Colonoscopy Videos Research Articles

P477 Performance of bowel preparation quality scales in patients with Crohn’s disease

Abstract Background The performance of bowel preparation (BP) in patients with Crohn’s disease (CD) is unknown and may be suboptimal due to the presence of mucosal inflammation, strictures, pseudopolyps and bowel resection. We evaluated the reliability and validity of available BP scales in patients with CD. Methods Bowel preparation, in colonoscopy videos (N=50) from patients with CD (N=40) that represented a range of bowel preparation quality and endoscopic disease activity, was independently rated twice, separated by at least 2 weeks, by 3 central readers using the Boston Bowel Preparation Score (BBPS), modified BBPS, Harefield Cleansing Scale (HCS), Food and Drug Administration bowel cleansing assessment scale (FDA BCAS), and a 100-mm visual analogue scale (VAS) of bowel cleanliness. Endoscopic activity was assessed with the Simple Endoscopic Score for CD (SES-CD). All assessments were on endoscope insertion and withdrawal and in the ileum, right colon, transverse colon, left colon, and rectum to facilitate evaluation of the relationship between BP and SES-CD scores. Reliability of BP assessment was quantified using the intraclass correlation coefficient (ICC) and interpreted with benchmarks defined by Landis and Koch.1 Validity was assessed by within-patient correlation between each BP scale and the VAS across segments using mixed effect models. Correlation between BP quality and SES-SD scores by location was assessed using Spearman’s rho. Results Substantial (ICC ≥0.61) inter-rater reliability was observed for all BP scales and the VAS during insertion and withdrawal, except for the FDA BCAS, which had moderate (ICC ≥0.41) inter-rater reliability on insertion (Table 1). Intra-rater reliability was similarly substantial for all BP scales and almost perfect (ICC ≥0.81) for the VAS during insertion and withdrawal. Coefficients for the validity of all BP scales based on correlation with the VAS of bowel cleanliness exceeded 0.58 on both insertion and withdrawal. BP and endoscopic disease activity were negatively correlated in the colon, particularly in the left colon, suggesting that lower BP scores in this segment may result in higher SES-CD scores (Table 2). Conclusion Existing scales are reliable for the assessment of BP in patients with CD. These results provide a framework for scoring of BP quality in clinical trials for CD and support the inclusion of patients with CD for the evaluation of novel BP agents. Reference: 1. Landis J.R. and Koch G. G. Biometrics. 1977:33;159-74

P098 The Endoscopic Severity Score Map (ESSM): An Artificial Intelligence scoring system providing accurate, objective and localised measurements of endoscopic disease severity in ulcerative colitis

Abstract Background Commonly used scoring schemes as the Mayo Endoscopic Subscore (MES) account for disease severity only at specific (i.e., the worst) segments and do not capture disease extent. However, for an accurate assessment of disease severity in patients with ulcerative colitis (UC), the measure of the precise extent of disease activity is necessary. Alternative systems that include disease extent have been proposed (Balint 2018), but their implementation is prohibited by the time and cost constraints of comprehensively scoring each location along the entire colon. Here, we present the Endoscopic Severity Score Map (ESSM), a scoring system based on Artificial Intelligence, capable of providing an assessment of disease severity and extent in UC in a fully automated manner. Methods The ESSM consists of 3 main elements (Fig. 1): 1) a quality algorithm which selects readable frames from a colonoscopy video, 2) a scoring system which assigns an MES to each readable frame (Gutierrez Becker 2020) and 3) a camera localisation algorithm that assigns each frame to an anatomical location within the colon (Yao 2022). The ESSM was trained and tested using 4,306 sigmoidoscopy videos from phase III Etrolizumab clinical trials (Hickory NCT02100696, Laurel NCT02165215, Hibiscus I NCT02163759, Hibiscus II NCT02171429 and Gardenia NCT02136069). Results We evaluate the performance of the ESSM by first assessing the agreement of scoring as compared to centrally read MES. The agreement between central reading and the ESSM at the colon section level was high (quadratic-weighted kappa k=0.81; Tab. 1). This was comparable to the agreement between central and local reading (k=0.84), suggesting that the ESSM shows levels of inter-rater variability comparable to experienced readers. Finally, we found correlations between the average ESSM at all anatomical locations and other disease activity markers to be moderate to high: faecal calprotectin rs=0.24, CRP rs=0.29, stool frequency rs=0.49, rectal bleeding rs=0.43 and physician global assessment rs=0.47 (Tab. 1). Conclusion Here, we introduced the ESSM, a fully-automated AI-based scoring system that enables accurate, objective and localised assessment of disease severity in UC. In brief, we show that the ESSM compares well with central reading (at the colon section level) and has clinical relevance when compared to other markers of disease activity. This tentatively suggests that the ESSM has the potential to augment the current way of assessing disease severity, both in clinical trials and everyday clinical practice.

