TYPE Original Research PUBLISHED 13 October 2022 DOI 10.3389/fmed.2022.1000726 Artificial intelligence to improve OPEN ACCESS polyp detection and screening time in colon capsule EDITED BY Zongxin Ling, Zhejiang University, China REVIEWED BY Konstantinos Triantafyllou, endoscopy National and Kapodistrian University of Athens, Greece Colm Antoine O. Morain, Pere Gilabert1*, Jordi Vitrià1 , Pablo Laiz1 , Trinity College Dublin, Ireland Carolina Malagelada2,3 , Angus Watson4 , Hagen Wenzek5 and *CORRESPONDENCE Pere Gilabert Santi Segui1 pere.gilabert@ub.edu 1 Departament de Matemàtiques i Informàtica, Universitat de Barcelona, Barcelona, Spain, 2 Digestive SPECIALTY SECTION System Research Unit, University Hospital Vall d’Hebron, Barcelona, Spain, 3 Department of This article was submitted to Medicine, Universitat Autònoma de Barcelona, Barcelona, Spain, 4 Department of Colorectal Surgery, Gastroenterology, Raigmore Hospital, NHS Highland, Inverness, United Kingdom, 5 CorporateHealth International ApS, a section of the journal Odense, Denmark Frontiers in Medicine RECEIVED 22 July 2022 ACCEPTED 23 September 2022 Colon Capsule Endoscopy (CCE) is a minimally invasive procedure which is PUBLISHED 13 October 2022 increasingly being used as an alternative to conventional colonoscopy. Videos CITATION Gilabert P, Vitrià J, Laiz P, recorded by the capsule cameras are long and require one or more experts’ Malagelada C, Watson A, Wenzek H time to review and identify polyps or other potential intestinal problems that and Segui S (2022) Artificial intelligence can lead to major health issues. We developed and tested a multi-platform to improve polyp detection and screening time in colon capsule web application, AI-Tool, which embeds a Convolution Neural Network (CNN) endoscopy. Front. Med. 9:1000726. to help CCE reviewers. With the help of artificial intelligence, AI-Tool is able to doi: 10.3389/fmed.2022.1000726 detect images with high probability of containing a polyp and prioritize them COPYRIGHT during the reviewing process. With the collaboration of 3 experts that reviewed © 2022 Gilabert, Vitrià, Laiz, Malagelada, Watson, Wenzek and 18 videos, we compared the classical linear review method using RAPID Reader Segui. This is an open-access article Software v9.0 and the new software we present. Applying the new strategy, distributed under the terms of the reviewing time was reduced by a factor of 6 and polyp detection sensitivity Creative Commons Attribution License (CC BY). The use, distribution or was increased from 81.08 to 87.80%. reproduction in other forums is permitted, provided the original KEYWORDS author(s) and the copyright owner(s) are credited and that the original colon capsule endoscopy, artificial intelligence, screening time, polyp detection, publication in this journal is cited, in colorectal cancer prevention accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. Introduction Colorectal cancer is the third most common type of cancer worldwide and ranks second on the list of most aggressive and deadly cancers (1). According to the Global Cancer Observatory, out of an estimated total of 1.9 million cases in 2020, this disease has caused the death of more than 935,000 people worldwide (1). This type of cancer has also been linked to unhealthy lifestyle habits such as smoking, alcohol consumption, unhealthy diets or sedentary lifestyle (2, 3). A clear relationship has been drawn between colon cancer and obesity (4, 5) along with genetic predisposition (6). One of the initial signs of the development of colon cancer is the appearance of polyps in the colon that grow in an uncontrolled manner (7). The detection of polyps when they are still small is crucial to prevent their Frontiers in Medicine 01 frontiersin.org Gilabert et al. 10.3389/fmed.2022.1000726 transformation into cancer. Screening programs are aimed to image frame to contain a polyp and a heatmap explaining detect early-stage cancer, improving the patient’s chances of the reasoning of the prediction. This heatmap, allows the survival (8, 9). With the disruption of colonoscopy screening expert to focus on the area of the image where the programs due to the COVID-19 pandemic, an increase in CNN suggests the polyp to be located. An experiment incidence of 0.2–0.9% and deaths of 