Search

A typical capsule endoscopy (CE) case generates tens of thousands of images, with abnormalities often confined to a just few frames. Omni Mode is a novel EndoCapsule software algorithm (Olympus, Tokyo, Japan) that proposes to intelligently remove duplicate images while maintaining accuracy in lesion detection.This prospective multicenter study took place across 9 European centers. Consecutive, unselected CE cases were read conventionally in normal mode, with every captured frame reviewed. Cases were subsequently anonymized and randomly allocated to another center where they were read using Omni Mode. Detected lesions and reading times were recorded, with findings compared between both viewing modes. The clinical significance of lesions was described according to the P classification (P0, P1, and P2). Where a discrepancy in lesion detection in either mode was found, expert blinded review at a consensus meeting was undertaken.The patient population undergoing CE had a mean age of 49.5 years (range, 18-91), with the investigation of anemia or GI bleeding accounting for 71.8% of cases. The average small-bowel transit time was 4 hours, 26 minutes. The mean reading time in normal mode was 42.5 minutes. The use of Omni Mode was significantly faster (P .0001), with an average time saving of 24.6 minutes (95% confidence interval, 22.8-26.9). The 2127 lesions were identified and classified according to the P classification as P0 (1234), P1 (656), and P2 (237). Lesions were identified using both reading modes in 40% (n = 936), and 1186 lesions were identified by either normal or Omni Mode alone. Normal mode interpretation was associated with 647 lesions being missed, giving an accuracy of .70. Omni Mode interpretation led to 539 lesions being missed, with an accuracy of .75. There was no significant difference in clinical conclusions made between either reading mode.This study shows that CE reading times can be reduced by an average of 40%, without any reduction in clinical accuracy.

Background and Aims Capsule endoscopy (CE) is an established, noninvasive modality for examining the small bowel. Minimum training requirements are based primarily on guidelines and expert opinion. A validated tool to assess the competence of CE is lacking. In this prospective, multicenter study, we determined the minimum number of CE procedures required to achieve competence during gastroenterology fellowship; validated a capsule competency test (CapCT); and evaluated any correlation between CE competence and endoscopy experience. Methods We included second- and third-year gastroenterology fellows from 3 institutions between 2013 and 2018 in a structured CE training program with supervised CE interpretation. Fellows completed the CapCT with a maximal score of 100. For comparison, expert faculty completed the same CapCT. Trainee competence was defined as a score ≥90% compared with the mean expert score. Fellows were tested after 15, 25, and 35 supervised CE interpretations. CapCT was validated using expert consensus and item analysis. Data were collected on the number of previous endoscopies. Results A total of 68 trainees completed 102 CapCTs. Fourteen CE experts completed the CapCT with a mean score of 94. Mean scores for fellows after 15, 25, and 35 cases were 83, 86, and 87, respectively. Fellows with at least 25 interpretations achieved a mean score ≥84 in all 3 institutions. CapCT item analysis showed high interobserver agreement among expert faculty (k = 0.85). There was no correlation between the scores and the number of endoscopies performed. Conclusion After a structured CE training program, gastroenterology fellows should complete a minimum of 25 supervised CE interpretations before assessing competence using the validated CapCT, regardless of endoscopy experience.

Background and Aims Acute upper GI bleeding is common and requires investigation with EGD, but endotherapy is not always necessary. Magnetically assisted capsule endoscopy (MACE) uses a capsule steerable by an external magnet and allows examination of the upper GI tract and small bowel, but its role in acute upper GI bleeding has not been assessed. Methods We conducted a prospective cohort study comparing the diagnostic yield of MACE and EGD in patients with suspected acute upper GI bleeding. Patient tolerance, mucosal visibility by MACE, and frequency of small-bowel bleeding were assessed. Whether or not MACE could safely predict discharge of patients was also determined. Results Thirty-three patients were included for analysis (median age, 60 years; 75.8% male). MACE detected more focal lesions (peptic, vascular, and fresh/altered blood without a clear source) than EGD (40 versus 25, respectively, P = .02) but statistical significance was not reached for significant lesions (considered to be the bleeding source; 14 vs 13, respectively, P = 1). Capsule endoscopy identified an additional cause for bleeding in the small bowel in 18%. Visualization by MACE was excellent in most areas; views of the esophagus, gastroesophageal junction, fundus, and duodenal bulb were suboptimal. MACE was better tolerated than unsedated EGD and correctly identified patients who were safe for discharge. Conclusions MACE had higher diagnostic yield for focal lesions and was better tolerated than EGD. It also correctly predicted safe discharge for patients with acute upper GI bleeding. (Clinical trials registration number: NCT02690376.)