Spatio-Temporal Feature Transformation Based Polyp Recognition for Automatic Detection: Higher Accuracy than Novice Endoscopists in Colorectal Polyp Detection and Diagnosis.

Artificial intelligence represents an emerging area with promising potential for improving colonoscopy quality. To develop a colon polyp detection model using STFT and evaluate its performance through a randomized sample experiment. Colonoscopy videos from the Digestive Endoscopy Center of the First Affiliated Hospital of Anhui Medical University, recorded between January 2018 and November 2022, were selected and divided into two datasets. To verify the model's practical application in clinical settings, 1500 colonoscopy images and 1200 polyp images of various sizes were randomly selected from the test set and compared with the STFT model's and endoscopists' recognition results with different years of experience. In the randomized sample trial involving 1500 colonoscopy images, the STFT model demonstrated significantly higher accuracy and specificity compared to endoscopists with low years of experience (0.902 vs. 0.809, 0.898 vs. 0.826, respectively). Moreover, the model's sensitivity was 0.904, which was higher than that of endoscopists with low, medium, or high years of experience (0.80, 0.896, 0.895, respectively), with statistical significance (P < 0.05). In the randomized sample experiment of 1200 polyp images of different sizes, the accuracy of the STFT model was significantly higher than that of endoscopists with low years of experience when the polyp size was ≤ 0.5cm and 0.6-1.0cm (0.902 vs. 0.70, 0.953 vs. 0.865, respectively). The STFT-based colon polyp detection model exhibits high accuracy in detecting polyps in colonoscopy videos, with a particular efficiency in detecting small polyps (≤ 0.5cm)(0.902 vs. 0.70, P < 0.001).

Density clustering-based automatic anatomical section recognition in colonoscopy video using deep learning

Recognizing anatomical sections during colonoscopy is crucial for diagnosing colonic diseases and generating accurate reports. While recent studies have endeavored to identify anatomical regions of the colon using deep learning, the deformable anatomical characteristics of the colon pose challenges for establishing a reliable localization system. This study presents a system utilizing 100 colonoscopy videos, combining density clustering and deep learning. Cascaded CNN models are employed to estimate the appendix orifice (AO), flexures, and "outside of the body," sequentially. Subsequently, DBSCAN algorithm is applied to identify anatomical sections. Clustering-based analysis integrates clinical knowledge and context based on the anatomical section within the model. We address challenges posed by colonoscopy images through non-informative removal preprocessing. The image data is labeled by clinicians, and the system deduces section correspondence stochastically. The model categorizes the colon into three sections: right (cecum and ascending colon), middle (transverse colon), and left (descending colon, sigmoid colon, rectum). We estimated the appearance time of anatomical boundaries with an average error of 6.31 s for AO, 9.79 s for HF, 27.69 s for SF, and 3.26 s for outside of the body. The proposed method can facilitate future advancements towards AI-based automatic reporting, offering time-saving efficacy and standardization.

Subjective and Objective Quality Assessment of Colonoscopy Videos.