0.6–1.6% is predicted over performed with 18 videos revised by 3 expert readers, the next 30 years (2020–2050) (10). To be able to mitigate has shown that the proposed system reduces screening these effects, well-resourced screening programs are urgently time significantly while it also increases the sensitivity needed. Unfortunately, not all healthcare systems can afford a of polyp detection. large increase in demand for colonoscopy which is usually a primary or secondary screening test. Therefore, other equally effective methods should be considered (10, 11). Among those alternatives, CCE has proven to be one of the most safe and effective tools, with an accuracy very similar to that of traditional Methods colonoscopies (12). It is less invasive for the patient (13), it does not cause discomfort during the procedure [only Study population minimal discomfort before the procedure due to required bowel preparation (13)] and no anesthesia is needed. However, the Eighteen videos of patients with at least one colon polyp reading of CCE videos is time-consuming and requires qualified obtained using the PillCam COLON 2 capsule (Medtronic) medical personnel (14–16). Further studies have compared CCE were randomly selected for this experiment following a Simple sensitivity and reading time using different RAPID Reader Randomization strategy. The data used in this study are viewing modes such as QuickView (17–19), as well reading retrospective CCE videos from patients that were conducted on performance of expert and non-expert physicians (20). For CCE behalf of the NHS Highland Raigmore Hospital in Inverness. to be considered as an alternative procedure in colorectal cancer All patients from this study came from referrals for symptoms detection, the use of AI could streamline the process without or were on surveillance lists within the Highlands and Islands compromising accuracy (21–23). area of Scotland and had a positive Fecal Immunochemical AI has been extensively applied to medical imaging Test (FIT). Referrals and final diagnoses were made based problems (24). In the colon capsule endoscopy field, multiple on local considerations, outside the influence of the teams methods have been presented to automatically detect ulcers (25), conducting CCE procedures. Bowel preparation in accordance polyps (26, 27), Crohn’s disease (28), bowel cleanliness (29) or with a standardized, PEG-based, split-dose cleansing protocol blood and mucosal lesions (30). These methods have shown was performed in all patients. All videos were obtained using promising results, but they are generally validated with limited PillCam COLON 2 which has two heads (front and rear). They data or biased datasets which does not guarantee a good were anonymized to protect patient information. Patients’ mean generalization in clinical practice (22, 31). To our knowledge, age was 58.1 ± 18.7 years (range, 18–92 years) and mean colon few attempts have been presented with similar evaluation transit time was 4 h 10 min (range, 0.17–14.2 h). to the one we propose in this paper. In Aoki et al. (32) Before the experiments began, the two videos obtained from a CNN-based method is presented to detect erosions and each of the heads of the capsule were meticulously reviewed ulcerations. In Beg et al. (33), a method to detect colonic by four independent CCE readers, experts from now on, in lesions is introduced comparing two reviewing procedures. order to create the ground truth for the experiments (gold There is no doubt that AI has potential benefits to both doctors standard). Each detected polyp was assigned a unique identifier, and patients, but its application to the clinical practice is the timestamp of the first and the last image where the polyp was challenging. AI is not yet at a point where it can completely visible, and from which head it was reported. The independent replace the intervention by human experts (22). Although analysis of the experts was then shared with all experts, reaching U.S. Food and Drug Administration (FDA) has approved a consensus in case of discrepancies. As experts, we have some assistance algorithms (34), no guidelines establishing arbitrarily considered CCE readers with at least 3 months of the role of AI currently exist. In order to achieve this, experience