CapsoView (CVV); CapsoCloud (CLD)

{0} [LOGO] FDA U.S. FOOD & DRUG ADMINISTRATION September 25, 2026 CapsoVision, Inc. Zane Liu Director, Regulatory Affairs 18805 Cox Avenue, Suite 250 Saratoga, California 95070 Re: K254230 Trade/Device Name: CapsoView (CVV); CapsoCloud (CLD) Regulation Number: 21 CFR 876.1540 Regulation Name: Gastrointestinal Capsule Endoscopy Analysis Software Device Regulatory Class: Class II Product Code: QZF Dated: August 26, 2026 Received: August 26, 2026 Dear Zane Liu: We have reviewed your section 510(k) premarket notification of intent to market the device referenced above and have determined the device is substantially equivalent (for the indications for use stated in the enclosure) to legally marketed predicate devices marketed in interstate commerce prior to May 28, 1976, the enactment date of the Medical Device Amendments, or to devices that have been reclassified in accordance with the provisions of the Federal Food, Drug, and Cosmetic Act (the Act) that do not require approval of a premarket approval application (PMA). You may, therefore, market the device, subject to the general controls provisions of the Act. Although this letter refers to your product as a device, please be aware that some cleared products may instead be combination products. The 510(k) Premarket Notification Database available at https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm identifies combination product submissions. The general controls provisions of the Act include requirements for annual registration, listing of devices, good manufacturing practice, labeling, and prohibitions against misbranding and adulteration. Please note: CDRH does not evaluate information related to contract liability warranties. We remind you, however, that device labeling must be truthful and not misleading. If your device is classified (see above) into either class II (Special Controls) or class III (PMA), it may be subject to additional controls. Existing major regulations affecting your device can be found in the Code of Federal Regulations, Title 21, Parts 800 to 898. In addition, FDA may publish further announcements concerning your device in the Federal Register. Before making any change significantly affecting the safety or effectiveness of the device, you must submit a new premarket notification in accordance with 21 CFR 807.81. Additional information about changes that may require a new premarket notification are provided in the FDA guidance documents entitled "Deciding U.S. Food & Drug Administration 10903 New Hampshire Avenue Silver Spring, MD 20993 www.fda.gov {1} K254230 - Zane Liu Page 2 When to Submit a 510(k) for a Change to an Existing Device" (https://www.fda.gov/media/99812/download) and "Deciding When to Submit a 510(k) for a Software Change to an Existing Device" (https://www.fda.gov/media/99785/download). Your device is also subject to, among other requirements, the Quality Management System Regulation (QMSR) (21 CFR Part 820), which includes, but is not limited to, ISO 13485 clause 7.3 (Design controls), ISO 13485 clause 8.3 (Nonconforming product), ISO 13485 clause 8.5.2 (Corrective action), and ISO 13485 clause 8.5.3 (Preventative action). Please note that regardless of whether a change requires premarket review, the QMSR requires device manufacturers to review and approve changes to device design and production (ISO 13485 clause 7.3 and ISO 13485 clause 7.5) and document changes and approvals in the Medical Device File (ISO 13485 clause 4.2.3). Please be advised that FDA's issuance of a substantial equivalence determination does not mean that FDA has made a determination that your device complies with other requirements of the Act or any Federal statutes and regulations administered by other Federal agencies. You must comply with all the Act's requirements, including, but not limited to: registration and listing (21 CFR Part 807); labeling (21 CFR Part 801); medical device reporting (reporting of medical device-related adverse events) (21 CFR Part 803) for devices or postmarketing safety reporting (21 CFR Part 4, Subpart B) for combination products (see https://www.fda.gov/combination-products/guidance-regulatory-information/postmarketing-safety-reporting-combination-products); good manufacturing practice requirements as set forth in the Quality Management System Regulation (QMSR) (21 CFR Part 820) for devices or current good manufacturing practices (21 CFR Part 4, Subpart A) for combination products; and, if applicable, the electronic product radiation control provisions (Sections 531-542 of the Act); 21 CFR Parts 1000-1050. All medical devices, including Class I and unclassified devices and combination product device constituent parts are required to be in compliance with the final Unique Device Identification System rule ("UDI Rule"). The UDI Rule requires, among other things, that a device bear a unique device identifier (UDI) on its label and package (21 CFR 801.20(a)) unless an exception or alternative applies (21 CFR 801.20(b)) and that the dates on the device label be formatted in accordance with 21 CFR 801.18. The UDI Rule (21 CFR 830.300(a) and 830.320(b)) also requires that certain information be submitted to the Global Unique Device Identification Database (GUDID) (21 CFR Part 830 Subpart E). For additional information on these requirements, please see the UDI System webpage at https://www.fda.gov/medical-devices/device-advice-comprehensive-regulatory-assistance/unique-device-identification-system-udi-system. Also, please note the regulation entitled, "Misbranding by reference to premarket notification" (21 CFR 807.97). For questions regarding the reporting of adverse events under the MDR regulation (21 CFR Part 803), please go to https://www.fda.gov/medical-devices/medical-device-safety/medical-device-reporting-mdr-how-report-medical-device-problems. For comprehensive regulatory information about medical devices and radiation-emitting products, including information about labeling regulations, please see Device Advice (https://www.fda.gov/medical-devices/device-advice-comprehensive-regulatory-assistance) and CDRH Learn (https://www.fda.gov/training-and-continuing-education/cdrh-learn). Additionally, you may contact the Division of Industry and Consumer Education (DICE) to ask a question about a specific regulatory topic. See the DICE website (https://www.fda.gov/medical-devices/device-advice-comprehensive-regulatory- {2} K254230 - Zane Liu Page 3 assistance/contact-us-division-industry-and-consumer-education-dice) for more information or contact DICE by email (DICE@fda.hhs.gov) or phone (1-800-638-2041 or 301-796-7100). Sincerely, SIVAKAMI VENKATACHALAM -S *for* Shanil P. Haugen, Ph.D. Assistant Director DHT3A: Division of Renal, Gastrointestinal, Obesity, and Transplant Devices OHT3: Office of Gastrorenal, ObGyn, General Hospital, and Urology Devices Office of Product Evaluation and Quality Center for Devices and Radiological Health Enclosure {3} K254230 DEPARTMENT OF HEALTH AND HUMAN SERVICES Food and Drug Administration # Indications for Use Form Approved: OMB No. 0910-0120 Expiration Date: 07/31/2026 See PRA Statement below. 510(k) Number (if known) K254230 Device Name CapsoView (CVV) CapsoCloud (CLD) Indications for Use (Describe) CapsoView is a software application for the download, transcode, and review of videos from CapsoCam Plus Capsule Endoscopes, and the generation of capsule endoscopy reports. CapsoCloud is a cloud-based software application for the review of videos from CapsoCam Plus Capsule Endoscopes, and the generation of capsule endoscopy reports. The software includes artificial intelligence (AI) tools for video analysis designed to aid clinicians in the review of small bowel capsule endoscopy images collected from adult and pediatric patients over two years of age, in whom the capsule endoscopy images were obtained for suspected small bowel bleeding. The “AI Highlights” tool aids small bowel capsule endoscopy reviewers by decreasing the time to review capsule endoscopy images. The “Suggested Landmarks” tool assists reviewers in the identification of the digestive tract location (esophagus, stomach, small bowel, and colon). The “Green Detection” tool identifies and sequesters obscured, uninterpretable frames (e.g., obscured by debris, bile, or bubbles) to streamline the review process. The AI tools are not intended to replace the clinician's diagnostic interpretation. Clinicians are responsible for conducting their own assessment of the AI-assisted findings by reviewing the entire video as clinically appropriate. Type of Use (Select one or both, as applicable) ☑ Prescription Use (Part 21 CFR 801 Subpart D) ☐ Over-The-Counter Use (21 CFR 801 Subpart C) # CONTINUE ON A SEPARATE PAGE IF NEEDED. This section applies only to requirements of the Paperwork Reduction Act of 1995. # *DO NOT SEND YOUR COMPLETED FORM TO THE PRA STAFF EMAIL ADDRESS BELOW.