Role of magnifying colonoscopy for diagnosis of colorectal neoplasms: From the perspective of Japanese colonoscopists

Artificial Intelligence Applied to Colonoscopy: Is It Time to Take a Step Forward?

Growing evidence indicates that artificial intelligence (AI) applied to medicine is here to stay. In gastroenterology, AI computer vision applications have been stated as a research priority. The two main AI system categories are computer-aided polyp detection (CADe) and computer-assisted diagnosis (CADx). However, other fields of expansion are those related to colonoscopy quality, such as methods to objectively assess colon cleansing during the colonoscopy, as well as devices to automatically predict and improve bowel cleansing before the examination, predict deep submucosal invasion, obtain a reliable measurement of colorectal polyps and accurately locate colorectal lesions in the colon. Although growing evidence indicates that AI systems could improve some of these quality metrics, there are concerns regarding cost-effectiveness, and large and multicentric randomized studies with strong outcomes, such as post-colonoscopy colorectal cancer incidence and mortality, are lacking. The integration of all these tasks into one quality-improvement device could facilitate the incorporation of AI systems in clinical practice. In this manuscript, the current status of the role of AI in colonoscopy is reviewed, as well as its current applications, drawbacks and areas for improvement.

Diagnosis of Depth of Submucosal Invasion in Colorectal Cancer with AI Using Deep Learning

The submucosal invasion depth predicts prognosis in early colorectal cancer. Although colorectal cancer with shallow submucosal invasion can be treated via endoscopic resection, colorectal cancer with deep submucosal invasion requires surgical colectomy. However, accurately diagnosing the depth of submucosal invasion via endoscopy is difficult. We developed a tool to diagnose the depth of submucosal invasion in early colorectal cancer using artificial intelligence. We reviewed data from 196 patients who had undergone a preoperative colonoscopy at the Osaka University Hospital and Osaka International Cancer Institute between 2011 and 2018 and were diagnosed pathologically as having shallow submucosal invasion or deep submucosal invasion colorectal cancer. A convolutional neural network for predicting invasion depth was constructed using 706 images from 91 patients between 2011 and 2015 as the training dataset. The diagnostic accuracy of the constructed convolutional neural network was evaluated using 394 images from 49 patients between 2016 and 2017 as the validation dataset. We also prospectively tested the tool from 56 patients in 2018 with suspected early-stage colorectal cancer. The sensitivity, specificity, accuracy, and area under the curve of the convolutional neural network for diagnosing deep submucosal invasion colorectal cancer were 87.2% (258/296), 35.7% (35/98), 74.4% (293/394), and 0.758, respectively. The positive predictive value was 84.4% (356/422) and the sensitivity was 75.7% (356/470) in the test set. The diagnostic accuracy of the constructed convolutional neural network seemed to be as high as that of a skilled endoscopist. Thus, endoscopic image recognition by deep learning may be able to predict the submucosal invasion depth in early-stage colorectal cancer in clinical practice.

Histopathological features for coexistent invasive cancer in large colorectal adenomatous polyps

Background

Histopathological features associated with coexistent invasive adenocarcinoma in large colorectal adenomas have not been described. This study aimed to determine the association of histopathological features in areas of low-grade dysplasia with coexistent invasive adenocarcinoma.

Methods

High-grade lesions (containing high-grade dysplasia or adenocarcinoma) from a cohort of large (at least 20 mm) colorectal adenomas removed by endoscopic resection were subjected to detailed histopathological analysis. The histopathological features in low-grade areas with coexistent adenocarcinoma were reviewed and their diagnostic performance was evaluated.

Results

Seventy-four high-grade lesions from 401 endoscopic resections of large adenomas were included. In the low-grade dysplastic areas, a coexistent invasive adenocarcinoma was associated significantly with a cribriform or trabecular growth pattern (P < 0.001), high nuclear grade (P < 0.001), multifocal intraluminal necrosis (P < 0.001), atypical mitotic figures (P = 0.006), infiltrative lesion edges (P < 0.001), a broad fibrous band (P = 0.001), ulceration (P < 0.001), expansile nodules (P < 0.001) and an extensive tumour-infiltrating lymphocyte pattern (P = 0.04). Lesions with coexistent invasive adenocarcinoma harboured at least one of these features. The area under the receiver operating characteristic curve (AUROC) for coexistent invasive adenocarcinoma, using frequencies of adverse histopathological factors in low-grade areas, was 0.92. The presence of two or more of these adverse histopathological features in low-grade areas had a sensitivity of 86 per cent and a specificity of 84 per cent for coexistent invasive adenocarcinoma.

Conclusion

Several histopathological features in low-grade dysplastic areas of adenomas could be predictive of coexistent adenocarcinoma.