in CCE. All of them have formal training in reviewing these systems would need to gain confidence of medical CCE videos and they analyze about 5–20 videos a week. On a experts and ethical and regulatory issues would need to be daily basis, they follow the standard review protocol to ensure solved (15, 35). This is why it is important to develop systems that each video is reviewed in a consistent, repeatable and well- that cooperate with the experts, facilitating their decision documented manner. The results of all of them were validated making. by a medical doctor with 2 years of experience who created In this paper, we present a novel CNN-based system, a final report about the results. In no case were any concerns AI-Tool, to assist physicians with the detection of colonic reported back by that clinician either from the review nor from polyps. Given a video, it outputs a probability score per any possibly follow-up procedure about the quality of the report. Frontiers in Medicine 02 frontiersin.org Gilabert et al. 10.3389/fmed.2022.1000726 Both, the final report and the gold standard used as ground- TABLE 1 Detection of polyps distinguishing by size, visibility and morphology. truth for the experiment in this study are the responsibility of the medical doctor that approved the results. #Polyps RAPID AI-Tool During this process, a total of 52 unique polyps were found. The video with the most polyps had 7, while Size Small (<6 mm) 29 76.92% 85.42% there were 5 videos with only one polyp. The polyps’ size Large (≥6 mm) 23 85.71% 91.18% was estimated with RAPID. A total of 23 polyps were Visibility Low (<4 frames) 9 58.33% 80.00% identified as large (≥6 mm) and 29 polyps as small (<6 Normal (4 − 10 frames) 15 73.91% 81.82% mm). There were 5 polyps larger than 10 mm and only High (>10 frames) 28 92.31% 93.33% one polyp smaller than 3 mm. The characterization of Morphology Pedunculated 4 100.00% 100.00% morphologies was done in accordance with the requirements Sessile 25 81.82% 92.86% from the referring clinicians, matching standard Paris Flat 23 75.76% 80.56% classifications where possible. In no case polyps were 52 81.08% 87.80% selected or discarded based on size or morphology, all polyps reported by the physicians were included in the study. that each video is analyzed in a consistent, repeatable way and Experimental design documented according to common standards. None of these three readers were part of the gold standard creation process. Three experienced CCE readers reviewed the videos selected for this experiment. Each reader reviewed half of the videos using the standard RAPID Reader Software v9.0 (Medtronic) Screening tool: RAPID reader v9.0 and the other half using the AI-Tool. Results obtained by the experts using each of the tools are reported in terms of number of RAPID Reader v9.0 is a Medtronic proprietary software, polyps detected (sensitivity) as well as time needed to complete used to review and interpret wireless capsule endoscopy videos. the reviews (screening time). This software is widely used by colon capsule endoscopy The experiment conducted in this study was restricted to the professionals designed to review the video in a comfortable way. images of the colon. Identification of the entrance and exit of The video is shown in temporal order and the app allows to label the colon was previously provided to the readers. When they images, to add comments to images and to measure any part of used RAPID software, they were asked to perform the standard the image with an integrated ruler function. The user can change screening procedure without any screening time limitation. No the replay speed of the video at any time and the application information other than a single video identifier was provided to shows the approximate area of the abdomen where the capsule the reviewers during the analysis of the videos. For the AI-Tool, is located. It also warns the user if the capsule is passing through the review time was limited to 30 min regardless of the video’s an area at high speed, so the user can know if special attention is length. required. RAPID has several modes to review videos: single view For both tools, readers were required to review the with the frontal camera, twin head mode showing both cameras videos without pauses