* The burden time for this collection of information is estimated to average 79 hours per response, including the time to review instructions, search existing data sources, gather and maintain the data needed and complete and review the collection of information. Send comments regarding this burden estimate or any other aspect of this information collection, including suggestions for reducing this burden, to: Department of Health and Human Services Food and Drug Administration Office of Chief Information Officer Paperwork Reduction Act (PRA) Staff PRAStaff@fda.hhs.gov "An agency may not conduct or sponsor, and a person is not required to respond to, a collection of information unless it displays a currently valid OMB number." FORM FDA 3881 (8/23) Page 1 of 1 PSC Publishing Services (301) 443-6740 EF {4} K254230 CapsoVision empowering strength innovation ### 510(k) Summary 510(k) #: K254230 Prepared on: September 24, 2026 ### Contact Details: 21 CFR 807.92(a)(1) Applicant: CapsoVision Inc. 18805 Cox Avenue Suite 250 Saratoga, CA, 95070 United States Applicant Contact: Azimun Jamal Senior Director, Quality Assurance and Regulatory Affairs Email: azimun.jamal@capsovision.com Phone: 408-866-6358 Correspondent Contact: Zane Liu Director, Regulatory Affairs Email: zane.liu@capsovision.com Phone: 408-416-4142 ### Device Name 21 CFR 807.92(a)(2) Device Trade Name: CapsoView (CVV) CapsoCloud (CLD) Common Name: Gastrointestinal capsule endoscopy analysis software device Classification Name: Gastrointestinal Capsule Endoscopy Analysis Software Device Regulation Number: 876.1540 Product Code(s): QZF ### Legally Marketed Predicate Devices 21 CFR 807.92(a)(3) Primary Predicate Device: Ankon Technologies Co., Ltd. NaviCam ProScan (DEN230027, Product Code: QZF) Reference Device: DigestAID - Artificial Intelligence Development, S.A., Deep Capsule (K250655, Product Code: QZF) ### Device Description Summary 21 CFR 807.92(a)(4) This submission focuses on the introduction of an artificial intelligence (AI)-assisted reading tool to the CapsoView and CapsoCloud reading software for the CapsoCam Plus (SV-3) Capsule Endoscopy System (K242643, "CapsoCam Plus"). CapsoView is proprietary workstation software used with CapsoCam Plus to download, view, and generate reports from capsule endoscopy images. CapsoCloud is a cloud-based (Web or mobile App) software application used to manage procedures and, if videos are uploaded onto the platform, review images from CapsoCam Plus and generate capsule endoscopy reports. This submission does not involve any changes to the hardware components cleared under K242643. K254230 510(k) Summary Page 1 {5} K254230 CapsoVision empowering through innovation The proposed artificial intelligence tools analyze capsule endoscopy images acquired by CapsoCam Plus to recognize and mark potential regions of interest, with the goal of reducing the time it takes small bowel capsule endoscopy reviewers to review capsule endoscopy images and to assist these reviewers in identifying the digestive tract location (oral cavity, esophagus, stomach, small bowel, cecum). These tools are: - The AI Highlights feature recognizes and marks images containing suspected abnormal lesions. During analysis, images are classified in a binary matter as "abnormal" or "normal," and for suspected "abnormal" frames, bounding boxes are additionally overlaid to identify the suspected lesion region. - The Suggested Landmarks feature identifies the specific segment of the gastrointestinal (GI) tract where an image was captured. By analyzing the capsule endoscopy video, this process can segment the video into distinct regions: esophagus, stomach, small bowel, and the regions before (ingestion and oral cavity) and after (cecum to excretion) the primary GI tract. - The Green Detection feature identifies images obscured by debris with no visible mucosa for diagnostic interpretation (i.e., often green in color). Frames detected to be likely uninterpretable are hidden, but not deleted, from the capsule endoscopy video with the goal of improving efficiency of review. Each classification task is performed through a separate machine learning algorithm trained on images captured by the CapsoCam Plus, including capsule endoscopy videos from real-world clinical use in the United States and those collected as a part of prior clinical studies involving the SV-3 capsule endoscope. These algorithms are run during the transcoding process for SV-3 videos on CapsoView. The outcomes of the AI analyses (e.g., predicted location, abnormality status, location of any bounding box overlays) are stored as metadata alongside the video file, which does not require the altering of any frame of the original video, and may be viewed with either CapsoView or CapsoCloud. The AI analysis tools are not intended to replace gastroenterologist assessment or, where applicable, histopathological sampling. ### Intended Use/Indications for Use 21 CFR 807.92(a)(5) CapsoView is a software application for the download, transcode, and review of videos from CapsoCam Plus Capsule Endoscopes, and the generation of capsule endoscopy reports. CapsoCloud is a cloud-based software application for the review of videos from CapsoCam Plus Capsule Endoscopes, and the generation of capsule endoscopy reports. The software includes artificial intelligence (AI) tools for video analysis designed to aid clinicians in the review of small bowel capsule endoscopy images collected from adult and pediatric patients over two years of age, in whom the capsule endoscopy images were obtained for suspected small bowel bleeding. The “AI Highlights” tool aids small bowel capsule endoscopy reviewers by decreasing the time to review capsule endoscopy images. The “Suggested Landmarks” tool assists reviewers in the identification of the digestive tract location (esophagus, stomach, small bowel, and colon). The “Green Detection” tool identifies and K254230 510(k) Summary Page 2 {6} K254230 CapsoVision empowering through innovation sequesters obscured, uninterpretable frames (e.g., obscured by debris, bile, or bubbles) to streamline the review process. The AI tools are not intended to replace the clinician's diagnostic interpretation. Clinicians are responsible for conducting their own assessment of the AI-assisted findings by reviewing the entire video as clinically appropriate. ### Indications for Use Comparison 21 CFR 807.92(a)(5) The proposed device and predicate device have the same intended use, as both devices are intended to aid reviewers of capsule endoscopy videos to decrease the reading time. The intended users of the two devices are the same. Neither device is intended to replace decision making. Rather, in both devices, the analysis findings of the software are presented as potential areas of interest, and the clinician is responsible for conducting their own assessment of the findings of the AI-assisted reading through review of the entire video when deemed clinically appropriate. The proposed device and predicate device have similar Indications for Use, each focused on the same claim of “reduction in reading time” and with the patient populations consisting of patients for whom capsule endoscopy was performed due to suspected bleeding. The primary difference in Indications for Use between the proposed device and the predicate device is with respect to intended patient population, which does not constitute a difference in intended use, but rather reflects the study population of each device. ### Technological Comparison 21 CFR 807.92(a)(6) The proposed device is similar to the predicate device with respect to technological characteristics. Both devices are AI-assisted reading tools for capsule endoscopy of the small bowel, built into the reading software of their respective capsule endoscopy systems, designed to analyze recorded videos and provide outputs which recognize and mark potential pathologies. For both devices, reading assistance is provided both directly, through the detection of suspected pathologies in the video, and indirectly, through the identification of frames of lower review priority (e.g., presumed “normal” frames and videos taken outside the small bowel). In both devices, potential pathologies are marked through a rectangular “bounding box” and regions without such detections remain unchanged. Both devices provide several “reading modes,” such that the user can filter through the video based on the detection outcome (e.g., frames with suspected pathology only) or to review the original video. There are technological differences between the two devices, including within their respective neural network-based detection algorithms (e.g., the