopy study conducted in 3 Asian centers. Patients were randomized to receive CADe or conventional white-light colonoscopy during the first withdrawal of the proximal colon (cecum to splenic flexure), immediately followed by tandem examination of the proximal colon with white light in both groups. The primary outcome was adenoma/polyp miss rate, which was defined as any adenoma/polyp detected during the second examination.
Of 223 patients (48.6% men; median age, 63 years) enrolled, 7 patients did not have tandem examination, leaving 108 patients in each group. There was no difference in the miss rate for proximal adenomas (CADe vs conventional: 20.0% vs 14.0%, P = .07) and polyps (26.7% vs 19.6%, P = .06). The CADe group, however, had significantly higher proximal polyp (58.0% vs 46.7%, P = .03) and adenoma (44.7% vs 34.6%, P = .04) detection rates than the conventional group. The mean number of proximal polyps and adenomas detected per patient during the first examination was also significantly higher in the CADe group (polyp: 1.20 vs .86, P = .03; adenoma, .91 vs .61, P = .03). Subgroup analysis showed that CADe enhanced proximal adenoma detection in patients with fair bowel preparation, shorter withdrawal time, and endoscopists with lower adenoma detection rate.
This multicenter trial from Asia confirmed that CADe can further enhance proximal adenoma and polyp detection but may not be able to reduce the number of missed proximal colonic lesions. (Clinical trial registration number: NCT04294355.)
Development of a predictive model for classifying colorectal cancer using principal component analysis
2023, Deep Learning in Personalized Healthcare and Decision Support
Colorectal cancer (CRC) is the leading second cause of cancer death cases in both genders in the United States. On the other hand, smoking, an unhealthy lifestyle, excessive alcohol use, physical inactivity, and excess body weight represent more than half of all cases and deaths, making them possibly avoidable. With effective screening and monitoring, CRC morbidity and death can be decreased. This study gives a review of colon-rectal cancer prediction techniques using machine learning classifiers. Machine learning (ML) approach is profound for cancer prediction and analysis, which is implemented through Juypter via a coding approach. An experimental analysis was performed on the dataset gotten from Kaggle using support vector machine, K-nearest neighbor, and random forest, it also made use of principal component analysis for dimensionality reduction. The dataset consists of 5000 instances and 785 attributes. The training set (80%) was used to determine the optimal combinations of variables that will generate a good predictive model and the testing set (20%) was used to provide an unbiased evaluation of a final model fit on the training dataset. The results of the classification technique showed that support vector machine, KNN, and random forest had accuracy of 77.75%, 81.00%, and 84.00%, respectively, without the application of PCA, while technique showed that SVM, KNN, and random forest had accuracy of 80.40%, 81.90%, and 96.00%, respectively, with the application of PCA. It is imperative to note that a good number of the classification techniques were able to predict CRC based on the Colon Cancer Gene Expression dataset. From the findings, support vector machine had the minimum accuracy 80.40%, sensitivity as well as specificity, while random forest had the highest accuracy 96.00%, sensitivity, and specificity and therefore was chosen as the best classifier for colon cancer prediction. Therefore, beyond accuracy, other performance evaluation metrics are critical for optimal algorithm selection for predictive analytics.
Approach of artificial intelligence in colorectal cancer and in precision medicine
2023, Computational Methods in Drug Discovery and Repurposing for Cancer Therapy
Artificial Intelligence (AI) is a computer-based technology where the computer is programmed to identify, learn, reason, and act like humans. AI has a groundbreaking resolution to different medical problems especially in cancer. Colorectal cancer (CRC) is the third most fatal cancer and upsurging in men and young adults due to unhealthy diet, heavy consumption of alcohol, smoking, and genetic and environmental conditions. The applications of AI to CRC include detection, diagnosis, grading, characterization, drug discovery, treatment, and prevention. With the increase in AI technology and research, a personalized medicine therapy called precision medicine is developed. This helps with accurate, unique drug design and medical treatment for the patients. CRC is a heterogeneous disease and is considered as model disease for the complete implementation of precision medicine concept. The applications, benefits, drawbacks, current challenges, and further prospects of AI in CRC and precision medicine are discussed in the article.
Current and Potential Applications of Artificial Intelligence in Endoscopic Imaging of Gastrointestinal Tumor: An Umbrella Review
2024, SSRN
Evaluating ChatGPT in pathology: towards multimodal AI in medical imaging
2024, Journal of Clinical Pathology

View all citing articles on Scopus

: DISCLOSURE: The following author disclosed financial relationships: W. K. Leung: Advisory board for NISI (Hong Kong) Company; consultant for Medtronics. All other authors disclosed no financial relationships.

: If you would like to chat with an author of this article, you may contact Dr Leung at [email protected].

View full text

Systematic review and meta-analysisAccuracy of artificial intelligence on histology prediction and detection of colorectal polyps: a systematic review and meta-analysis

Background and Aims

Method

Results

Conclusions

Section snippets

Definition

Literature search and bias assessment

Discussion

Gastrointest Endosc

Gastrointest Endosc

Clin Gastroenterol Hepatol

Gastrointest Endosc

Gastrointest Endosc

Gastrointest Endosc

Gastrointest Endosc

Gastroenterology

Gastrointest Endosc

Gastrointest Endosc

Gastrointest Endosc

Gastrointest Endosc

Narrow band imaging to differentiate neoplastic and non-neoplastic colorectal polyps in real time: a meta-analysis of diagnostic operating characteristics

Gut

Usefulness of magnifying narrow-band imaging endoscopy for the diagnosis of gastric and colorectal lesions

Digestion

Proposal of a new “resect and discard” strategy using magnifying narrow band imaging: pilot study of diagnostic accuracy

Dig Endosc

Validation of a simple classification system for endoscopic diagnosis of small colorectal polyps using narrow-band imaging

Gastroenterology

A resect and discard strategy would improve cost-effectiveness of colorectal cancer screening

Clin Gastroenterol Hepatol

Advanced endoscopic imaging: European Society of Gastrointestinal Endoscopy (ESGE) technology review

Endoscopy

Computer-aided classification of colorectal polyps based on vascular patterns: a pilot study

Endoscopy

Software for automated classification of probe-based confocal laser endomicroscopy videos of colorectal polyps

World J Gastroenterol

In vivo real-time assessment of colorectal polyp histology using an optical biopsy forceps system based on laser-induced fluorescence spectroscopy

Endoscopy

Impact of an automated system for endocytoscopic diagnosis of small colorectal lesions: an international web-based study

Endoscopy

Real-time differentiation of adenomatous and hyperplastic diminutive colorectal polyps during analysis of unaltered videos of standard colonoscopy using a deep learning model

Gut

Systematic review and meta-analysis
Accuracy of artificial intelligence on histology prediction and detection of colorectal polyps: a systematic review and meta-analysis