Captured colonoscopy videos usually suffer from multiple real-world distortions, such as motion blur, low brightness, abnormal exposure, and object occlusion, which impede visual interpretation. However, existing works mainly investigate the impacts of synthesized distortions, which differ from real-world distortions greatly. This research aims to carry out an in-depth study for colonoscopy Video Quality Assessment (VQA). In this study, we advance this topic by establishing both subjective and objective solutions. Firstly, we collect 1,000 colonoscopy videos with typical visual quality degradation conditions in practice and construct a multi-attribute VQA database. The quality of each video is annotated by subjective experiments from five distortion attributes (i.e., temporal-spatial visibility, brightness, specular reflection, stability, and utility), as well as an overall perspective. Secondly, we propose a Distortion Attribute Reasoning Network (DARNet) for automatic VQA. DARNet includes two streams to extract features related to spatial and temporal distortions, respectively. It adaptively aggregates the attribute-related features through a multi-attribute association module to predict the quality score of each distortion attribute. Motivated by the observation that the rating behaviors for all attributes are different, a behavior guided reasoning module is further used to fuse the attribute-aware features, resulting in the overall quality. Experimental results on the constructed database show that our DARNet correlates well with subjective ratings and is superior nine state-of-the-art methods.

A comprehensive study on automatic non‐informative frame detection in colonoscopy videos

AbstractDespite today's developing healthcare technology, conventional colonoscopy is still a gold‐standard method to detect colon abnormalities. Due to the folded structure of the intestine and visual disturbances caused by artifacts, it can be hard for specialists to detect abnormalities during the procedure. Frames that include artifacts such as specular reflection, improper contrast levels from insufficient or excessive illumination gastric juice, bubbles, or residuals should be detected to increase an accurate diagnosis rate. In this work, both conventional machine learning and transfer learning methods have been used to detect non‐informative frames in colonoscopy videos. The conventional machine learning part consists of 5 different types of texture features, which are gray level co‐occurrence matrix (GLCM), gray level run length matrix (GLRLM), neighborhood gray‐tone difference matrix (NGTDM), focus measure operators (FMOs), and first‐order statistics. In addition to these methods, we utilized 8 different transfer learning models: AlexNet, SqueezeNet, GoogleNet, ShuffleNet, ResNet50, ResNet18, NasNetMobile, and MobileNet. The results showed that FMOs and decision tree combination gave the best accuracy and f‐measure values with almost 89% and 0.79%, respectively, for the conventional machine learning part. When the transfer learning part is taken into account, AlexNet (99.85%) and SqueezeNet (98.80%) have the highest performance metric results. This study shows the potential of both transfer learning and conventional machine learning algorithms to provide fast and accurate non‐informative frame detection to be used during a colonoscopy, which may be considered the initial step in identifying and classifying colon‐related diseases automatically to help guide physicians.

Exponential Chef based Optimization Enabled Deep Learning for Polyp Frame Detection and Polyp Segmentation using Colonoscopy Videos

Exponential Chef based Optimization Enabled Deep Learning for Polyp Frame Detection and Polyp Segmentation using Colonoscopy Videos

Hyperplastic and tubular polyp classification using machine learning and feature selection

PurposeThe aim of this study is to develop an effective approach for differentiating between hyperplastic and tubular adenoma colon polyps, which is one of the most difficult tasks in colonoscopy procedures. The main research challenge is how to improve the classification of these polyp subtypes applying various focusing levels on the polyp images, data preprocessing approaches, and classification algorithms. MethodsThis study employed 202 colonoscopy videos from a total of 201 patients, focusing on 59 videos containing hyperplastic and tubular adenoma polyps. Manually extract key frames and several feature extraction and classification techniques were applied. The influence of different datasets with various focuses as well as data preprocessing steps on the performance of classification was examined, and AUC values were calculated using ten classifiers. ResultsThe study discovered that the optimal dataset, data preprocessing method, and classification algorithm all had significant effects on classification results. The Random Forest model with the Recursive Feature Elimination (RFE) feature selection approach, for example, consistently outperformed other models and achieved the highest AUC value of 0.9067. In terms of accuracy, F1 score, recall, and AUC, the suggested model outperformed a gastroenterologist, nevertheless precision remained slightly lower. ConclusionThis study emphasizes the importance of dataset selection, data preprocessing, and feature selection in enhancing the classification of difficult colon polyp subtypes. The suggested model offers a promising model for the clinical differentiation of hyperplastic and tubular adenoma polyps, potentially improving diagnostic accuracy in gastroenterology.