Dye-based chromoendoscopy following polypectomy reduces incomplete polyp resection

Background and study aims  The completeness of a polyp resection is an important determinant of quality in colonoscopy, and may reduce incidence of interval cancers. Incomplete resection rates (IRR) vary widely and range from 6.5 % to 22.7 %. Residual disease is more likely with larger polyps, for sessile serrated adenomas, and with more proximal lesions. Chromoendoscopy is increasingly employed in lesion detection. The aims of this study were to assess local IRR, and to determine whether chromoendoscopy could correctly identify residual disease post polypectomy. Patients and methods  This was a prospective study examining post polypectomy sites. Chromoendoscopy (0.13 % indigo carmine) was applied to resection bases to identify residual disease. Targeted base biopsies were taken from identified residual disease (positive group) or random base biopsies were taken when a clear base was visualised (negative group). Overall rates of incomplete resection were documented. Reported rates post chromoendoscopy and actual histological rates were documented and compared. Results  A total of 102 polyps were identified for inclusion, of which 15 % (n = 16) were excluded. Resection quality was evaluated in 86 polyps of 61 patients (female n = 33 54 %; mean Age 62.3 years). Polyps were mainly removed by cold snare (n = 71, 82.5 %). Most polyps (n = 58, 67 %) measured between 5 to 10 mm. Polyps were largely located in the right colon (n = 57, 66 %). Overall histological residual disease occurred in 17 /86 (19.6 %). Chromoendoscopy correctly identified residual disease in 13 of 17 bases (76.5 %). Only four of /86 (4.6 %) of polyp bases were missclassified post-chromoendoscopy (odds ratio 0.284 (95 % CI 0.0857-0.9409), P  = 0.03). Conclusion  Indigo carmine chromoendoscopy improves early detection of residual disease post polypectomy, reducing incomplete resection rates.

Diagnostic accuracy of computed tomography colonography for tumor depth in colorectal cancer: A systematic review and meta-analysis

BACKGROUND

Accurate preoperative determination of tumor depth is a major concern in colorectal cancer. Assessment of tumor depth can influence the treatment strategy for colorectal cancer, but there is currently no universal diagnostic standard. This review sought to evaluate the diagnostic performance of computed tomography (CT) colonography for tumor infiltration depth in colorectal cancer.

METHODS

The PubMed, Scopus, and Cochrane Central Register of Controlled Trials databases were searched to identify studies that compared the tumor depth using CT colonography with that determined using pathological assessment. Relevant data were extracted from the eligible studies by two review authors working independently. Any disagreement was resolved by discussion. The quality of the studies was assessed using the QUADAS-2 tool. The diagnostic performance of CT colonography for tumor depth in colorectal cancer was evaluated by meta-analysis using sensitivity and specificity values and diagnostic odds ratios.

RESULTS

The systematic literature search identified 16 eligible studies. The pooled sensitivity and specificity values and diagnostic odds ratios were 97.8% (95% confidence interval [CI] 88.1-99.6), 76.4% (95% CI 59.9-87.5), and 142.6 (95% CI 20.0-1071.0) for T2 invasion, 97.7% (95% CI 92.1-99.3), 83.5% (95% CI 75.9-89.0), and 212.1 (95% CI 50.6-888.1) for T3 invasion, and 84.9% (95% CI 61.7-95.1), 97.5% (95% CI 93.2-99.1), and 219.5 (95% CI 37.8-1273.6) for T4 invasion, respectively.

CONCLUSIONS

CT colonography was found to have high diagnostic performance and was useful for determining tumor depth in colorectal cancer.

Comparison of endoscopic ultrasonography and magnifying endoscopy for assessment of the invasion depth of shallow gastrointestinal neoplasms: a systematic review and meta-analysis

OBJECTIVES

To conduct a meta-analysis to provide accurate evidence regarding the preferred diagnostic method, magnifying endoscopy (ME) or endoscopic ultrasonography (EUS), for assessment of the depth of invasion of the gastrointestinal neoplasms.

METHODS

PubMed, EMBASE, Ovid Medline, and the Cochrane Library databases were searched for studies published between January 1946 and October 2016, regarding the use of EUS and ME to assess the invasion depth of gastrointestinal cancers. The quality of diagnostic studies was evaluated using the QUADAS2 instrument. The Meta-DiSc software (version 1.4) was used for meta-analysis of the pooled data regarding the diagnostic accuracy of EUS and ME of the invasion depth of gastrointestinal neoplasms.

RESULTS

Our meta-analysis included the data of 754 patients with gastrointestinal cancers contributed by seven prospective studies. All studies were of high quality (QUADAS2). The receiver operating characteristic (ROC) planes were not observed in shoulder and arm forms for either EUS or ME, with Spearman's correlation coefficients of -0.821 and 0.234 for EUS and ME, respectively. The p values of the diagnostic odds ratio for EUS and ME were 0.0038 and 0.0131, respectively. The sensitivity and specificity of EUS for the diagnosis of the depth of invasion of gastrointestinal cancers were 0.75 (95% CI 0.69-0.81) and 0.84 (95% CI 0.79-0.88), respectively. In comparison, the sensitivity and specificity for ME were 0.74 (95% CI 0.67-0.69) and 0.85 (95% CI 0.80-0.89), respectively. The values of area under the summary ROC (SROC) curves for EUS and ME were 0.8499 and 0.8757, respectively, with a non-significant Z value between EUS and MR (0.296 < 1.96).

CONCLUSIONS

Both EUS and ME provide a comparable performance for judging the depth of invasion of gastrointestinal neoplasms. However, there is heterogeneity between studies contributed by non-threshold effects.