proposed device utilizes U.S. training data whereas the predicate device utilizes OUS [Chinese] data) and in the discrete classifications output by each device, but they do not raise different questions of safety or effectiveness. In both devices, the indicated patient population is consistent with that for which the underlying artificial intelligence algorithm has been developed and tested. Furthermore, as both devices provide the same form of reading assistance, their labeling each discloses the potential pathologies detectable by each device, neither device replaces clinical decision making, and the clinician retains the responsibility to exercise clinical judgment on the device findings through review of the full video, this technological difference in detectable pathology types thus does not raise different K254230 510(k) Summary Page 3 {7} K254230 CapsoVision empowering through innovation questions of safety or effectiveness. As such, CapsoVision is of the opinion that the proposed device is substantially equivalent to the predicate device with respect to technological characteristics. # **Non-Clinical and/or Clinical Tests Summary & Conclusions** 21 CFR 807.92(b) # **Software/Cybersecurity** CapsoVision has conducted comprehensive software verification and validation testing on both CapsoView and CapsoCloud. The results of these tests demonstrate that both software perform as intended, in accordance with their respective Software Requirements Specifications. CapsoView/CapsoCloud was identified as having a basic level of concern as defined in the FDA guidance document “Guidance for the Content of Premarket Submissions for Software Contained in Medical Devices.” The software documentation included: 1. Software Description 2. Risk Management File 3. Software Requirement Specification 4. System and Software Architecture Design 5. Software Design Specification 6. Software Development, Configuration Management, and Maintenance Practices 7. Software Testing as Part of Verification and Validation 8. Software Version History 9. Unresolved Software Anomalies Risk analysis was provided for the software with a description of the hazards, their causes and severity as well as acceptable methods for control of the identified risks. CapsoView/CapsoCloud provided a description, with test protocols including pass/fail criteria and report of results, of acceptable verification and validation activities at the unit, integration and system level. All testing met design specifications and passed successfully. Cybersecurity documentation included recommended information from the FDA guidance document “FDA Cybersecurity in Medical Devices: Quality System Considerations and Content of Premarket Submissions.” This includes threat identification, vulnerability assessment, likelihood and impact assessment, cybersecurity mitigation information, security policies and controls, continuous monitoring and review activities, regular auditing and cybersecurity testing. # **Development Dataset and Ground Truth Establishment** All training and testing data used during the development phase are sourced from CapsoCloud, which consists of patient capsule endoscopy videos and reports acquired from real-world clinical use of the CapsoCam Plus, with a maximum of one capsule recording per patient. A total of K254230 510(k) Summary Page 4 {8} K254230 CapsoVision empowering through innovation 1,432 patients contributed to the developmental datasets, with a 2:1 partition by patient across the training and internal testing sets. Following anonymization, initial screening is performed by CapsoVision staff to identify videos where the accompanying reports indicated the presence of pathology. Video clips of 7~11 consecutive frames are extracted based on this initial screen, and these clips are then provided to at least two experienced, independent capsule endoscopists for labeling. Only raw images, blinded from the keywords used for parsing, to the readers, with no text or any other information. The capsule endoscopists are asked to 1) draw contours of the pathologies and 2) select one of the 14 classes for each contour. Agreement between two independent labelers is required to confirm the label of the training data. A total of 8,176 confirmed positive images and 27,296 confirmed negative images were ultimately collected and labeled. Table 1 below summarizes the demographic characteristics of the training and internal testing datasets. Table 1. Comparison of Demographic Characteristics in Training and Internal Testing Datasets | | Training (n = 886 patients) | Internal Testing (n = 546 patients) | | --- | --- | --- | | Number of Images | | | | Ground Truth Positive | 5365 | 2811 | | Ground Truth Negative | 17342 | 9954 | | Patients by Age | | | | 0–5 | 1 | 0 | | 6–12 | 6 | 8 | | 13–21 | 42 | 16 | | 22–39 | 64 | 51 | | 40–59 | 180 | 127 | | 60+ | 593 | 344 | | Patients by U.S. Census Region of Hospital/Clinic | | | | West | 221 | 178 | | Midwest | 99 | 60 | | South | 471 | 257 | | Northeast | 95 | 51 | | Patients by Finding [1] | | | | Classified “Abnormal” | 464 | 260 | | Angiodysplasia/angiectasia (AVM, ectasia) | 206 | 97 | | Bleeding/blood | 99 | 54 | | Ulcers or aphthoid erosion | 151 | 93 | | Protruding Lesions (Polyps, Lipomas, Nodules, Masses) | 79 | 42 | | Scalloped mucosa | 9 | 5 | | Diverticulum | 17 | 16 | | Edematous villi | 13 | 11 | | Classified “Normal” | 675 | 420 | | Normal Mucosa | 365 | 195 | | Lymphangiectasia | 163 | 104 | | Xanthoma | 16 | 11 | | Phlebectasia | 44 | 30 | | Duodenal Papilla | 33 | 22 | K254230 510(k) Summary Page 5 {9} CapsoVision empowering through innovation K254230 | Erythema | 195 | 131 | | --- | --- | --- | | Petechiae | 139 | 66 | [1] The count represents the number of unique patients who contributed frames depicting a finding. The AI Highlights feature recognizes and marks images containing suspected abnormal lesions. The software is specifically trained to recognize seven pathology types: Angiodysplasia/angiectasia (AVM, ectasia), Bleeding/Blood, Ulcers or aphthoid erosion, Protruding Lesions (Polyps, Lipomas, Nodules, Masses), Scalloped mucosa, Diverticulum, and Edematous villi. During analysis, images are classified in a binary matter as "abnormal" or "normal," and for suspected "abnormal" frames, bounding boxes are additionally overlaid to identify the suspected lesion region. Although the device was trained on these specific pathology types, and is capable of recognizing potential small bowel pathologies, it was not validated to differentiate these pathologies. ### Standalone/Clinical Performance Testing Following completion of training and internal testing and locking of the device artificial intelligence algorithms which are the subject of the proposed device, CapsoVision has conducted comprehensive standalone and clinical performance testing on the artificial intelligence algorithms to evaluate both the detection performance of each algorithm, for their respective regions of interest, and the collective performance of these algorithms in assisting capsule endoscopists in the detection of clinically-relevant small bowel pathologies during capsule endoscopy to reduce the reading time of capsule endoscopy videos. ### Study Population and Reference Standard: The evaluation dataset consisted of 111 deidentified capsule endoscopy videos collected retrospectively from the commercial use of the CapsoCam Plus (SV-3) system in the United States via the CapsoCloud® platform. To ensure independence, the dataset had no patient-level overlap with the data used for algorithm training. The population included both adult and pediatric patients from a diverse range of 70 U.S. hospitals and medical centers to ensure representation of real-world clinical practice. The patients ranged between 4 and 89 years of age at the time of procedure (Median: 49.5, SD: 28.1) with 33 patients in the pediatric subgroup and 78 in the adult subgroup. Table 2 presents the patient demographics and clinical characteristics. Table 2. Summary of Study Population | Age, years | n | 111 | | --- | --- | --- | | | Mean (SD) | 49.5 (28.1) | | | Median | 61.0 | | | Min, Max | 4.0, 89.0 | | Indication for Capsule Endoscopy | Bleeding | 63 (57%) | | | Non-Bleeding | 45 (41%) | | | Not Reported | 3 (2%) | | Finding of Blood, n (%) | Bleeding | 72 (65%) | | | No Bleeding | 39 (35%) | K254230 510(k) Summary Page 6 {10} CapsoVision empowering through innovation K254230 | Age group, n (%) | 4 – 5 | 5 (5%) | | --- | --- | --- | | | 6 – 12 | 12 (11%) | | | 13 – 21 | 16 (14%) | | | 22 – 39 | 8 (7%) | | | 40 - 59 | 13 (12%) | | | 60+ | 57 (51%) | | Gender | Male | 13 (12%) | | | Female | 18 (16%) | | | Unknown | 80 (72%) | | U.S. Census Region of Original Site | West | 43 (38.7%) | | | Midwest | 21 (18.9%) | | | South | 34 (30.6%) | | | Northeast | 13 (11.7%) | A "ground truth" reference standard for 1) the presence of pathology (AI Highlights), 2) the presence of obstructing debris (green detection), and 3) GI landmark locations (Suggested Landmarks), was established by a panel of independent, expert gastroenterologists utilizing a consensus methodology. Following this process, 82 videos were found with at least one clinically-relevant pathology and 29 videos included no such pathology. A total of 6,073 green frames were identified across the videos. Duodenal and cecal landmarks were established for all 111 videos. The esophageal landmark was established for 98 videos. The gastric landmark was established for 100 videos. ### Standalone Performance Assessment A standalone performance assessment was conducted to characterize the accuracy of the software's algorithms (Pathology Detection, Green Detection, and Landmark Segmentation) independent of clinician interaction. The algorithms were tested against a statistically powered dataset of videos independent of the training data. Performance was evaluated against the expert consensus reference standard. The results of the study were as follows: - Pathology Detection (AI Highlights): The algorithm achieved a per-patient sensitivity of 100% (95% CI: 95.60%, 100%) and a per-patient specificity of 0% (95% CI: 0.00%, 11.94%). As the algorithm is designed to recommend frames for every video for consideration by the physician, this specificity result was expected. The per-frame sensitivity was observed at 85.34% (95% CI: 71.98%, 90.89%) and per-frame specificity was observed at 87.29% (95% CI: 85.39%, 89.07%). Bounding box accuracy in a random sampling of 2,368 true positive frames was observed at 98.09% (95% CI: 96.93%, 98.92%). - Green Detection: The algorithm achieved a specificity of 98.88% (95% CI: 98.38%, 99.29%) and a sensitivity of 68.86% (95% CI: 61.59%, 76.13%) for identifying obscured frames. K254230 510(k) Summary Page 7 {11} CapsoVision empowering through innovation K254230 - Landmark Segmentation (Suggested Landmarks): Table 3 presents the sensitivity and specificity for detecting each GI segment (oral cavity, esophagus, stomach, small bowel, and colon). Please note that for patients for whom the capsule was delivered endoscopically, the upper GI (oral cavity, esophagus, and stomach) landmarks may not be visible, depending on the location of endoscopic delivery. Table 3. Summary of Landmark Segmentation Algorithm Performance | GI Segment | Sample Size (n) | Sensitivity (95% CI) | Specificity (95% CI) | | --- | --- | --- | --- | | Oral Cavity | 98 | 74.82% (61.41%, 85.91%) | 99.51% (99.02%, 99.82%) | | Esophagus | 98 | 72.68% (47.53%, 85.89%) | 99.41% (98.90%, 99.81%) | | Stomach | 100 | 59.83% (50.07%, 68.73%) | 99.44% (98.87%, 99.87%) | | Small Bowel | 111 | 94.78% (91.96%, 97.20%) | 90.97% (87.98%, 93.60%) | | Colon | 111 | 99.38% (98.88%, 99.76%) | 96.77% (94.71%, 98.43%) | Considering that the user interface of CapsoView/CapsoCloud require the confirmation of each “landmark” at the beginning of video review, performance of the algorithm is acceptable for providing an initial recommendation to a trained capsule reader. Lower sensitivity in early GI segments (e.g., stomach) is expected in real-world capsule endoscopy due to rapid bolus transit, variable gastric distension, or un-prepped mucosal debris. In addition, the sensitivities and specificities for the comparable feature within the predicate device, NaviCam ProScan (DEN230027), represent the results from internal testing utilizing a "split" of patients from the same dataset used for algorithm training, which was enrolled solely from OUS [Chinese] sites. Furthermore, data suitable for training typically requires a high quality of bowel cleanliness, which does not reflect real-world capsule endoscopy conditions. As such, the reported sensitivity and specificity figures may overestimate the real-world performance of the feature, as would be reflected if such testing had been conducted on a fully independent, U.S. clinical cohort, as was executed for the proposed device. In addition, subgroup analyses were performed for each endpoint by region (Midwest, West, Northeast, and South U.S., per U.S. Census Bureau regions), bowel preparation quality, patient age, patient gender, indications for imaging (bleeding related vs. not), and presence of blood in findings (found vs. not). The following presents the detailed subgroup analyses conducted as a part of the standalone performance tests for Pathology Detection (Table 4 [patient] and Table 5 [frame]), Green Detection (Table 6), and Landmark Segmentation for the Small Bowel (Table 7). Table 4. Standalone Study Outcomes for the Pathology Detection Algorithm (Patient-Level) | | Total / Positive / Negative Patient Count | | | Sensitivity | Specificity ^{[1]} | | --- | --- | --- | --- | --- | --- | | **Overall** | Total: 111 | Positive | Negative | 100.00 [95.60, 100.00] | 0.00 [0.00, 11.94] | | | Ground Truth | 82 | 29 | | | | | AI | 111 | 0 | | | | **Subgroup – U.S. Region** | | | | | | | Midwest | Total: 21 | Positive | Negative | 100.00 [79.41, 100.00] | 0.00 [0.00, 52.18] | | | Ground Truth | 16 | 5 | | | | | AI | 21 | 0 | | | K254230 510(k) Summary Page 8 {12} K254230  | | Total / Positive / Negative Patient Count | | | Sensitivity | Specificity ^{[1]} | | --- | --- | --- | --- | --- | --- | | Northeast | Total: 13 | Positive | Negative | 100.00 [73.54, 100.00] | 0.00 [0.00, 97.50] | | | Ground Truth | 12 | 1 | | | | | AI | 13 | 0 | | | | South | Total: 34 | Positive | Negative | 100.00 [87.66, 100.00] | 0.00 [0.00, 45.93] | | | Ground Truth | 28 | 6 | | | | | AI | 34 | 0 | | | | West | Total: 43 | Positive | Negative | 100.00 [86.77, 100.00] | 0.00 [0.00, 19.51] | | | Ground Truth | 26 | 17 | | | | | AI | 43 | 0 | | | | **Subgroup – Bowel Prep** | | | | | | | Fair | Total: 36 | Positive | Negative | 100.00 [88.43, 100.00] | 0.00 [0.00, 45.93] | | | Ground Truth | 30 | 6 | | | | | AI | 36 | 0 | | | | Poor | Total: 1 | Positive | Negative | N/A^{[2]} | N/A^{[2]} | | | Ground Truth | 1 | 0 | | | | | AI | 1 | 0 | | | | Good | Total: 74 | Positive | Negative | 100.00 [93.02, 100.00] | 0.00 [0.00, 14.82] | | | Ground Truth | 51 | 23 | | | | | AI | 74 | 0 | | | | **Subgroup – Age** | | | | | | | 2-5 | Total: 5 | Positive | Negative | 100.00 [47.82, 100.00] | N/A^{[2]} | | | Ground Truth | 5 | 0 | | | | | AI | 5 | 0 | | | | 6-12 | Total: 12 | Positive | Negative | 100.00 [63.06, 100.00] | 0.00 [0.00, 60.24] | | | Ground Truth | 8 | 4 | | | | | AI | 12 | 0 | | | | 13-21 | Total: 16 | Positive | Negative | 100.00 [66.37, 100.00] | 0.00 [0.00, 40.96] | | | Ground Truth | 9 | 7 | | | | | AI | 16 | 0 | | | | 22-39 | Total: 8 | Positive | Negative | 100.00 [29.24, 100.00] | 0.00 [0.00, 52.18] | | | Ground Truth | 3 | 5 | | | | | AI | 8 | 0 | | | | 40-59 | Total: 13 | Positive | Negative | 100.00 [71.51, 100.00] | 0.00 [0.00, 84.19] | | | Ground Truth | 11 | 2 | | | | | AI | 13 | 0 | | | | 60+ | Total: 57 | Positive | Negative | 100.00 [92.29, 100.00] | 0.00 [0.00, 28.49] | | | Ground Truth | 46 | 11 | | | | | AI | 57 | 0 | | | | **Subgroup – Gender** | | | | | | | Male | Total: 13 | Positive | Negative | 100.00 [66.37, 100.00] | 0.00 [0.00, 60.24] | | | Ground Truth | 9 | 4 | | | | | AI | 13 | 0 | | | | Female | Total: 18 | Positive | Negative | 100.00 [71.51, 100.00] | 0.00 [0.00, 40.96] | | | Ground Truth | 11 | 7 | | | | | AI | 18 | 0 | | | | Unknown | Total: 80 | Positive | Negative | 100.00 [94.22, 100.00] | 0.00 [0.00, 18.53] | | | Ground Truth | 62 | 18 | | | | | AI | 80 | 0 | | | | **Subgroup – Indications for Imaging** | | | | | | K254230 510(k) Summary Page 9 {13} K254230 CapsoVision empowering through innovation | | Total / Positive / Negative Patient Count | | | Sensitivity | Specificity [1] | | --- | --- | --- | --- | --- | --- | | Bleeding | Total: 63 | Positive | Negative | 100.00 [92.89, 100.00] | 0.00 [0.00, 24.71] | | | Ground Truth | 50 | 13 | | | | | AI | 63 | 0 | | | | No Bleeding | Total: 45 | Positive | Negative | 100.00 [88.78, 100.00] | 0.00 [0.00, 23.16] | | | Ground Truth | 31 | 14 | | | | | AI | 45 | 0 | | | | Unknown | Total: 3 | Positive | Negative | 100.00 [2.50, 100.00] | 0.00 [0.00, 84.19] | | | Ground Truth | 1 | 2 | | | | | AI | 3 | 0 | | | | Subgroup – Finding of Bleeding | | | | | | | Bleeding | Total: 72 | Positive | Negative | 100.00 [94.04, 100.00] | 0.00 [0.00, 26.46] | | | Ground Truth | 60 | 12 | | | | | AI | 72 | 0 | | | | No Bleeding | Total: 39 | Positive | Negative | 100.00 [84.56, 100.00] | 0.00 [0.00, 19.51] | | | Ground Truth | 22 | 17 | | | | | AI | 39 | 0 | | | [1]: All patients were false positives; therefore, a bootstrap CI was not computed, and an exact Clopper-Pearson CI is reported. [2]: Insufficient number