Assessing clinical efficacy of polyp detection models using open-access datasets.

Ensuring accurate polyp detection during colonoscopy is essential for preventing colorectal cancer (CRC). Recent advances in deep learning-based computer-aided detection (CADe) systems have shown promise in enhancing endoscopists' performances. Effective CADe systems must achieve high polyp detection rates from the initial seconds of polyp appearance while maintaining low false positive (FP) detection rates throughout the procedure. We integrated four open-access datasets into a unified platform containing over 340,000 images from various centers, including 380 annotated polyps, with distinct data splits for comprehensive model development and benchmarking. The REAL-Colon dataset, comprising 60 full-procedure colonoscopy videos from six centers, is used as the fifth dataset of the platform to simulate clinical conditions for model evaluation on unseen center data. Performance assessment includes traditional object detection metrics and new metrics that better meet clinical needs. Specifically, by defining detection events as sequences of consecutive detections, we compute per-polyp recall at early detection stages and average per-patient FPs, enabling the generation of Free-Response Receiver Operating Characteristic (FROC) curves. Using YOLOv7, we trained and tested several models across the proposed data splits, showcasing the robustness of our open-access platform for CADe system development and benchmarking. The introduction of new metrics allows for the optimization of CADe operational parameters based on clinically relevant criteria, such as per-patient FPs and early polyp detection. Our findings also reveal that omitting full-procedure videos leads to non-realistic assessments and that detecting small polyp bounding boxes poses the greatest challenge. This study demonstrates how newly available open-access data supports ongoing research progress in environments that closely mimic clinical settings. The introduced metrics and FROC curves illustrate CADe clinical efficacy and can aid in tuning CADe hyperparameters.

Real-time detection of colon polyps during colonoscopy using YOLOv7

Deep learning has made brilliant achievements in detecting colonic polyps in colonoscopy videos in recent years. However, the detection of colonic polyps in colonoscopy videos is problematic because of the complex environment of the colon and the various shapes of polyps. Therefore, researchers need to spend a lot of time searching for real-time detection systems with good performance and that are suitable for the current equipment and working environment. This paper aimed to investigate the polyp detection potential of the state-of-the-art deep learning model You Only Look Once version 7. We implemented, trained, and tested the polyp detection model using open public datasets: Kvasir-Seg, CVC-ClinicDB, CVC_ColonDB, and ETIS-LaribPolypDB. Validation of the test set utilizing Recall, Precision, F1 Score, and Average Precision (AP) showed that the model achieved the highest performance on CVC-ClinicDB with 83.3% Recall, 80.6% Precision, 81.9% F1 Score, 75% AP@0.5, 51.8% AP and the mean processing time per frame was 20ms. The automatic polyp detection model exhibited good performance, as evidenced by the high detection sensitivity and rapid processing. This model can help endoscopists improve polyp detection performance during the colonoscopy procedure.

Development and validation of a three-dimensional deep learning-based system for assessing bowel preparation on colonoscopy video.

The performance of existing image-based training models in evaluating bowel preparation on colonoscopy videos was relatively low, and only a few models used external data to prove their generalization. Therefore, this study attempted to develop a more precise and stable AI system for assessing bowel preparation of colonoscopy video. We proposed a system named ViENDO to assess the bowel preparation quality, including two CNNs. First, Information-Net was used to identify and filter out colonoscopy video frames unsuitable for Boston bowel preparation scale (BBPS) scoring. Second, BBPS-Net was trained and tested with 5,566 suitable short video clips through three-dimensional (3D) convolutional neural network (CNN) technology to detect BBPS-based insufficient bowel preparation. Then, ViENDO was applied to complete withdrawal colonoscopy videos from multiple centers to predict BBPS segment scores in clinical settings. We also conducted a human-machine contest to compare its performance with endoscopists. In video clips, BBPS-Net for determining inadequate bowel preparation generated an area under the curve of up to 0.98 and accuracy of 95.2%. When applied to full-length withdrawal colonoscopy videos, ViENDO assessed bowel cleanliness with an accuracy of 93.8% in the internal test set and 91.7% in the external dataset. The human-machine contest demonstrated that the accuracy of ViENDO was slightly superior compared to most endoscopists, though no statistical significance was found. The 3D-CNN-based AI model showed good performance in evaluating full-length bowel preparation on colonoscopy video. It has the potential as a substitute for endoscopists to provide BBPS-based assessments during daily clinical practice.