or external stimuli that could lead simultaneously (Figure 1) and collage mode which shows a batch to distractions. During the review, the readers labeled the of frames selected by traditional computer vision techniques. images that they identified as a polyp using the tools provided within each of the applications. In the case of RAPID, experts were asked to tag all the unique polyps they found. Screening tool: AI-Tool When using the AI-Tool, experts were asked to make a decision, polyp, clear or other, for each sequence that was The AI-Tool is a software designed to assist clinicians in the presented to them. detection of polyps, by complementing any proprietary video reviewing software, such as RAPID. It embeds a Convolutional Neural Network (CNN) into a web tool that presents images CCE readers with potential polyps to the user in a sequence of declining certainty. Therefore, images that are very likely to contain a The expert readers are endoscopy nurses with at least 2 polyp will appear first. At the time we started the study, the years of experience with CCE. They have a formal CCE training CNN from Laiz et al. (26) was the state of the art for polyp and conduct between 5 and 20 video analyses per week. On a detection so it was chosen among other candidates (36–38) as daily basis, they follow a standard operating procedure to ensure the core of the AI-Tool. The hyperparameters of the model Frontiers in Medicine 03 frontiersin.org Gilabert et al. 10.3389/fmed.2022.1000726 FIGURE 1 RAPID Reader Software v9.0a : screen with images from both camera heads (green/yellow) and marked thumbnails. a https://www.medtronic. com/covidien/en-gb/products/capsule-endoscopy/pillcam-software.html. FIGURE 2 Candidate polyp sequence displayed in the AI-Tool. Each colored bar shows the probabilities for a polyp in one head of the capsule. The proposed image is presented in the center frame and 4 context images are placed by each side. were fixed after a 5-Fold cross validation process using 120 image zones that allow the CNN to classify an image as CCE videos (2,080 polyp images and 246 k negative images). polyp. Different hyperparameters were tested (the same for the five Each potential polyp image is displayed along models) and those that gave the best results in the validation with eight context frames, the four preceding and the sets were selected. A single model was then trained using the four following it (Figure 2). For each frame, a colored 120 CCE and embedded into the AI-Tool. The experimental square is shown to indicate the probability of it being validation of the network has shown a sensitivity over 90% a polyp using a colorblind friendly palette that can be at a specificity of 95% when evaluated in a fully automatic customized when the application starts. Each image can setting (when no expert is involved) using full videos. All 120 be enlarged by clicking on it, then further information videos used in the training of the CNN were excluded for this is presented such as the probability or the timestamp of experiment. the sequence. The image sequence can be also displayed The AI-Tool computes two outputs using a CNN: a as a video using the left and right keyboard arrows. probability score per frame to contain a polyp and a heatmap Heatmaps can always be activated showing the most to visualize the reasoning behind the score using CAM (39), likely area to contain a polyp. A further benefit of an algorithm that uses the values of the latest CNN layers the heatmap is that it helps readers to understand the to display the image areas most relevant for classification. In reasoning behind the polyp probability and as a consequence, this particular case, this method presents the most relevant increase their trust in the system. Frontiers in Medicine 04 frontiersin.org Gilabert et al. 10.3389/fmed.2022.1000726 Results The average time required for the experiments performed with RAPID was 47.11 min (11.6 min for each hour of CCE video Polyp detection reviewed) with a maximum of 126 min. Let us recall that the time for analysis using the AI-Tool was fixed at 30 min for all The overall sensitivity of polyp detection using RAPID the experiments. as the screening procedure was 81.08% while using the AI- Figure 3 shows the average sensitivity curve as a