for calculation. Table 5. Standalone Study Outcomes for the Pathology Detection Algorithm (Frame-Level) | | Total / Positive / Negative Frame Count | | | Sensitivity | Bounding Box Accuracy | Specificity | | --- | --- | --- | --- | --- | --- | --- | | Overall | Total: 1,920,609 | Positive | Negative | 85.34 [71.98, 90.89] | 98.09 [96.93, 98.92] | 87.29 (85.38, 89.07) | | | Ground Truth | 26,930 | 1,893,679 | | | | | | AI | 263,617 | 1,656,992 | | | | | | Bounding Box | 2,368 | N/A | | | | | Subgroup – U.S. Region | | | | | | | | Midwest | Total: 398,846 | Positive | Negative | 89.81 [74.82, 95.65] | 99.15 [98.37, 100.00] | 85.81 [81.34, 90.38] | | | Ground Truth | 6,555 | 392,291 | | | | | | AI | 61,526 | 337,320 | | | | | | Bounding Box | 818 | N/A | | | | | Northeast | Total: 226,277 | Positive | Negative | 48.31 [17.02, 95.63] | 98.28 [93.71, 100.00] | 89.24 [86.17, 92.32] | | | Ground Truth | 1,629 | 224,648 | | | | | | AI | 24,951 | 201,326 | | | | | | Bounding Box | 174 | N/A | | | | | South | Total: 659,390 | Positive | Negative | 87.72 [71.05, 94.43] | 98.81 [97.61, 99.66] | 88.26 [84.6, 91.55] | | | Ground Truth | 5,790 | 653,600 | | | | | | AI | 81,794 | 577,596 | | | | | | Bounding Box | 852 | N/A | | | | K254230 510(k) Summary Page 10 {14} K254230 CapsoVision empowering through innovation | | Total / Positive / Negative Frame Count | | | Sensitivity | Bounding Box Accuracy | Specificity | | --- | --- | --- | --- | --- | --- | --- | | West | Total: 636,096 | Positive | Negative | 86.67 [57.59, 91.75] | 95.24 [91.81, 98.07] | 86.50 [83.35, 89.13] | | | Ground Truth | 12,956 | 623,140 | | | | | | AI | 95,346 | 540,750 | | | | | | Bounding Box | 524 | N/A | | | | | Subgroup – Bowel Prep | | | | | | | | Poor | Total: 12,271 | Positive | Negative | 89.23 [89.23, 89.23] | 100.00 [100.00, 100.00] | N/A [2] | | | Ground Truth | 65 | 12,206 | | | | | | AI | 1,362 | 10,909 | | | | | | Bounding Box | 22 | N/A | | | | | Fair | Total: 681,621 | Positive | Negative | 89.31 [76.36, 94.61] | 98.04 [96.03, 99.14] | 82.40 [79.26, 85.32] | | | Ground Truth | 12,524 | 669,097 | | | | | | AI | 128,889 | 552,732 | | | | | | Bounding Box | 1,375 | N/A | | | | | Good | Total: 1,226,717 | Positive | Negative | 81.86 [52.4, 90.14] | 98.13 [96.61, 99.27] | 89.97 [87.78, 91.89] | | | Ground Truth | 14,341 | 1,212,376 | | | | | | AI | 133,366 | 1,093,351 | | | | | | Bounding Box | 971 | N/A | | | | | Subgroup – Age | | | | | | | | 2-5 | Total: 115,863 | Positive | Negative | 85.84 [81.75, 98.80] | 99.44 [98.98, 100.00] | 77.71 [69.98, 86.13] | | | Ground Truth | 1,963 | 113,900 | | | | | | AI | 27,058 | 88,805 | | | | | | Bounding Box | 179 | N/A | | | | | 6-12 | Total: 154,651 | Positive | Negative | 80.09 [50.46, 93.79] | 94.12 [89.54, 97.87] | 79.28 [71.85, 84.31] | | | Ground Truth | 3,727 | 150,924 | | | | | | AI | 34,253 | 120,398 | | | | | | Bounding Box | 286 | N/A | | | | | 13-21 | Total: 265,422 | Positive | Negative | 40.52 [24.15, 72.03] | 94.34 [78.26, 100.00] | 90.64 [89.36, 91.92] | | | Ground Truth | 575 | 264,847 | | | | | | AI | 25,022 | 240,400 | | | | | | Bounding Box | 106 | N/A | | | | | 22-39 | Total: 105,570 | Positive | Negative | 92.42 [50.00, 92.49] | 96.83 [96.36, 100.00] | 92.94 [88.68, 94.7] | | | Ground Truth | 8,416 | 97,154 | | | | | | AI | 14,628 | 90,942 | | | | | | Bounding Box | 62 | N/A | | | | | 40-59 | Total: 209,518 | Positive | Negative | 95.41 [73.59, 96.66] | 99.30 [96.83, 100.00] | 82.17 [73.95, 90.56] | | | Ground Truth | 3,920 | 205,598 | | | | | | AI | 40,390 | 169,128 | | | | | | Bounding Box | 426 | N/A | | | | K254230 510(k) Summary Page 11 {15} K254230 CapsoVision empacking through innovation | | Total / Positive / Negative Frame Count | | | Sensitivity | Bounding Box Accuracy | Specificity | | --- | --- | --- | --- | --- | --- | --- | | 60+ | Total: 1,069,585 | Positive | Negative | 78.77 [55.68, 91.72] | 98.77 [97.92, 99.52] | 89.1 [86.96, 91.08] | | | Ground Truth | 8,329 | 1,061,256 | | | | | | AI | 122,266 | 947,319 | | | | | | Bounding Box | 1,309 | N/A | | | | | Subgroup – Gender | | | | | | | | Male | Total: 202,807 | Positive | Negative | 75.74 [46.64, 92.58] | 95.90 [92.47, 98.56] | 82.60 [74.82, 88.55] | | | Ground Truth | 2,403 | 200,404 | | | | | | AI | 36,692 | 166,115 | | | | | | Bounding Box | 269 | N/A | | | | | Female | Total: 269,070 | Positive | Negative | 84.38 [38.62, 97.26] | 95.21 [89.22, 100.00] | 87.82 [83.96, 90.69] | | | Ground Truth | 1,799 | 267,271 | | | | | | AI | 34,059 | 235,011 | | | | | | Bounding Box | 165 | N/A | | | | | Unknown | Total: 1,448,732 | Positive | Negative | 86.43 [70.23, 92.07] | 98.65 [97.65, 99.33] | 87.85 [85.63, 89.96] | | | Ground Truth | 22,728 | 1,426,004 | | | | | | AI | 192,866 | 1,255,866 | | | | | | Bounding Box | 1,934 | N/A | | | | | Subgroup – Indications for Imaging | | | | | | | | Bleeding | Total: 1,090,684 | Positive | Negative | 86.25 [59.18, 92.55] | 98.79 [97.85, 99.42] | 89.14 [86.81, 91.25] | | | Ground Truth | 16,960 | 1,073,724 | | | | | | AI | 131,212 | 959,472 | | | | | | Bounding Box | 1,339 | N/A | | | | | No Bleeding | Total: 773,830 | Positive | Negative | 85.70 [72.33, 92.86] | 97.66 [95.63, 99.06] | 84.64 [81.39, 87.66] | | | Ground Truth | 9,693 | 764,137 | | | | | | AI | 125,633 | 648,197 | | | | | | Bounding Box | 1,023 | N/A | | | | | Unknown | Total: 56,095 | Positive | Negative | 16.97 [16.97, 16.97] | 28.57 [28.57, 28.57] | 87.95 [80.72, 93.83] | | | Ground Truth | 277 | 55,818 | | | | | | AI | 6,772 | 49,323 | | | | | | Bounding Box | 6 | N/A | | | | | Subgroup – Finding of Bleeding | | | | | | | | Bleeding | Total: 1,303,755 | Positive | Negative | 83.29 [68.58, 91.26] | 98.29 [97.06, 99.12] | 84.81 [82.32, 87.17] | | | Ground Truth | 17,226 | 1,286,529 | | | | K254230 510(k) Summary Page 12 {16} K254230 CapsoVision empowering strength innovation | | Total / Positive / Negative Frame Count | | | Sensitivity | Bounding Box Accuracy | Specificity | | --- | --- | --- | --- | --- | --- | --- | | | AI | 209,725 | 1,094,030 | | | | | | Bounding Box | 1,990 | N/A | | | | | No Bleeding | Total: 616,854 | Positive | Negative | 88.98 [47.64, 92.58] | 97.07 [92.93, 99.27] | 92.54 [91.18, 93.82] | | | Ground Truth | 9,704 | 607,150 | | | | | | AI | 53,892 | 562,962 | | | | | | Bounding Box | 378 | N/A | | | | [1]: As all patients were false positives, a bootstrap confidence interval was not computed, and an exact confidence interval is reported. [2]. Analysis not conducted for the bowel prep poor subgroup as it comprised of one patient. Table 6. Standalone Study Outcomes for the Green Detection Algorithm (Frame-Level) | | Total / Green / Not Green Frame Count | | | Sensitivity | Specificity | | --- | --- | --- | --- | --- | --- | | Overall | Total: 1,920,609 | Green | Not Green | 68.86[61.59, 76.13] | 98.88[98.38, 99.29] | | | Ground Truth | 6,073 | 1,914,536 | | | | | AI | 25,665 | 1,894,944 | | | | Subgroup – U.S. Region | | | | | | | Midwest | Total:398,846 | Green | Not Green | 65.98 [51.30, 86.83] | 99.10 [98.41, 99.63] | | | Ground Truth | 823 | 398,023 | | | | | AI | 4,133 | 394,713 | | | | Northeast | Total:226,277 | Green | Not Green | 90.82 [60.13, 97.22] | 98.57 [97.20, 99.66] | | | Ground Truth | 523 | 225,754 | | | | | AI | 3,699 | 222,578 | | | | South | Total:659,390 | Green | Not Green | 72.10 [43.14, 88.65] | 98.15 [96.88, 99.14] | | | Ground Truth | 3,308 | 656,082 | | | | | AI | 14,492 | 644,898 | | | | West | Total:636,096 | Green | Not Green | 54.90 [31.76, 79.41] | 99.60 [99.32, 99.81] | | | Ground Truth | 1,419 | 634,677 | | | | | AI | 3,341 | 632,755 | | | | Subgroup – Bowel Prep | | | | | | | Poor | Total:12,271 | Green | Not Green | N/A [1] | N/A [1] | | | Ground Truth | 153 | 12,118 | | | | | AI | 3,106 | 9,165 | | | | Fair | Total:681,621 | Green | Not Green | 71.59 [54.99, 83.84] | 98.52 [97.76, 99.14] | | | Ground Truth | 4,755 | 676,866 | | | | | AI | 13,421 | 668,200 | | | | Good | Total:1,226,717 | Green | Not Green | 53.91 [19.46, 87.58] | 99.31 [98.91, 99.64] | | | Ground Truth | 1,165 | 1,225,552 | | | | | AI | 9,138 | 1,217,579 | | | | Subgroup – Age | | | | | | K254230 510(k) Summary Page 13 {17} CapsoVision empacking through innovation K254230 | | Total / Green / Not Green Frame Count | | | Sensitivity | Specificity | | --- | --- | --- | --- | --- | --- | | 2-5 | Total: 115,863 | Green | Not Green | 69.34 [63.21, 86.92] | 99.00 [98.37, 99.65] | | | Ground Truth | 486 | 115,377 | | | | | AI | 1,492 | 114,371 | | | | 6-12 | Total: 154,651 | Green | Not Green | 22.36 [14.96, 59.24] | 99.85 [99.74, 99.94] | | | Ground Truth | 559 | 154,092 | | | | | AI | 357 | 154,294 | | | | 13-21 | Total: 265,422 | Green | Not Green | 81.18 [31.71, 91.02] | 99.88 [99.75, 99.97] | | | Ground Truth | 186 | 265,236 | | | | | AI | 481 | 264,941 | | | | 22-39 | Total: 105,570 | Green | Not Green | 44.44 [30.0, 47.73] | 99.47 [98.74, 99.86] | | | Ground Truth | 54 | 105,516 | | | | | AI | 586 | 104,984 | | | | 40-59 | Total: 209,518 | Green | Not Green | 28.79 [8.07, 71.86] | 99.39 [98.85, 99.75] | | | Ground Truth | 771 | 208,747 | | | | | AI | 1,488 | 208,030 | | | | 60+ | Total: 1,069,585 | Green | Not Green | 82.72 [68.99, 90.57] | 98.32 [98.85, 99.75] | | | Ground Truth | 4,017 | 1,065,568 | | | | | AI | 21,261 | 1,048,324 | | | | **Subgroup – Gender** | | | | | | | Male | Total: 202,807 | Green | Not Green | 51.20 [23.66, 79.25] | 99.82 [99.66, 99.93] | | | Ground Truth | 832 | 201,975 | | | | | AI | 785 | 202,022 | | | | Female | Total: 269,070 | Green | Not Green | 72.03 [45.29, 92.77] | 99.19 [98.28, 99.88] | | | Ground Truth | 479 | 268,591 | | | | | AI | 2,524 | 266,546 | | | | Unknown | Total: 1,448,732 | Green | Not Green | 71.63 [51.48, 85.63] | 98.69 [98.06, 99.21] | | | Ground Truth | 4,762 | 1,443,970 | | | | | AI | 22,356 | 1,426,376 | | | | **Subgroup – Indications for Imaging** | | | | | | | Bleeding | Total: 1,090,684 | Green | Not Green | 82.15 [66.68, 90.27] | 98.30 [97.47, 98.98] | | | Ground Truth | 3,687 | 1,086,997 | | | | | AI | 21,507 | 1,069,177 | | | | No Bleeding | Total: 773,830 | Green | Not Green | 51.27 [30.44, 73.93] | 99.62 [99.42, 99.79] | | | Ground Truth | 2,204 | 771,626 | | | | | AI | 4,091 | 769,739 | | | | Unknown | Total: 56,095 | Green | Not Green | 12.64 [11.17, 100.00] | 99.92 [99.87, 99.99] | | | Ground Truth | 182 | 55,913 | | | | | AI | 67 | 56,028 | | | K254230 510(k) Summary Page 14 {18} K254230 CapsoVision empacking strength innovation | | Total / Green / Not Green Frame Count | | | Sensitivity | Specificity | | --- | --- | --- | --- | --- | --- | | Subgroup – Finding of Bleeding | | | | | | | Bleeding | Total: 1,303,755 | Green | Not Green | 68.14 [49.69, 82.04] | 98.72 [98.04, 99.27] | | | Ground Truth | 5,571 | 1,298,184 | | | | | AI | 20,351 | 1,283,404 | | | | No Bleeding | Total: 616,854 | Green | Not Green | 76.89 [54.30, 86.47] | 99.20 [98.62, 99.67] | | | Ground Truth | 502 | 616,352 | | | | | AI | 5,314 | 611,540 | | | [1]. Analysis not conducted for the bowel prep poor subgroup as it comprised of one patient. Table 7. Standalone Study Outcomes for the Suggested Landmark Feature for the Small Bowel (Frame-Level) | | Total / Yes (in Sm. Bowel) / No Frame Count | | | Sensitivity | Specificity | | --- | --- | --- | --- | --- | --- | | Overall | Total: 3,424,556 | Yes | No | 94.78[91.96, 97.20] | 90.97[87.98, 93.60] | | | Ground Truth | 1,920,609 | 1,503,947 | | | | | AI | 1,956,030 | 1,468,526 | | | | Subgroup – U.S. Region | | | | | | | Midwest | Total: 682,449 | Yes | No | 96.54 [92.06, 99.67] | 88.29 [80.83, 94.86] | | | Ground Truth | 398,846 | 283,603 | | | | | AI | 418,240 | 264,209 | | | | Northeast | Total: 423,708 | Yes | No | 96.17 [87.67, 99.88] | 92.62 [88.54, 96.25] | | | Ground Truth | 226,277 | 197,431 | | | | | AI | 232,185 | 191,523 | | | | South | Total: 1,061,739 | Yes | No | 91.58 [85.58, 96.70] | 89.96 [82.90, 95.10] | | | Ground Truth | 659,390 | 402,349 | | | | | AI | 644,300 | 417,439 | | | | West | Total: 1,256,660 | Yes | No | 96.48 [92.46, 99.19] | 92.33 [87.69, 96.16] | | | Ground Truth | 636,096 | 620,564 | | | | | AI | 661,305 | 595,355 | | | | Subgroup – Bowel Prep | | | | | | | Poor | Total: 36,396 | Yes | No | N/A [1] | N/A [1] | | | Ground Truth | 12,271 | 24,125 | | | | | AI | 11,360 | 25,036 | | | | Fair | Total: 1,087,793 | Yes | No | 90.30 [84.35, 95.31] | 89.81 [82.57, 95.41] | | | Ground Truth | 681,621 | 406,172 | | | | | AI | 656,905 | 430,888 | | | | Good | Total: 2,300,367 | Yes | No | 97.29 [94.60, 99.39] | 91.21 [87.92, 94.08] | | | Ground Truth | 1,226,717 | 1,073,650 | | | | | AI | 1,287,765 | 1,012,602 | | | K254230 510(k) Summary Page 15 {19} CapsoVision empacking through innovation K254230 | | Total / Yes (in Sm. Bowel) / No Frame Count | | | Sensitivity | Specificity | | --- | --- | --- | --- | --- | --- | | Subgroup – Age | | | | | | | 2-5 | Total: 166,809 | Yes | No | 96.22 [88.45, 99.96] | 98.29 [97.1, 100.00] | | | Ground Truth | 115,863 | 50,946 | | | | | AI | 112,355 | 54,454 | | | | 6-12 | Total: 319,430 | Yes | No | 89.82 [76.10, 98.88] | 91.62 [82.56, 99.55] | | | Ground Truth | 154,651 | 164,779 | | | | | AI | 152,715 | 166,715 | | | | 13-21 | Total: 454,480 | Yes | No | 99.91 [99.81, 99.97] | 90.93 [79.23, 98.26] | | | Ground Truth | 265,422 | 189,058 | | | | | AI | 282,315 | 172,165 | | | | 22-39 | Total: 239,240 | Yes | No | 97.84 [92.67, 99.98] | 94.48 [88.36, 99.40] | | | Ground Truth | 105,570 | 133,670 | | | | | AI | 110,675 | 128,565 | | | | 40-59 | Total: 421,622 | Yes | No | 95.90 [89.50, 99.90] | 93.53 [88.50, 97.31] | | | Ground Truth | 209,518 | 212,104 | | | | | AI | 214,635 | 206,987 | | | | 60+ | Total: 1,822,975 | Yes | No | 93.54 [89.19, 97.11] | 89.00 [84.38, 92.92] | | | Ground Truth | 1,069,585 | 753,390 | | | | | AI | 1,083,335 | 739,640 | | | | Subgroup – Gender | | | | | | | Male | Total: 366,793 | Yes | No | 93.46 [82.09, 99.72] | 93.22 [87.05, 98.94] | | | Ground Truth | 202,807 | 163,986 | | | | | AI | 200,650 | 166,143 | | | | Female | Total: 507,439 | Yes | No | 97.08 [93.80, 99.90] | 95.64 [90.92, 99.10] | | | Ground Truth | 269,070 | 238,369 | | | | | AI | 271,610 | 235,829 | | | | Unknown | Total: 2,550,324 | Yes | No | 94.53 [91.11, 97.40] | 89.63 [85.78, 93.02] | | | Ground Truth | 1,448,732 | 1,101,592 | | | | | AI | 1,483,770 | 1,066,554 | | | | Subgroup – Indications for Imaging | | | | | | | Bleeding | Total: 1,975,457 | Yes | No | 94.52 [90.26, 97.82] | 89.91 [85.75, 93.37] | | | Ground Truth | 1,090,684 | 884,773 | | | | | AI | 1,120,200 | 855,257 | | | | No Bleeding | Total: 1,360,984 | Yes | No | 94.77 [90.70, 98.27] | 94.69 [91.76, 97.31] | | | Ground Truth | 773,830 | 587,154 | | | | | AI | 764,550 | 596,434 | | | | Unknown | Total: 88,115 | Yes | No | 99.73 [98.61, 100.00] | 52.10 [23.30, 99.35] | | | Ground Truth | 56,095 | 32,020 | | | K254230 510(k) Summary Page 16 {20} CapsoVision empowering through innovation K254230 | | Total / Yes (in Sm. Bowel) / No Frame Count | | | Sensitivity | Specificity | | --- | --- | --- | --- | --- | --- | | | AI | 71,280 | 16,835 | | | | Subgroup – Finding of Bleeding | | | | | | | Bleeding | Total: 2,216,758 | Yes | No | 93.02 [89.01, 96.42] | 89.92 [85.79, 93.48] | | | Ground Truth | 1,303,755 | 913,003 | | | | | AI | 1,304,720 | 912,038 | | | | No Bleeding | Total: 1,207,798 | Yes | No | 98.49 [96.63, 99.78] | 92.59 [88.21, 95.89] | | | Ground Truth | 616,854 | 590,944 | | | | | AI | 651,310 | 556,488 | | | [1]. Analysis not conducted for the bowel prep poor subgroup as it comprised of one patient. ### Clinical Performance Evaluation (Reader Study) A retrospective, multi-reader, multi-case (MRMC) study was conducted to evaluate the clinical performance of the proposed device. The study involved 15 independent, board-certified U.S. gastroenterologists reviewing capsule endoscopy videos collected from real-world clinical use. A split plot, crossover design was employed where subgroups of videos were allocated to “batches” of three readers each. Within each batch, each reader reviewed videos in both modalities (Conventional and AI-Assisted) on CapsoView in a randomized order, separated by a minimum washout period of four weeks to minimize memory bias. The study aimed to demonstrate the clinical utility of the AI Highlights, Suggested Landmarks, and Green Detection Features, in reducing time to review small bowel capsule endoscopy videos. The co-primary endpoints of the study sought to demonstrate that the diagnostic accuracy (sensitivity) of endoscopists using the AI-assisted mode is non-inferior to conventional reading, and that the AI-assisted mode provides a statistically significant reduction in video reading time. Key additional endpoints evaluated other diagnostic metrics for reader performance between the two reading conditions, such as diagnostic yield. The Reader Study included data from the same 111 patients in the Standalone Study. The analysis demonstrated that patient-level sensitivity with the CapsoView AI Software was non-inferior to conventional reading, with an observed sensitivity of 92.3% (95% CI: 87.8%, 96.7%) in the AI-assisted arm compared to 71.0% (95% CI: 59.2%, 82.7%) in the conventional arm. The difference in sensitivity was 21.3% (95% CI: 12.8%, 29.8%), which constituted non-inferiority (primary endpoint) and superiority (key secondary endpoint). The specificities of the AI-assisted and conventional arms were 26.4% (95% CI: 11.2%, 41.6%) and 56.2% (95% CI: 35.7%, 76.7%), a decrease comparable to that observed in prior studies. Furthermore, the study demonstrated a statistically significant