Analysis of the expression of matrix RNA of a panel of genes of morphologically unchanged rectal epithelium as a method of early diagnosis of colon pathology

Introduction. The absence of specific clinical symptoms in the early stages of colorectal cancer development leads to the fact that a quarter of patients who seek help for the first time have a metastatic stage of the disease. For the timely detection of pre-tumor disorders or hidden foci of malignancy, the possibilities of modern molecular biological technologies are being actively studied today.Aim. To develop a method for diagnosing tumor diseases of the colon based on molecular genetic analysis of morphologically unchanged intestinal epithelium distant from the focus of the tumor lesion.Materials and methods. We examined the matrix RNA (mRNA) expression profile of 63 candidate genes potentially associated with the pathogenesis of neoplastic changes in rectal mucosal samples. Samples were obtained during prophylactic and/or diagnostic video colonoscopy of 122 patients, 41 of whom had no history of breast cancer (“Normal”), 32 patients were diagnosed with breast cancer polyps (“Polyposis”) and 49 patients were diagnosed with breast cancer (“colorectal cancer”). mRNA expression was assessed by reverse transcription polymerase chain reaction.Results. Using the discriminant analysis method, it was established that the cellular material of scrapings from the rectum in the “colorectal cancer” group reliably, with a classification accuracy above 96 %, differs in expression phenotype from the “Normal” and “Polyposis” groups.Conclusion. The data obtained are a prerequisite for the development of a minimally invasive diagnostic method that can be used as part of an outpatient examination to assess the risk of colon tumor disease.

Colonoscopy 3D video dataset with paired depth from 2D-3D registration.

Screening colonoscopy is an important clinical application for several 3D computer vision techniques, including depth estimation, surface reconstruction, and missing region detection. However, the development, evaluation, and comparison of these techniques in real colonoscopy videos remain largely qualitative due to the difficulty of acquiring ground truth data. In this work, we present a Colonoscopy 3D Video Dataset (C3VD) acquired with a high definition clinical colonoscope and high-fidelity colon models for benchmarking computer vision methods in colonoscopy. We introduce a novel multimodal 2D-3D registration technique to register optical video sequences with ground truth rendered views of a known 3D model. The different modalities are registered by transforming optical images to depth maps with a Generative Adversarial Network and aligning edge features with an evolutionary optimizer. This registration method achieves an average translation error of 0.321 millimeters and an average rotation error of 0.159 degrees in simulation experiments where error-free ground truth is available. The method also leverages video information, improving registration accuracy by 55.6% for translation and 60.4% for rotation compared to single frame registration. 22 short video sequences were registered to generate 10,015 total frames with paired ground truth depth, surface normals, optical flow, occlusion, six degree-of-freedom pose, coverage maps, and 3D models. The dataset also includes screening videos acquired by a gastroenterologist with paired ground truth pose and 3D surface models. The dataset and registration source code are available at https://durr.jhu.edu/C3VD.

Intestinal endometriosis: A rare clinical image of sigmoid presentation

A 47-year-old female patient with a history of umbilical endometriosis and iron deficiency anemia presented to the gastroenterology department with a year-long complaint of abdominal pain, rectal pain, and hematochezia during menstruation. The physical examination and the results of laboratory studies of the patient were unremarkable. A contrast MRI of the abdomen revealed pathological thickening of 1.5 cm in the middle segment of the rectum and chocolate cysts on both ovaries, and the patient was referred for colonoscopic evaluation with suspicion of infiltrative endometriosis or colon neoplasm. An upper endoscopy and colonoscopy were performed. An actively bleeding mass in sigmoid colon, partially restricting the lumen but allowing for the passage of the video colonoscope, was observed (Figure 1A,1B: Sigmoid lesion). Biopsy specimens from the sigmoidal mass showed features consistent with endometriosis. A diagnosis of endometriosis was made. The patient underwent surgical intervention by sigmoid colon resection, hysterectomy, and bilateral oophorectomy. The complaints ceased on postoperative follow-up.