function Tool the sensitivity increased to 87.80%. Table 1 shows the of time for both applications. AI-Tool took only 8.00 min to percentage of polyps found using both tools distinguishing reach the same accuracy as RAPID (point A). Since the mean of between three categories: polyp size (in millimeters), visibility RAPID experiments was 47.11 min we can state that the AI-Tool (in number of frames) and morphology. The sensitivity using reduces the time needed to reach the same accuracy as RAPID RAPID turned out to be 76.92% for polyps smaller than 6 mm by a factor of 47.11/8.00 ≈ 6. We can also see that RAPID and 85.71% for larger polyps. Both numbers increased when experiments reached a 55.41% of sensitivity (point B) by minute the AI-Tool was used (85.42 and 91.18% for small and large 30 when the AI-Tool experiments finished. polyps, respectively). The shape of the curves is also an aspect worth considering. The biggest difference between both tools was observed in While the RAPID curve has an almost linear behavior, the AI- the visibility of the polyp. For polyps appearing in a few frames Tool curve shows an initial steep slope and, after minute 16, it (low visibility), the table shows a significant improvement when is almost flat. In the first 10 min of the analysis 84.14% of the using the AI-Tool. While RAPID achieved an accuracy of polyps are detected. In the next 10 min, this number rises to 58.33%, the AI-Tool reached 80.00%. This represents an increase 86.59%, which represents an increase of 2.45% in this period. of 21.67% in this category. Smaller improvements using the AI- Finally, only 1.21% of the polyps are detected in the last 10 min of Tool were also reported for polyps appearing in a larger number the experiment. This indicates that our application is proposing of frames. the relevant images in the first minutes of visualization. It is also These results show that small polyps and polyps that appear worth mentioning that without limiting the time to 30 min as for only a few frames are more likely to be detected using the we did, the detection of polyps may slightly increase because AI-Tool than using the RAPID application. those with a very low score would have been presented to the reviewers. However, this was not the objective of our study, since we aimed to see if the video review could be done better and in a CCE screening time shorter period of time. One of the aims of this study was to compare the time needed for the detection of polyps using RAPID and our AI-Tool. Qualitative results Heatmaps are a key element of our tool. They allow the medical staff to trust the system and make an informed decision on each image sequence. Figure 4 shows the heatmaps activation for three different categories. First, in the image on the left, we see high-scoring polyps. We observe how the heatmaps generated by the CAM algorithm are well defined and show the area where the polyp is located. In the center, we see polyps with a very low score, which the network erroneously classifies as negative. In this case, the heatmaps are not activated. Finally, in the image on the right we can see images that do not contain any polyp but to which the system assigns a high score. It can also be seen that the false positive patterns correspond to textures and morphologies compatible with polyps. In almost all of them a rounded area is displayed which, without the context of the other images, can be difficult to classify as a polyp FIGURE 3 image or not. In addition, this is a valuable information for the Mean sensitivity curve of the experiments using both applications. In green the Super-Expert curve (gold standard) reviewers of the video, as the heatmap shows the area on which that represents the maximum value that the blue line could it is important to focus their attention. reach. This curve has been calculated simulating an expert who We now focus on showing the images of polyps that have not never makes mistakes when identifying a polyp while using the AI-Tool. been detected with either of the two applications (specificity). Figure 5 shows those polyps not detected with the AI-Tool Frontiers in Medicine 05 frontiersin.org Gilabert et al. 10.3389/fmed.2022.1000726 FIGURE 4 Left: Images of correctly identified polyps with their respective heatmaps (True Positives). Center: Images of polyps with a very low score (False Negatives). Right: Images that do not contain a polyp but still have a very high score (False Positives). FIGURE 5 Polyp frames to which the app has attributed a small score and, therefore, none of the experts have been able to review in the first 30 min. Polyps are circled in white. because of the imposed time restriction (30 min). The score given to these image does not exceed 15%, therefore, the experts never reviewed them. In fact, these four images are the polyps with the lowest score of this study. The polyps of these images are difficult to find since they are partially occluded or do not present a regular morphology. Figure 6 shows images of polyps reviewed and discarded by all the experts even though the AI-Tool assigned them a remarkably high probability. These missed polyps are the result of human error or discrepancies between the video reviewer and the experts who generated the ground truth. Finally, Figure 7 shows some examples of polyp images FIGURE 6 missed in RAPID experiments but correctly detected using the Images that all experts have reviewed and found not to be AI-Tool. Due to the size of the polyp and the fast movement polyps. Polyps are circled in white. of the capsule that took few images, these polyps are especially difficult to find using RAPID. In contrast, they are easy to find for AI-Tool users as they are presented with the clearest image. experts. redWe consider that these improvements further improve the value proposition of CCE as a clinically viable Conclusions alternative to traditional imaging methods of the gastrointestinal tract. In this paper, thanks to the AI enhancement, we mitigate The proposed method, AI-Tool, uses the output of a one of the main drawbacks of CCE, the required time CNN architecture. It scores frames based on the probability for the analysis while increasing the detection rate of the of containing a polyp and then it reorders the video images Frontiers in Medicine 06 frontiersin.org Gilabert et al. 10.3389/fmed.2022.1000726 FIGURE 7 Example of polyps missed in RAPID experiments. Polyp frames are tagged with a red square. Polyps are circled in white. to present the most relevant ones first. The validation was previous versions of the manuscript, study conception, design, performed by three clinical experts that analyzed 18 videos, read, and approved the final manuscript. comparing the standard method with and without the proposed AI-enhanced application. With the assistance of the AI-Tool, Funding the time required to review the videos was reduced by a factor of 6 and the sensitivity increased from 81.08 to 87.80%. In the This work has been also supported by MINECO Grant case of small polyps (<6 mm), the improvement in sensitivity RTI2018-095232-B-C21, SGR 1742, Instituto de Salud Carlos III obtained by the AI-Tool was 8.50%, and, for polyps with low as well as the Innovate UK project 104633. visibility (seen in <3 frames), the improvement in detection was 21.67%. Acknowledgments Data availability statement The authors would like to thank the team from CorporateHealth International ApS for their feedback. The data used during the present study are not publicly available as they are property of National Conflict of interest Services Scotland (NHS Highland) but are available through co-author AW (angus.watson@nhs.scot) upon Author HW is co-founder of CorporateHealth International, reasonable request. a company that may be affected by the research reported in the enclosed manuscript. The remaining authors declare that the research was Ethics statement conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of The studies involving human participants were reviewed interest. and approved by University of Barcelona’s Bioethics Commission, Institutional Review Board IRB00003099. The patients/participants provided their written informed consent to Publisher’s note participate in this study. All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated Author contributions organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or Data collection was performed by HW. The first draft of claim that may be made by its manufacturer, is not guaranteed the manuscript was written by PG. All authors commented on or endorsed by the publisher. Frontiers in Medicine 07 frontiersin.org Gilabert et al. 10.3389/fmed.2022.1000726 References 1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. of a novel software algorithm. Int J Colorectal Dis. (2019) 34:1857–63. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality doi: 10.1007/s00384-019-03393-0 worldwide for 36 cancers in 185 countries. CA Cancer J Clin. (2021) 71:209–49. 21. Byrne MF, Donnellan F. Artificial intelligence and capsule endoscopy: doi: 10.3322/caac.21660 is the truly “smart” capsule nearly here? Gastrointest Endosc. (2019) 89:195–7. 2. O’Sullivan DE, Metcalfe A, Hillier TW, King WD, Lee S, Pader doi: 10.1016/j.gie.2018.08.017 J, et al. Combinations of modifiable lifestyle behaviours in relation to 22. Yang YJ. The future of capsule endoscopy: the role of artificial intelligence and colorectal cancer risk in Alberta’s Tomorrow Project. Sci Rep. (2020) 10:1–11. other technical advancements. Clin Endosc. (2020) 53:387. doi: 10.5946/ce.2020.133 doi: 10.1038/s41598-020-76294-w 23. Chetcuti Zammit S, Sidhu R. Capsule endoscopy-Recent developments 3. Ahmed M. Colon cancer: a clinician’s perspective in 2019. Gastroenterol Res. and future directions. Expert Rev Gastroenterol Hepatol. (2021) 15:127–37. (2020) 13:1. doi: 10.14740/gr1239 doi: 10.1080/17474124.2021.1840351 4. Ye P, Xi Y, Huang Z, Xu P. Linking obesity with colorectal 24. Topol EJ. High-performance medicine: the convergence of human and cancer: epidemiology and mechanistic insights. Cancers. (2020) 12:1408. artificial intelligence. Nat Med. (2019) 25:44–56. doi: 10.1038/s41591-018-0300-7 doi: 10.3390/cancers12061408 25. Rahim T, Usman MA, Shin SY. A survey on contemporary computer- 5. Lauby-Secretan B, Scoccianti C, Loomis D, Grosse Y, Bianchini F, Straif K. aided tumor, polyp, and ulcer detection methods in wireless capsule Body fatness and cancer–viewpoint of the IARC working group. N Engl J Med. endoscopy imaging. Comput Med Imaging Graphics. (2020) 85:101767. (2016) 375:794–8. doi: 10.1056/NEJMsr1606602 doi: 10.1016/j.compmedimag.2020.101767 6. Valle L. Genetic predisposition to colorectal cancer: where we 26. Laiz P, Vitrià J, Wenzek H, Malagelada C, Azpiroz F, Seguí S. WCE polyp stand and future perspectives. World J Gastroenterol. (2014) 20:9828–49. detection with triplet based embeddings. Comput Med Imaging Graphics. (2020) doi: 10.3748/wjg.v20.i29.9828 86:101794. doi: 10.1016/j.compmedimag.2020.101794 7. Bond J. Colon polyps and cancer. Endoscopy. (2003) 35:27–35. 27. Jin EH, Lee D, Bae JH, Kang HY, Kwak MS, Seo JY, et al. Improved doi: 10.1055/s-2003-36410 accuracy in optical diagnosis of colorectal polyps using convolutional neural 8. Levin TR, Corley DA, Jensen CD, Schottinger JE, Quinn VP, Zauber AG, et al. networks with visual explanations. Gastroenterology. (2020) 158:2169–79.e8. Effects of organized colorectal cancer screening on cancer incidence and mortality doi: 10.1053/j.gastro.2020.02.036 in a large community-based population. Gastroenterology. (2018) 155:1383–91. 28. Goran L, Negreanu AM, Stemate A, Negreanu L. Capsule endoscopy: current doi: 10.1053/j.gastro.2018.07.017 status and role in Crohn’s disease. World J Gastrointest Endosc. (2018) 10:184. 9. Loveday C, Sud A, Jones ME, Broggio J, Scott S, Gronthound F, et al. doi: 10.4253/wjge.v10.i9.184 Prioritisation by FIT to mitigate the impact of delays in the 2-week wait colorectal 29. Noorda R, Nevárez A, Colomer A, Beltrán VP, Naranjo V. cancer referral pathway during the COVID-19 pandemic: a UK modelling study. Automatic evaluation of degree of cleanliness in capsule endoscopy Gut. (2021) 70:1053–60. doi: 10.1136/gutjnl-2020-321650 based on a novel CNN architecture. Sci Rep. (2020) 10:1–13. doi: 10. 10. de Jonge L, Worthington J, van Wifferen F, Iragorri N, Peterse EF, Lew 1038/s41598-020-74668-8 JB, et al. Impact of the COVID-19 pandemic on faecal immunochemical test- 30. Mascarenhas M, Ribeiro T, Afonso J, Ferreira JP, Cardoso H, Andrade P, et based colorectal cancer screening programmes in Australia, Canada, and the al. Deep learning and colon capsule endoscopy: automatic detection of blood and Netherlands: a comparative modelling study. Lancet Gastroenterol Hepatol. (2021) colonic mucosal lesions using a convolutional neural network. Endoscopy Int Open. 6:304–14. doi: 10.1016/S2468-1253(21)00003-0 (2022) 10:E171–7. doi: 10.1055/a-1675-1941 11. Laghi L, Cameletti M, Ferrari C, Ricciardiello L. Impairment of colorectal cancer screening during the COVID-19 pandemic. Lancet Gastroenterol Hepatol. 