Background and Aims Video capsule endoscopy (VCE) is limited by poor image quality and incomplete small-bowel transit. This study was designed to evaluate the diving method for VCE in the examination of small-intestinal disease. Methods From July 2017 to September 2017, eligible patients were randomly assigned to 2 groups, the diving group and the control group. For the diving group, 500 mL of water was administered every hour when the capsule reached the small bowel. The primary outcomes were image quality and positive findings. Secondary outcomes were the completion rate of examination, gastric transit time (GTT), small-bowel transit time (SBTT), lesion detection rate, adverse events, and patient satisfaction. Results One hundred forty patients were included. The scores of endoscopic images in the proximal third and middle third of the small bowel in the diving group were significantly higher than that in the control group (3.47 ± .60 vs 3.11 ± .63 [P = .007] and 3.24 ± .59 vs 2.78 ± .74 [P = .002], respectively). The positive findings in the distal third of the small bowel were significantly different between the 2 groups (P = .005). The completion rate in the diving group was significantly higher (92.19% vs 76.32%, respectively; P = .012). The GTT, SBTT, and lesion detection rate were similar in 2 groups (P = .282, .067, and .577, respectively). No discomfort or adverse events were reported except for a few cases of frequent urination. Conclusions The diving method for VCE examination effectively improves the endoscopic view in the proximal and middle thirds of the small bowel and the positive findings in the distal small intestine and increases the completion rate. (Clinical trial registration number: ChiCTR-RDR-17011823.)

Background and Aims Magnetically controlled capsule endoscopy (MCE) has become an efficient diagnostic modality for gastric diseases. We developed a novel automatic gastric lesion detection system to assist in diagnosis and reduce inter-physician variations. This study aimed to evaluate the diagnostic capability of the computer-aided detection system for MCE images. Methods We developed a novel automatic gastric lesion detection system based on a convolutional neural network (CNN) and faster region-based convolutional neural network (RCNN). A total of 1,023,955 MCE images from 797 patients were used to train and test the system. These images were divided into 7 categories (erosion, polyp, ulcer, submucosal tumor, xanthoma, normal mucosa, and invalid images). The primary endpoint was the sensitivity of the system. Results The system detected gastric focal lesions with 96.2% sensitivity (95% confidence interval [CI], 95.7%-96.5%), 76.2% specificity (95% CI, 75.97%-76.3%), 16.0% positive predictive value (95% CI, 15.7%-16.3%), 99.7% negative predictive value (95% CI, 99.74%-99.79%), and 77.1% accuracy (95% CI, 76.9%-77.3%) (sensitivity was 99.3% for erosions; 96.5% for polyps; 89.3% for ulcers; 87.2% for submucosal tumors; 90.6% for xanthomas; 67.8% for normal; and 96.1% for invalid images). Analysis of the receiver operating characteristic curve showed that the area under the curve for all positive images was 0.84. Image processing time was 44 milliseconds per image for the system and 0.38 ± 0.29 seconds per image for clinicians (P Conclusions The CNN faster-RCNN-based diagnostic program system showed good performance in diagnosing gastric focal lesions in MCE images.