reduction in the time required to review the capsule video, with the mean reading time decreasing from 45.7 minutes in the conventional arm to 34.4 minutes in the AI-assisted arm (difference: -11.3 minutes; 95% CI: -13.5 minutes, -9.0 minutes). The study also demonstrated the superiority of the AI-assisted mode in diagnostic yield, meeting a key secondary endpoint. Diagnostic Yield increased from 63.3% (95% CI: 49.8%, 76.8%) in the conventional arm to 87.6% (95% CI: 81.2%, 93.9%) with AI assistance, a difference of 26.7%. K254230 510(k) Summary Page 17 {21} K254230 CapsoVision empowering through innovation Subgroup analyses were performed for each endpoint by region (Midwest, West, Northeast, and South U.S., per U.S. Census Bureau regions), bowel preparation quality, patient age, patient gender, indications for imaging (bleeding related vs. not), and presence of blood in findings (found vs. not). The following presents the detailed subgroup analyses of the reader study results for both patient-level diagnostic performance and reading time, which include all primary and key secondary endpoints which were met by the study results. K254230 510(k) Summary Page 18 {22} CapsoVision empowering through innovation K254230 Table 8. Summary of Reader Study Results – Diagnostic Performance – Patient Level | | Total / Positive / Negative Patient/Read (3/Video/Condition) Count | | | Sensitivity | Specificity | Diagnostic Yield | | --- | --- | --- | --- | --- | --- | --- | | Overall | Total: 111 patients | Positive | Negative | Aided:0.923[0.878, 0.967]Unaided:0.710[0.592, 0.827]Difference:0.213[0.128, 0.298] | Aided:0.264[0.112, 0.416]Unaided:0.562[0.357, 0.767]Difference:-0.298[-0.454, -0.142] | Aided:0.876[0.812, 0.939]Unaided:0.633[0.498, 0.768]Difference:0.243[0.160, 0.325] | | | Ground Truth (pt) | 82 | 29 | | | | | | AI Assisted (Reads) | 291 | 42 | | | | | | Unassisted (Reads) | 209 | 124 | | | | | Subgroup – U.S. Region | | | | | | | | Midwest | Total: 21 patients | Positive | Negative | Aided:0.981[NE, NE]Unaided:0.838[NE, NE]Difference:0.143[NE, NE] | Aided:0.417[NE, NE]Unaided:0.458[NE, NE]Difference:-0.042[NE, NE] | Aided:0.876[0.761, 0.990]Unaided:0.761[0.589, 0.933]Difference:0.115[-0.025, 0.255] | | | Ground Truth (pt) | 16 | 5 | | | | | | AI Assisted (Reads) | 55 | 8 | | | | | | Unassisted (Reads) | 42 | 21 | | | | | Northeast | Total: 13 patients | Positive | Negative | Aided:0.878[NE, NE]Unaided:0.711[NE, NE]Difference:0.167[NE, NE] | Aided:N/A [1]Unaided:N/A [1]Difference:N/A [1] | Aided:0.878[0.760, 0.996]Unaided:0.678[0.448, 0.908]Difference:0.200[0.003, 0.397] | | | Ground Truth (pt) | 12 | 1 | | | | | | AI Assisted (Reads) | 35 | 4 | | | | | | Unassisted (Reads) | 28 | 11 | | | | | South | Total: 34 patients | Positive | Negative | Aided:0.976[0.938, 1.000]Unaided:0.717[0.552, 0.882]Difference:0.259[0.118, 0.401] | Aided:0.200[NE, NE]Unaided:0.600[NE, NE]Difference:-0.400[NE, NE] | Aided:0.945[0.875, 1.000]Unaided:0.661[0.496, 0.825]Difference:0.285[0.150, 0.419] | | | Ground Truth (pt) | 28 | 6 | | | | | | AI Assisted (Reads) | 97 | 5 | | | | | | Unassisted (Reads) | 71 | 31 | | | | | West | Total: 43 patients | Positive | Negative | Aided:0.842[0.693, 0.991]Unaided:0.625[0.433, 0.817]Difference:0.217[0.064, 0.370] | Aided:0.250[0.060, 0.440]Unaided:0.606[0.390, 0.821]Difference:-0.356[-0.532, -0.179] | Aided:0.814[0.696, 0.932]Unaided:0.541[0.353, 0.729]Difference:0.273[0.173, 0.373] | | | Ground Truth (pt) | 26 | 17 | | | | | | AI Assisted (Reads) | 104 | 25 | | | | | | Unassisted (Reads) | 68 | 61 | | | | | Subgroup – Bowel Prep [2] | | | | | | | | Fair | Total: 36 patients | Positive | Negative | Aided:0.922[0.844, 1.000] | Aided:0.037[NE, NE] | Aided:0.917[0.840, 0.994] | | | Ground Truth (pt) | 30 | 6 | | | | | | AI Assisted (Reads) | 101 | 7 | | | | K254230 510(k) Summary Page 19 {23} K254230 CapsoVision empacking strength innovation | | Total / Positive / Negative Patient/Read (3/Video/Condition) Count | | | Sensitivity | Specificity | Diagnostic Yield | | --- | --- | --- | --- | --- | --- | --- | | | Unassisted (Reads) | 72 | 36 | Unaided:0.690[0.524, 0.856]Difference:0.232[0.065, 0.398] | Unaided:0.278[NE, NE]Difference:-0.241[NE, NE] | Unaided:0.679[0.516, 0.843]Difference:0.238[0.074, 0.402] | | Good | Total: 74 patients | Positive | Negative | Aided:0.927[0.871, 0.982]Unaided:0.722[0.609, 0.835]Difference:0.205[0.128, 0.281] | Aided:0.333[NE, NE]Unaided:0.615[NE, NE]Difference:-0.283[NE, NE] | Aided:0.840[0.746, 0.935]Unaided:0.613[0.468, 0.759]Difference:0.227[0.147, 0.306] | | | Ground Truth (pt) | 51 | 23 | | | | | | AI Assisted (Reads) | 188 | 34 | | | | | | Unassisted (Reads) | 134 | 88 | | | | | Subgroup – Age | | | | | | | | 2-5 | Total: 5 patients | Positive | Negative | Aided:1.000 [NE, NE]Unaided:0.833 [NE, NE]Difference:0.167 [NE, NE] | Aided:\( ^{1} \)Unaided:\( ^{1} \)Difference:\( ^{1} \) | Aided:1.000[NE, NE]Unaided:0.833[NE, NE]Difference:0.167[NE, NE] | | | Ground Truth (pt) | 5 | 0 | | | | | | AI Assisted (Reads) | 15 | 0 | | | | | | Unassisted (Reads) | 11 | 4 | | | | | 6-12 | Total: 12 patients | Positive | Negative | Aided:0.889[0.761, 1.000]Unaided:0.815[0.629, 1.000]Difference:0.074[-0.139, 0.288] | Aided:0.167[NE, NE]Unaided:0.444[NE, NE]Difference:-0.278[NE, NE] | Aided:0.892[NE, NE]Unaided:0.742[NE, NE]Difference:0.150[NE, NE] | | | Ground Truth (pt) | 8 | 4 | | | | | | AI Assisted (Reads) | 30 | 6 | | | | | | Unassisted (Reads) | 25 | 11 | | | | | 13-21 | Total: 16 patients | Positive | Negative | Aided:0.875[NE, NE]Unaided:0.646[NE, NE]Difference:0.229[NE, NE] | Aided:0.278[NE, NE]Unaided:0.542[NE, NE]Difference:-0.264[NE, NE] | Aided:0.795[0.649, 0.940]Unaided:0.554[0.291, 0.816]Difference:0.241[0.087, 0.395] | | | Ground Truth (pt) | 9 | 7 | | | | | | AI Assisted (Reads) | 37 | 11 | | | | | | Unassisted (Reads) | 24 | 24 | | | | | 22-39 | Total: 8 patients | Positive | Negative | Aided:0.556[NE, NE] | Aided:0.389[0.000, 0.789] | Aided:0.688[NE, NE] | | | Ground Truth (pt) | 3 | 5 | | | | | | AI Assisted (Reads) | 14 | 10 | | | | K254230 510(k) Summary Page 20 {24} CapsoVision empowering through innovation K254230 | | Total / Positive / Negative Patient/Read (3/Video/Condition) Count | | | Sensitivity | Specificity | Diagnostic Yield | | --- | --- | --- | --- | --- | --- | --- | | | Unassisted (Reads) | 6 | 18 | Unaided:0.444[NE, NE]Difference:0.111[NE, NE] | Unaided:0.833[0.444, 1.000]Difference:-0.444[-0.788, -0.101] | Unaided:0.417[NE, NE]Difference:0.271[NE, NE] | | 40-59 | Total: 13 patients | Positive | Negative | Aided:0.944[0.862, 1.000]Unaided:0.569[0.302, 0.837]Difference:0.375[0.149, 0.601] | Aided:0.000[NE, NE]Unaided:0.500[NE, NE]Difference:-0.500[NE, NE] | Aided:0.944[0.862, 1.000]Unaided:0.507[0.230, 0.784]Difference:0.438[0.192, 0.683] | | | Ground Truth (pt) | 11 | 2 | | | | | | AI Assisted (Reads) | 37 | 2 | | | | | | Unassisted (Reads) | 21 | 18 | | | | | 60+ | Total: 57 patients | Positive | Negative | Aided:0.965[0.935, 0.995]Unaided:0.774[0.659, 0.888]Difference:0.191[0.104, 0.278] | Aided:0.206[NE, NE]Unaided:0.517[NE, NE]Difference:-0.311[NE, NE] | Aided:0.931[0.876, 0.987]Unaided:0.722[0.591, 0.852]Difference:0.210[0.118, 0.302] | | | Ground Truth (pt) | 46 | 11 | | | | | | AI Assisted (Reads) | 158 | 13 | | | | | | Unassisted (Reads) | 122 | 49 | | | | | Subgroup – Gender | | | | | | | | Male | Total: 13 patients | Positive | Negative | Aided:0.833[NE, NE]Unaided:0.750[NE, NE]Difference:0.083[NE, NE] | Aided:0.556[NE, NE]Unaided:0.778[NE, NE]Difference:-0.222[NE, NE] | Aided:0.833[NE, NE]Unaided:0.650[NE, NE]Difference:0.183[NE, NE] | | | Ground Truth (pt) | 9 | 4 | | | | | | AI Assisted (Reads) | 31 | 8 | | | | | | Unassisted (Reads) | 22 | 17 | | | | | Female | Total: 18 patients | Positive | Negative | Aided:0.933[NE, NE]Unaided:0.692[NE, NE]Difference:0.242[NE, NE] | Aided:0.233[NE, NE]Unaided:0.667[NE, NE]Difference:-0.433[NE, NE] | Aided:0.833[NE, NE]Unaided:0.489[NE, NE]Difference:0.344[NE, NE] | | | Ground Truth (pt) | 11 | 7 | | | | | | AI Assisted (Reads) | 46 | 8 | | | | | | Unassisted (Reads) | 31 | 23 | | | | | Unknown | Total: 80 patients | Positive | Negative | Aided:0.943[0.896, 0.989] | Aided:0.244[0.086, 0.403] | Aided:0.898 [0.828, 0.967] | | | Ground Truth (pt) | 62 | 18 | | | | | | AI Assisted (Reads) | 214 | 26 | | | | K254230 510(k) Summary Page 21 {25} CapsoVision empacking strength innovation K254230 | | Total / Positive / Negative Patient/Read (3/Video/Condition) Count | | | Sensitivity | Specificity | Diagnostic Yield | | --- | --- | --- | --- | --- | --- | --- | | | Unassisted (Reads) | 156 | 84 | Unaided:0.730[0.611, 0.850]Difference:0.212[0.123, 0.302] | Unaided:0.511[0.243, 0.779]Difference:-0.267[-0.532, -0.002] | Unaided:0.666 [0.534, 0.798]Difference:0.232 [0.145, 