31. Saurin JC, Beneche N, Chambon C, Pioche M. Challenges and (2021) 6:425–6. doi: 10.1016/S2468-1253(21)00098-4 future of wireless capsule endoscopy. Clin Endosc. (2016) 49:26. doi: 10. 5946/ce.2016.49.1.26 12. Vuik FE, Nieuwenburg SA, Moen S, Spada C, Senore C, Hassan C, et al. Colon capsule endoscopy in colorectal cancer screening: a systematic review. Endoscopy. 32. Aoki T, Yamada A, Aoyama K, Saito H, Fujisawa G, Odawara N, et al. (2021) 53:815–24. doi: 10.1055/a-1308-1297 Clinical usefulness of a deep learning-based system as the first screening on small-bowel capsule endoscopy reading. Digest Endoscopy. (2020) 32:585–91. 13. Ismail MS, Murphy G, Semenov S, McNamara D. Comparing colon doi: 10.1111/den.13517 capsule endoscopy to colonoscopy; a symptomatic patient’s perspective. BMC Gastroenterol. (2022) 22:1–8. doi: 10.1186/s12876-021-02081-0 33. Beg S, Wronska E, Araujo I, Suárez BG, Ivanova E, Fedorov E, et al. Use of rapid reading software to reduce capsule endoscopy reading 14. Maieron A, Hubner D, Blaha B, Deutsch C, Schickmair T, Ziachehabi A, et times while maintaining accuracy. Gastrointest Endosc. (2020) 91:1322–7. al. Multicenter retrospective evaluation of capsule endoscopy in clinical routine. doi: 10.1016/j.gie.2020.01.026 Endoscopy. (2004) 36:864–8. doi: 10.1055/s-2004-825852 34. Muehlematter UJ, Daniore P, Vokinger KN. Approval of artificial intelligence 15. Koulaouzidis A, Dabos K, Philipper M, Toth E, Keuchel M. and machine learning-based medical devices in the USA and Europe (2015- How should we do colon capsule endoscopy reading: a practical guide. 20): a comparative analysis. Lancet Digital Health. (2021) 3:e195-e203. doi: 10. Therapeutic Adv Gastrointestinal Endoscopy. (2021) 14:26317745211001983. 1016/S2589-7500(20)30292-2 doi: 10.1177/26317745211001983 35. Ahmad OF, Stoyanov D, Lovat LB. Barriers and pitfalls for artificial 16. Rondonotti E, Pennazio M, Toth E, Koulaouzidis A. How to read small bowel intelligence in gastroenterology: ethical and regulatory issues. Tech Innovat capsule endoscopy: a practical guide for everyday use. Endoscopy Int Open. (2020) Gastrointest Endoscopy. (2020) 22:80–4. doi: 10.1016/j.tgie.2019.150636 8:E1220–4. doi: 10.1055/a-1210-4830 17. Farnbacher MJ, Krause HH, Hagel AF, Raithel M, Neurath MF, Schneider 36. Yuan Y, Meng MQH. Deep learning for polyp recognition in T. QuickView video preview software of colon capsule endoscopy: reliability wireless capsule endoscopy images. Med Phys. (2017) 44:1379–89. doi: 10. in presenting colorectal polyps as compared to normal mode reading. Scand J 1002/mp.12147 Gastroenterol. (2014) 49:339–46. doi: 10.3109/00365521.2013.865784 37. Yuan Y, Qin W, Ibragimov B, Zhang G, Han B, Meng MQH, et al. Densely 18. Günther U, Daum S, Zeitz M, Bojarski C. Capsule endoscopy: comparison connected neural network with unbalanced discriminant and category sensitive of two different reading modes. Int J Colorectal Dis. (2012) 27:521–5. constraints for polyp recognition. IEEE Trans Automat Sci Eng. (2019) 17:574–83. doi: 10.1007/s00384-011-1347-9 doi: 10.1109/TASE.2019.2936645 19. Shiotani A, Honda K, Kawakami M, Kimura Y, Yamanaka Y, Fujita 38. Guo X, Yuan Y. Triple ANet: Adaptive abnormal-aware attention network M, et al. Analysis of small-bowel capsule endoscopy reading by using for WCE image classification. In: International Conference on Medical Image Quickview mode: training assistants for reading may produce a high diagnostic Computing and Computer-Assisted Intervention. Shenzhen: Springer (2019). pp. yield and save time for physicians. J Clin Gastroenterol. (2012) 46:e92–5. 293–301. doi: 10.1097/MCG.0b013e31824fff94 39. Zhou B, Khosla A, Lapedriza A, Oliva A, Torralba A. Learning deep features 20. Hausmann J, Linke JP, Albert JG, Masseli J, Tal A, Kubesch A, for discriminative localization. In: Proceedings of the IEEE Conference on Computer et al. Time-saving polyp detection in colon capsule endoscopy: evaluation Vision and Pattern Recognition. Las Vegas, NV: IEEE (2016). p. 2921–9. Frontiers in Medicine 08 frontiersin.org