Delayed gastric transit of the capsule may lead to incomplete small bowel examination, reducing the diagnostic yield. Thus, this study was designed to determine if magnetic steering could enhance capsule gastric emptying and mucosal visualization within the duodenum.The intervention group comprised 100 patients undergoing magnetic-controlled capsule endoscopy between May to September 2017 in whom magnetic control was used to assist transpyloric passage of the capsule and duodenal inspection. A cohort of 100 patients who had undergone the procedure before May 2017 was randomly selected from the database as an historic control group in whom transpyloric movement of the capsule occurred spontaneously (without magnetic assistance). The difference in the pyloric transit time (PTT) and duodenal papilla detection rate (DPDR) between the 2 groups were compared, and related factors were also investigated.Transpyloric passage of the capsule under magnetic control was successfully performed in 59 patients (59%). Median PTT was greatly reduced in the intervention group from 58.38 minutes (range, 13.45-87.47) to 4.69 minutes (range, 1.56-55.00; P .001), and DPDR was also greatly improved with magnetic steering (30.5% vs 9%, P .001). Magnetic steering, male gender, and higher body mass index were independently associated with reduced gastric transit time and magnetic steering with an enhanced DPDR.Magnetic steering of the capsule can enhance gastric emptying of the capsule and may prove useful in nonobese and female patients who appeared to have longer gastric transit time and achieved a better DPDR than that under the action of peristalsis alone. (Clinical trial registration number: NCT03441945.).

Background and Aims Controversy remains regarding the type and amount of precapsule bowel cleansing required for small-bowel video capsule endoscopy (VCE). This study aims to assess the efficacy and tolerance of 2 active preparations and a control group of clear fluids only. Methods Patients with clinical indications for VCE were randomized to (1) clear fluids only the evening before VCE, (2) 2 sachets of sodium picosulfate plus magnesium sulfate (P/MC) the evening before, or (3) 2 L of polyethylene glycol (PEG) the evening before. Diet instructions were the same for all 3 groups. Small-bowel cleansing was assessed in 3 ways: a 5-point ordinal scale (primary outcome), the percentage of time the small-bowel view was clear, and a validated computerized assessment of cleansing. Results In total, 198 patients were randomized and 175 patients completed the trial with a mean age of 49.2 years. There was no clear benefit of active preparation with either P/MC or PEG over clear fluids only in the overall 5-point rating scale or in the distal fourth of each examination. There was no difference in diagnostic yield between groups. Significant differences were seen concerning tolerance of the preparations, with a higher proportion rating it as easy or very easy in the clear fluids–only group (93%) and the P/MC group (67%) than in the PEG group (13%) ( P Conclusions Small-bowel cleansing for VCE remains a controversial topic. This randomized control trial demonstrates no benefit in overall or distal small-bowel visualization with active preparation using either PEG or P/MC compared with clear fluids only. (Clinical trial registration number: NCT00677794.)

Background and Aims Diagnosis of GI ulcers and/or hemorrhage by wireless capsule endoscopy (WCE) is limited by the physician-dependent, tedious, time-consuming process of image and/ or video classification. Computer-aided diagnosis (CAD) by convolutional neural network (CNN)-based machine learning may help reduce this burden. Our aim was to conduct a meta-analysis and appraise the reported data. Methods Multiple databases were searched (from inception to November 2019), and studies that reported on the performance of CNN in the diagnosis of GI ulcerations and/or hemorrhage on WCE were selected. A random-effects model was used to calculate the pooled rates. In cases where multiple 2 × 2 contingency tables were provided for different thresholds, we assumed the data tables were independent from each other. Heterogeneity was assessed by I2% and 95% prediction intervals. Results Nine studies were included in our final analysis that evaluated the performance of CNN-based CAD of GI ulcers and/or hemorrhage by WCE. The pooled accuracy was 95.4% (95% confidence interval [CI], 94.3-96.3), sensitivity was 95.5% (95% CI, 94-96.5), specificity was 95.8% (95% CI, 94.7-96.6), positive predictive value was 95.8% (95% CI, 90.5-98.2), and negative predictive value was 96.8% (95% CI, 94.9-98.1). I2% heterogeneity was negligible except for the pooled positive predictive value. Conclusions Based on our meta-analysis, CNN-based CAD of GI ulcerations and/or hemorrhage on WCE achieves a high-level performance. The quality of the evidence is robust, and therefore CNN-based CAD has the potential to become the first choice of machine learning to optimize WCE image/video reading.