CCRA: A colon cleanliness rating algorithm based on colonoscopy video analysis

ObjectiveA Colon Cleanliness Rating Algorithm (CCRA) based on colonoscopy image analysis is proposed in this paper, in order to solve the problem that the results of Colon Cleanliness (or Bowel Preparation Quality) rating caused by manual inspection are inconsistent. MethodsFirstly, CCRA intercepts images from the colonoscopy video. Secondly, each colonoscopy image's stool area is segmented by U-Net to obtain the 2-classification segmentation results. Finally, the colon cleanliness is obtained by comparing the average area of the stool area with the standard proportion. ResultsAfter testing, the pixel accuracy of the U-Net model is 97.02 %, IoU is 83.67 %, accuracy is 92.17 %, recall is 90.21 %, F1-Score is 90.95 %. The accuracy of CCRA is 92.45 %–99.275 % ConclusionThe experimental results show that the CCRA proposed in this paper can quickly and accurately output the colon cleanliness rating of patients without manpower.

Comparative Analysis of U-Net and DeepLab for Automatic Polyp Segmentation in Colonoscopic Frames Using CVC-ClinicDB Dataset

In the context of medical image segmentation, accurate polyp detection in colonoscopy videos is crucial for early colorectal disease diagnosis. This study compares U-Net and DeepLab deep learning models in automatically segmenting polyps using the CVC-ClinicDB dataset. The dataset comprises 612 images from 31 colonoscopy sequences with a resolution of 384×288.Our primary metric is Mean Intersection over Union (IoU), measuring the overlap between predicted and ground truth masks. We also evaluate Mean Dice Loss for comprehensive comparison. U-Net demonstrates superior performance, with a Mean IoU Score of 0.9897 and a low Mean Dice Loss of 0.0523, indicating consistent and accurate polyp segmentation. In contrast, DeepLab achieves a Mean IoU Score of 0.9676 and a slightly higher Mean Dice Loss of 0.0417, showing good results but being outperformed by U-Net.In conclusion, U-Net excels in automatic polyp segmentation, offering high accuracy and robustness. These findings advance computer-aided diagnosis for colorectal diseases, potentially enhancing early and precise polyp detection.

Direct comparison of multiple computer-aided polyp detection systems.

Artificial intelligence (AI)-based systems for computer-aided detection (CADe) of polyps receive regular updates and occasionally offer customizable detection thresholds, both of which impact their performance, but little is known about these effects. This study aimed to compare the performance of different CADe systems on the same benchmark dataset. 101 colonoscopy videos were used as benchmark. Each video frame with a visible polyp was manually annotated with bounding boxes, resulting in 129 705 polyp images. The videos were then analyzed by three different CADe systems, representing five conditions: two versions of GI Genius, Endo-AID with detection Types A and B, and EndoMind, a freely available system. Evaluation included an analysis of sensitivity and false-positive rate, among other metrics. Endo-AID detection Type A, the earlier version of GI Genius, and EndoMind detected all 93 polyps. Both the later version of GI Genius and Endo-AID Type B missed 1 polyp. The mean per-frame sensitivities were 50.63 % and 67.85 %, respectively, for the earlier and later versions of GI Genius, 65.60 % and 52.95 %, respectively, for Endo-AID Types A and B, and 60.22 % for EndoMind. This study compares the performance of different CADe systems, different updates, and different configuration modes. This might help clinicians to select the most appropriate system for their specific needs.

Interobserver Variation of Bowel Preparation for Colonoscopy.