0.319] | | Subgroup – Indications for Imaging [3] | | | | | | | | Bleeding | Total: 63 patients | Positive | Negative | Aided:0.947[0.901, 0.993]Unaided:0.737[0.610, 0.864]Difference:0.210[0.119, 0.301] | Aided:0.239[0.069, 0.409]Unaided:0.583[0.316, 0.851]Difference:-0.344[-0.597, -0.092] | Aided:0.904[0.834, 0.975]Unaided:0.671[0.534, 0.808]Difference:0.233[0.140, 0.325] | | | Ground Truth (pt) | 50 | 13 | | | | | | AI Assisted (Reads) | 170 | 19 | | | | | | Unassisted (Reads) | 123 | 66 | | | | | No Bleeding | Total: 45 patients | Positive | Negative | Aided:0.899[0.809, 0.989]Unaided:0.656[0.469, 0.844]Difference:0.243[0.044, 0.442] | Aided:0.389[NE, NE]Unaided:0.678[NE, NE]Difference:-0.289[NE, NE] | Aided:0.843[0.724, 0.961]Unaided:0.571[0.382, 0.760]Difference:0.272[0.128, 0.416] | | | Ground Truth (pt) | 31 | 14 | | | | | | AI Assisted (Reads) | 115 | 20 | | | | | | Unassisted (Reads) | 82 | 53 | | | | | Unknown | Total: 3 patients | Positive | Negative | Aided:\( ^{3} \)Unaided:\( ^{3} \)Difference:\( ^{3} \) | Aided:\( ^{3} \)Unaided:\( ^{3} \)Difference:\( ^{3} \) | Aided:\( ^{3} \)Unaided:\( ^{3} \)Difference:\( ^{3} \) | | | Ground Truth (pt) | 1 | 2 | | | | | | AI Assisted (Reads) | 6 | 3 | | | | | | Unassisted (Reads) | 4 | 5 | | | | | Subgroup – Finding of Bleeding | | | | | | | | Bleeding | Total: 72 patients | Positive | Negative | Aided:0.941[0.900, 0.982]Unaided:0.770[0.660, 0.879]Difference:0.171[0.083, 0.259] | Aided:0.150[NE, NE]Unaided:0.506[NE, NE]Difference:-0.356[NE, NE] | Aided:0.928[0.877, 0.978]Unaided:0.729[0.611, 0.846]Difference:0.199[0.108, 0.290] | | | Ground Truth (pt) | 60 | 12 | | | | | | AI Assisted (Reads) | 201 | 15 | | | | | | Unassisted (Reads) | 162 | 54 | | | | | No Bleeding | Total: 39 patients | Positive | Negative | Aided:0.892[NE, NE]Unaided:0.531[NE, NE]Difference:0.361[NE, NE] | Aided:0.340[NE, NE]Unaided:0.604[NE, NE]Difference:-0.264[NE, NE] | Aided:0.785[0.643, 0.927]Unaided:0.474[0.300, 0.648]Difference:0.311[0.193, 0.429] | | | Ground Truth (pt) | 22 | 17 | | | | | | AI Assisted (Reads) | 90 | 27 | | | | | | Unassisted (Reads) | 47 | 70 | | | | K254230 510(k) Summary Page 22 {26} CapsoVision empowering through innovation K254230 | | Total / Positive / Negative Patient/Read (3/Video/Condition) Count | Sensitivity | Specificity | Diagnostic Yield | | --- | --- | --- | --- | --- | [1]: Insufficient number for calculation. [2]: Analysis not conducted for the bowel prep poor subgroup as it comprised of one patient. [3] Analysis not conducted for the unknown indication as it comprises of less than 5 patients. Table 9. Summary of Reading Time Analysis – Patient Level | | Small Bowel Reading Time [minutes] | | | | --- | --- | --- | --- | | Overall | Total: 111 patients | Read Count (3/pt/condition) | Aided: 34.393 [18.013, 50.773]Unaided: 45.687 [29.306, 62.067]Difference: -11.294 [-13.545, -9.043] | | | AI Assisted (Reads) | 329 | | | | Unassisted (Reads) | 331 | | | Subgroup – U.S. Region | | | | | Midwest | Total: 21 patients | Read Count (3/pt/condition) | Aided: 37.244 [17.831, 56.658]Unaided: 50.879 [31.465, 70.293]Difference: -13.635 [-19.439, -7.831] | | | AI Assisted (Reads) | 63 | | | | Unassisted (Reads) | 63 | | | Northeast | Total: 13 patients | Read Count (3/pt/condition) | Aided: 32.368 [20.198, 44.538]Unaided: 44.684 [32.514, 56.853]Difference: -12.316 [-18.366, -6.266] | | | AI Assisted (Reads) | 38 | | | | Unassisted (Reads) | 38 | | | South | Total: 34 patients | Read Count (3/pt/condition) | Aided: 36.022 [20.613, 51.431]Unaided: 47.717 [32.305, 63.129]Difference: -11.695 [-15.994, -7.395] | | | AI Assisted (Reads) | 101 | | | | Unassisted (Reads) | 102 | | | West | Total: 43 patients | Read Count (3/pt/condition) | Aided: 32.160 [15.191, 49.129]Unaided: 41.734 [24.765, 58.704]Difference: -9.574 [-13.036, -6.112] | | | AI Assisted (Reads) | 127 | | | | Unassisted (Reads) | 128 | | | Subgroup – Bowel Prep [1] | | | | | Fair | Total: 36 patients | Read Count (3/pt/condition) | Aided: 41.224 [23.043, 59.405]Unaided: 46.193 [28.010, 64.376]Difference: -4.969 [-8.803, -1.135] | | | AI Assisted (Reads) | 107 | | | | Unassisted (Reads) | 108 | | | Poor | Total: 1 patient | Read Count (3/pt/condition) | Aided: N/A [1]Unaided: N/A [1]Difference: N/A [1] | | | AI Assisted (Reads) | 3 | | | | Unassisted (Reads) | 3 | | | Good | Total: 74 patients | Read Count (3/pt/condition) | Aided: 30.420 [14.593, 46.247]Unaided: 45.258 [29.431, 61.084]Difference: -14.837 [-17.582, -12.093] | | | AI Assisted (Reads) | 219 | | | | Unassisted (Reads) | 220 | | | Subgroup – Age | | | | | 2-5 | Total: 5 patients | Read Count (3/pt/condition) | Aided: 26.116 [-67.887, 120.119]Unaided: 36.751 [-63.543, 137.044]Difference: -10.635 [-16.584, -4.687] | | | AI Assisted (Reads) | 14 | | | | Unassisted (Reads) | 15 | | | 6-12 | Total: 12 patients | Read Count (3/pt/condition) | Aided: 37.730 [13.860, 61.601]Unaided: 40.369 [16.499, 64.240]Difference: -2.639 [-9.028, 3.750] | | | AI Assisted (Reads) | 36 | | | | Unassisted (Reads) | 36 | | | 13-21 | Total: 16 patients | Read Count (3/pt/condition) | Aided: 31.701 [8.017, 55.385]Unaided: 45.557 [21.899, 69.216]Difference: -13.856 [-19.317, -8.396] | | | AI Assisted (Reads) | 47 | | | | Unassisted (Reads) | 47 | | | 22-39 | Total: 8 patients | Read Count (3/pt/condition) | Aided: 26.516 [4.864, 48.167]Unaided: 40.182 [18.531, 61.834]Difference: -13.667 [-22.403, -4.930] | | | AI Assisted (Reads) | 24 | | | | Unassisted (Reads) | 24 | | | 40-59 | Total: 13 patients | Read Count (3/pt/condition) | Aided: 41.919 [21.102, 62.737]Unaided: 47.047 [26.230, 67.865]Difference: -5.128 [-12.588, 2.331] | | | AI Assisted (Reads) | 39 | | | | Unassisted (Reads) | 39 | | | 60+ | Total: 57 patients | Read Count (3/pt/condition) | Aided: 35.224 [19.875, 50.572] | K254230 510(k) Summary Page 23 {27} CapsoVision empowering through innovation K254230 | | Small Bowel Reading Time [minutes] | | | | --- | --- | --- | --- | | | AI Assisted (Reads) | 169 | Unaided: 48.831 [33.481, 64.181] Difference: -13.608 [-16.842, -10.373] | | | Unassisted (Reads) | 170 | | | Subgroup – Gender | | | | | Male | Total: 13 patients | Read Count (3/pt/condition) | Aided: 36.204 [19.564, 52.845] Unaided: 42.906 [26.248, 59.564] Difference: -6.702 [-13.564, 0.161] | | | AI Assisted (Reads) | 37 | | | | Unassisted (Reads) | 39 | | | Female | Total: 18 patients | Read Count (3/pt/condition) | Aided: 32.180 [12.945, 51.416] Unaided: 44.336 [25.109, 63.564] Difference: -12.156 [-17.294, -7.019] | | | AI Assisted (Reads) | 54 | | | | Unassisted (Reads) | 53 | | | Unknown | Total: 80 patients | Read Count (3/pt/condition) | Aided: 35.082 [19.506, 50.658] Unaided: 47.013 [31.436, 62.590] Difference: -11.931 [-14.602, -9.261] | | | AI Assisted (Reads) | 238 | | | | Unassisted (Reads) | 239 | | | Subgroup – Indications for Imaging^{[2]} | | | | | Bleeding | Total: 63 patients | Read Count (3/pt/condition) | Aided: 34.739 [18.210, 51.268] Unaided: 47.235 [30.704, 63.765] Difference: -12.496 [-15.656, -9.335] | | | AI Assisted (Reads) | 187 | | | | Unassisted (Reads) | 188 | | | No Bleeding | Total: 45 patients | Read Count (3/pt/condition) | Aided: 32.866 [18.410, 47.321] Unaided: 42.120 [27.665, 56.576] Difference: -9.254 [-12.567, -5.942] | | | AI Assisted (Reads) | 133 | | | | Unassisted (Reads) | 134 | | | Unknown | Total: 3 patients | Read Count (3/pt/condition) | Aided: N/A^{[1]} Unaided: N/A^{[1]} Difference: N/A^{[1]} | | | AI Assisted (Reads) | 9 | | | | Unassisted (Reads) | 9 | | | Subgroup – Finding of Bleeding | | | | | Bleeding | Total: 72 patients | Read Count (3/pt/condition) | Aided: 37.785 [20.268, 55.302] Unaided: 45.673 [28.156, 63.190] Difference: -7.888 [-10.645, -5.131] | | | AI Assisted (Reads) | 213 | | | | Unassisted (Reads) | 214 | | | No Bleeding | Total: 39 patients | Read Count (3/pt/condition) | Aided: 29.104 [13.486, 44.722] Unaided: 46.603 [30.984, 62.222] Difference: -17.499 [-21.315, -13.683] | | | AI Assisted (Reads) | 116 | | | | Unassisted (Reads) | 117 | | [1]: Analysis not conducted for the bowel prep poor subgroup as it comprised of one patient. [2] Analysis not conducted for the unknown indication as it comprises of less than 5 patients. ## Conclusion The results of the standalone and clinical performance studies support that the proposed device functions as intended, and that use of the device achieves decreased small bowel capsule endoscopy reading time relative to “unassisted” review. CapsoVision believes that the collective results of non-clinical and clinical performance testing support the safety and performance of the proposed device for its Indications for Use, and that that the proposed device is at least as safe and as effective as the predicate device for the same Intended Use. K254230 510(k) Summary Page 24