A deep convolutional neural network (CNN) system could be a high-level screening tool for capsule endoscopy (CE) reading but has not been established for targeting various abnormalities. We aimed to develop a CNN-based system and compare it with the existing QuickView mode in terms of their ability to detect various abnormalities.We trained a CNN system using 66,028 CE images (44,684 images of abnormalities and 21,344 normal images). The detection rate of the CNN for various abnormalities was assessed per patient, using an independent test set of 379 consecutive small-bowel CE videos from 3 institutions. Mucosal breaks, angioectasia, protruding lesions, and blood content were present in 94, 29, 81, and 23 patients, respectively. The detection capability of the CNN was compared with that of QuickView mode.The CNN picked up 1,135,104 images (22.5%) from the 5,050,226 test images, and thus, the sampling rate of QuickView mode was set to 23% in this study. In total, the detection rate of the CNN for abnormalities per patient was significantly higher than that of QuickView mode (99% vs 89%, P .001). The detection rates of the CNN for mucosal breaks, angioectasia, protruding lesions, and blood content were 100% (94 of 94), 97% (28 of 29), 99% (80 of 81), and 100% (23 of 23), respectively, and those of QuickView mode were 91%, 97%, 80%, and 96%, respectively.We developed and tested a CNN-based detection system for various abnormalities using multicenter CE videos. This system could serve as an alternative high-level screening tool to QuickView mode.

Patients presenting with nonhematemesis GI bleeding (NHGIB) represent a diagnostic challenge for physicians. We performed a randomized controlled trial to assess the benefits of deployment of a video capsule soon after admission in the management of patients presenting with melena, hematochezia, or severe anemia compared with standard of care management.Patients admitted with NHGIB were randomized and placed into 1 of 2 study groups. In the experimental group, patients ingested a video capsule soon after admission to the hospital. These patients had further endoscopic workup based on the findings from the capsule. Patients in the control group underwent endoscopic evaluation (ie, upper endoscopy, capsule endoscopy, and/or colonoscopy) to identify the source of bleeding as directed by the attending gastroenterologist's interpretation of their clinical presentation. The primary endpoint for this study was the rate of localization of bleeding during hospitalization.Eighty-seven patients were included in this study: 45 randomized to the standard of care arm and 42 to the early capsule arm. A bleeding source was localized in 64.3% of the patients in the early capsule arm and in 31.1% of the patients in the standard of care arm (P .01). The likelihood of endoscopic localization of bleeding over time was greater for patients receiving early capsule endoscopy compared with those in the standard of care arm (adjusted hazard ratio, 2.77; 95% confidence interval, 1.36-5.64).For patients admitted to the hospital for NHGIB, early capsule endoscopy is a safe and effective alternative for the detection of the source of bleeding. (Clinical trial registration number: NCT02442830.).