Several bowel preparation scales have been developed, but they are variably validated to guide clinical practice. The aim of this study was to assess the interobserver reliability between internal and external endoscopists using the Harefield Cleaning Scale (HCS). Data was collected from a study evaluating the efficacy of oral sulfate tablets in patients undergoing colonoscopy. The assessments of bowel cleansing were carried out by four internal endoscopists at the time of the procedure, and colonoscopy video recordings were reviewed independently by three external endoscopists. Interobserver reliability was quantified using Cohen's Kappa coefficient, and scores and grades of each segment were compared between internal and external endoscopist groups. Among 83 colonoscopy videos, interobserver reliability yielded a fair agreement between internal and external endoscopists, with a Cohen's Kappa coefficient of 0.39 (95% CI 0.08-0.70). Successful overall preparation rates (100% vs 98.8%) and high-quality preparation rates (88% vs 94%) were excellent and comparable between the two groups. In contrast, one case of unsuccessful cleansing (grade C) was documented, only in the external group. In addition, the interobserver reliability between internal and external endoscopists comparing each segmental score yielded a slight agreement in the descending colon (0.11, 95% CI - 0.02 to 0.23) and sigmoid colon (0.04, 95% CI - 0.18 to 0.26). Interobserver reliability between internal and external endoscopists using the HCS was fair. However, the HCS may be limited as a reliable scale for assessing bowel preparation quality in the left colon, where liquid is mainly retained in the left lateral position.

Measuring the observer (Hawthorne) effect on adenoma detection rates.

Background and study aims An independent observer can improve procedural quality. We evaluated the impact of the observer (Hawthorne effect) on important quality metrics during colonoscopies. Patients and Methods In a single-center comparative study, consecutive patients undergoing routine screening or diagnostic colonoscopy were prospectively enrolled. In the index group, all procedural steps and quality metrics were observed and documented, and the procedure was video recorded by an independent research assistant. In the reference group, colonoscopies were performed without independent observation. Colonoscopy quality metrics such as polyp, adenoma, serrated lesions, and advanced adenoma detection rates (PDR, ADR, SLDR, AADR) were compared. The probabilities of increased quality metrics were evaluated through regression analyses weighted by the inversed probability of observation during the procedure. Results We included 327 index individuals and 360 referents in the final analyses. The index group had significantly higher PDRs (62.4% vs. 53.1%, P =0.02) and ADRs (39.4% vs. 28.3%, P =0.002) compared with the reference group. The SLDR and AADR were not significantly increased. After adjusting for potential confounders, the ADR and SLDR were 50% (relative risk [RR] 1.51; 95%, CI 1.05-2.17) and more than twofold (RR 2.17; 95%, CI 1.05-4.47) more likely to be higher in the index group than in the reference group. Conclusions The presence of an independent observer documenting colonoscopy quality metrics and video recording the colonoscopy resulted in a significant increase in ADR and other quality metrics. The Hawthorne effect should be considered an alternative strategy to advanced devices to improve colonoscopy quality in practice.

Assisted documentation as a new focus for artificial intelligence in endoscopy: the precedent of reliable withdrawal time and image reporting.

BACKGROUND : Reliable documentation is essential for maintaining quality standards in endoscopy; however, in clinical practice, report quality varies. We developed an artificial intelligence (AI)-based prototype for the measurement of withdrawal and intervention times, and automatic photodocumentation. METHOD: A multiclass deep learning algorithm distinguishing different endoscopic image content was trained with 10 557 images (1300 examinations, nine centers, four processors). Consecutively, the algorithm was used to calculate withdrawal time (AI prediction) and extract relevant images. Validation was performed on 100 colonoscopy videos (five centers). The reported and AI-predicted withdrawal times were compared with video-based measurement; photodocumentation was compared for documented polypectomies. RESULTS: Video-based measurement in 100 colonoscopies revealed a median absolute difference of 2.0 minutes between the measured and reported withdrawal times, compared with 0.4 minutes for AI predictions. The original photodocumentation represented the cecum in 88 examinations compared with 98/100 examinations for the AI-generated documentation. For 39/104 polypectomies, the examiners' photographs included the instrument, compared with 68 for the AI images. Lastly, we demonstrated real-time capability (10 colonoscopies). CONCLUSION : Our AI system calculates withdrawal time, provides an image report, and is real-time ready. After further validation, the system may improve standardized reporting, while decreasing the workload created by routine documentation.