Background and Aims Deep learning is an innovative algorithm based on neural networks. Wireless capsule endoscopy (WCE) is considered the criterion standard for detecting small-bowel diseases. Manual examination of WCE is time-consuming and can benefit from automatic detection using artificial intelligence (AI). We aimed to perform a systematic review of the current literature pertaining to deep learning implementation in WCE. Methods We conducted a search in PubMed for all original publications on the subject of deep learning applications in WCE published between January 1, 2016 and December 15, 2019. Evaluation of the risk of bias was performed using tailored Quality Assessment of Diagnostic Accuracy Studies-2. Pooled sensitivity and specificity were calculated. Summary receiver operating characteristic curves were plotted. Results Of the 45 studies retrieved, 19 studies were included. All studies were retrospective. Deep learning applications for WCE included detection of ulcers, polyps, celiac disease, bleeding, and hookworm. Detection accuracy was above 90% for most studies and diseases. Pooled sensitivity and specificity for ulcer detection were .95 (95% confidence interval [CI], .89-.98) and .94 (95% CI, .90-.96), respectively. Pooled sensitivity and specificity for bleeding or bleeding source were .98 (95% CI, .96-.99) and .99 (95% CI, .97-.99), respectively. Conclusions Deep learning has achieved excellent performance for the detection of a range of diseases in WCE. Notwithstanding, current research is based on retrospective studies with a high risk of bias. Thus, future prospective, multicenter studies are necessary for this technology to be implemented in the clinical use of WCE.

Background and Aims Protruding lesions of the small bowel vary in wireless capsule endoscopy (WCE) images, and their automatic detection may be difficult. We aimed to develop and test a deep learning–based system to automatically detect protruding lesions of various types in WCE images. Methods We trained a deep convolutional neural network (CNN), using 30,584 WCE images of protruding lesions from 292 patients. We evaluated CNN performance by calculating the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity, using an independent set of 17,507 test images from 93 patients, including 7507 images of protruding lesions from 73 patients. Results The developed CNN analyzed 17,507 images in 530.462 seconds. The AUC for detection of protruding lesions was 0.911 (95% confidence interval [Cl], 0.9069–0.9155). The sensitivity and specificity of the CNN were 90.7% (95% CI, 90.0%–91.4%) and 79.8% (95% CI, 79.0%–80.6%), respectively, at the optimal cut-off value of 0.317 for probability score. In a subgroup analysis of the category of protruding lesions, the sensitivities were 86.5%, 92.0%, 95.8%, 77.0%, and 94.4% for the detection of polyps, nodules, epithelial tumors, submucosal tumors, and venous structures, respectively. In individual patient analyses (n = 73), the detection rate of protruding lesions was 98.6%. Conclusion We developed and tested a new computer-aided system based on a CNN to automatically detect various protruding lesions in WCE images. Patient-level analyses with larger cohorts and efforts to achieve better diagnostic performance are necessary in further studies.

Background and Aims Seronegative villous atrophy (SNVA), raised intraepithelial lymphocytes (IELs), and crypt hyperplasia on duodenal histology can be secondary to celiac disease (CD) or other causes such as medications or infections. Our aims were to assess the role of small-bowel capsule endoscopy (SBCE) in these patients and to ascertain whether findings on SBCE at diagnosis can predict disease outcome. Methods Patients (n = 177) with SNVA, IELs, ± crypt hyperplasia on duodenal histology were studied. These patients all had an equivocal diagnosis of CD. Results Overall, 56 patients (31.6%) had a positive SBCE. Thirty-three patients (58.9%) had disease affecting the proximal third of the small bowel (SB). The diagnostic yield of SBCE was 40.0% (22 patients), 51.4% (18 patients), 27.0% (10 patients), and 14.0% (7 patients) in patients with an unknown cause for SNVA (SNVA-UO), patients with SNVA who responded to a gluten-free diet (SNVA-CD), patients with a known cause for SNVA, and patients with railed IELs ± crypt hyperplasia, respectively. In SNVA-UO, SBCE at diagnosis was more likely to be positive in patients with persistent SNVA (10, 90.9%) and persistent SNVA with lymphoproliferative features (4, 80.4%) than patients with spontaneous resolution of SNVA (8, 20.5%) (P = .0001). All patients in the SNVA-CD group who eventually developed adverse events had a positive SBCE (P = .022). They also had more extensive SB disease than those without adverse events (50% vs 1% P = .002). More extensive SB disease on SBCE correlated with a higher SNVA-related mortality in patients with SNVA-UO and SNVA-CD (P = .019). Severity of histology did not correlate with mortality (P = .793). Conclusions A positive SBCE at diagnosis predicts a worse outcome. More importantly, more extensive disease in these patients is associated with poor survival. Targeting patients with extensive disease at diagnosis with more aggressive therapy can help to improve prognosis.

Background and Aims Magnetically assisted capsule endoscopy (MACE) potentially offers a comfortable, patient friendly, and community-based alternative to gastroscopy (EGD). This pilot study aims to explore whether this approach can be used to accurately diagnose Barrett’s esophagus (BE) and esophageal varices. Method The MiroCam Navi capsule system was used to examine the upper GI tract in patients due to undergo a clinically indicated EGD. A total of 50 participants were enrolled: 34 had known pathology (17 BE, 17 esophageal varices [EV]) and 16 controls. Patients underwent the MACE procedure with the operator blinded to the indication and any previous endoscopic diagnoses. The subsequent EGD was performed by an endoscopist blinded to the MACE findings. Diagnostic yield, comfort, and patient preference between the 2 modalities were compared. Results The mean age of the participants was 61 years, the male/female ratio was 2.1:1, the mean body mass index was 29.5 kg/m2, and the average chest measurement was 105.3 cm. Forty-seven patients underwent both procedures; 3 patients were unable to swallow the capsule. With the use of the magnet, it was possible to hold the capsule within the esophagus for a mean duration of 190 seconds and up to a maximum of 634 seconds. A correct real-time MACE diagnosis was made in 11 of 15 patients with EV (sensitivity 73.3% [95% confidence interval (CI), 44.9%-92.2%] and specificity 100% [95% CI, 89.1%-100%]) and 15 of 16 patients with BE (sensitivity 93.8% [95% CI, 69.8%-99.8%] and specificity of 100% [95% CI, 88.8%-100%]). MACE was considered more comfortable than conventional endoscopy (P Conclusion This pilot study demonstrates that MACE is both safe and well tolerated by patients. Accuracy for the diagnosis of BE was high, and therefore MACE may have a role in screening for this condition. (Clinical trial registration number: NCT02852161.)

Background and Aims The aim of our study was to develop and evaluate a deep learning algorithm for the automated detection of small-bowel ulcers in Crohn's disease (CD) on capsule endoscopy (CE) images of individual patients. Methods We retrospectively collected CE images of known CD patients and control subjects. Each image was labeled by an expert gastroenterologist as either normal mucosa or containing mucosal ulcers. A convolutional neural network was trained to classify images into either normal mucosa or mucosal ulcers. First, we trained the network on 5-fold randomly split images (each fold with 80% training images and 20% images testing). We then conducted 10 experiments in which images from n – 1 patients were used to train a network and images from a different individual patient were used to test the network. Results of the networks were compared for randomly split images and for individual patients. Area under the curves (AUCs) and accuracies were computed for each individual network. Results Overall, our dataset included 17,640 CE images from 49 patients: 7391 images with mucosal ulcers and 10,249 images of normal mucosa. For randomly split images results were excellent, with AUCs of .99 and accuracies ranging from 95.4% to 96.7%. For individual patient-level experiments, the AUCs were also excellent (.94-.99). Conclusions Deep learning technology provides accurate and fast automated detection of mucosal ulcers on CE images. Individual patient-level analysis provided high and consistent diagnostic accuracy with shortened reading time; in the future, deep learning algorithms may augment and facilitate CE reading.

Background and Aims Second-generation colon capsule endoscopy (CCE-2) has shown promising accuracy for the diagnosis of overall neoplasia. Advanced neoplasia (AN) represents the main target of colorectal cancer screening programs. Our aim was to assess the diagnostic accuracy of CCE-2 for the detection of AN in patients with a positive result for the fecal immunochemical test (FIT) who are undergoing screening. Methods Patients aged 50 to 69 years with a positive result for the FIT in 4 population screening programs in Italy and Spain were enrolled. Screenees were asked to undergo CCE-2, followed by traditional colonoscopy (TC). TC was performed the same day or the following morning. Bowel preparation included a split-dose polyethylene glycol–based regimen, with sodium phosphate (NaP) with gastrografin as boosters. The CCE-2 video was read by an endoscopist blinded to the results of TC. The main outcomes were CCE-2 accuracy in terms of sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for AN when using 2 different size thresholds for TC referral (ie, polyps ≥6 mm and ≥10 mm). Results Two hundred twenty-two patients were enrolled, and 178 patients completed both CCE-2 and TC (87.7%). Overall, 59 cases of AN were detected at TC. CCE-2 sensitivity was 90%, specificity was 66.1%, PPV was 57.4%, and NPV was 92.9% for AN when using a 6-mm cut-off (TC referral rate, 52.8%) and 76.7%, 90.7%, 80.7%, and 88.4% when using a 10-mm cut-off (TC referral rate, 32%), respectively. CCE-2 detected that 8 of 9 already developed colorectal cancers. Among the 41 false positives at the 6-mm cut-off, 34 (82.9%) presented with a nonadvanced adenoma at TC. Mean transit time was 4 hours and 4 minutes, and ≥70% of patients excreted the capsule within 5 hours. Conclusions In an enriched disease setting, we showed the high sensitivity of CCE-2 for the diagnosis of AN at a 6-mm cut-off. The apparently low CCE-2 specificity is related to the choice of AN as the main outcome. (Clinical trial registration number: ISRCTN 62158762.)

International audience; Background and AimsGastrointestinal angiectasia (GIA) is the most common small bowel (SB) vascular lesion, with an inherent risk of bleeding. SB Capsule Endoscopy (SB-CE) is the currently accepted diagnostic procedure. The aim of this study was to develop a computer-assisted diagnosis (CAD) tool for the detection of GIA. MethodsDeidentified SB-CE still frames featuring annotated typical GIA and normal control still frames, were selected from a database. A semantic segmentation images approach associated with a convolutional neural network (CNN) was used for deep feature extractions and classification. Two datasets of still frames were created and used for machine-learning and for algorithm testing. ResultsThe GIA detection algorithm yielded a sensitivity of 100%, a specificity of 96%, a positive predictive value of 96%, and a negative predictive value of 100%. Reproducibility was optimal. The reading process for an entire SB-CE video would take 2340 seconds (39 minutes).Conclusion The developed CNN-based algorithm had high diagnostic performances allowing detection of GIA in SB-CE still frames. This study paves the way for future automated CNN-based SB-CE reading softwares.

Background Patients frequently consult primary care physicians and gastroenterologists when experiencing chronic abdominal pain. Although its diagnostic efficacy in these settings is uncertain, small-bowel capsule endoscopy (SBCE) has been used to evaluate the unexplained reasons for abdominal pain. Objective To evaluate the diagnostic yield of SBCE in patients with unexplained chronic abdominal pain. Design We performed a retrospective review of publications reporting the diagnostic yield of SBCE in patients with unexplained chronic abdominal pain and calculated the overall diagnostic yield. Setting Two investigators independently searched studies from databases and analyzed the results. Patients A total of 1520 patients from 21 studies were included. Interventions Small-bowel capsule endoscopy. Main Outcome Measurements Per-patient diagnostic yield, with 95% confidence intervals (CI), was evaluated by a random-effect model. Clear categorical analysis also was performed. Results The pooled diagnostic yield of SBCE in patients with unexplained chronic abdominal pain was 20.9% (95% CI, 15.9%-25.9%), with high heterogeneity ( I 2 = 80.0%; P ＜ .001). Inflammatory lesions were the most common (78.3%) positive findings, followed by tumors (9.0%). Limitations Heterogeneity among studies, retrospective design, variable chronicity of abdominal pain, and different previous examinations before SBCE. Conclusion SBCE provides a noninvasive diagnostic tool for patients with unexplained chronic abdominal pain, but the diagnostic yield is limited (20.9%). Among patients with positive findings, inflammatory lesions are the most common.