Company Spotlight: Ever Fortune.AI APPEN-CT-100 - Flagging Appendicitis

Sam: An AI system just proved it can flag appendicitis on abdominal CT scans with a sensitivity of 0.928 and a specificity of 0.932, which could really reshape emergency room triage.

Maya: Today we are unpacking the FDA 510(k) summary K253163 for the EFAI Ersuite CT Appendicitis Assessment System, published by the U.S. Food and Drug Administration on April 22, 2026.

Sam: And a quick reminder upfront: our voices are AI-generated, and this episode is a summary of a publicly available regulatory document, not company marketing material.

Maya: This is a Company Spotlight episode, but we remain completely independent. If a detail is not in this specific 510(k) clearance summary, we will not discuss it. Everything we cover comes straight from the FDA text.

Sam: Let us start with the basics from the clearance letter itself. The application was submitted by Ever Fortune.AI, Co., Ltd., located in Taichung City, Taiwan. The contact person listed is Ti-Hao Wang, their Chief Technology Officer, along with Joseph Chang as the applicant contact.

Maya: The FDA classifies this product, which is commonly referred to in the document as EFAI APPENDICITIS, as a Class II device. The regulation name is Radiological Computer Aided Triage And Notification Software, under regulation number 21 CFR 892.2080, with the product code QAS.

Sam: The clearance letter, signed by Jessica Lamb, Assistant Director of the Division of Radiological Imaging Devices and Electronic Products, contains a lot of standard regulatory stipulations for a Class II device.

Maya: It does. It reminds the sponsor that they are subject to general controls, including annual registration, listing of devices, and good manufacturing practice. Specifically, it highlights compliance with the Quality Management System Regulation, or QMSR, under 21 CFR Part 820.

Sam: It even calls out specific ISO 13485 clauses they must follow, like clause 7.3 for design controls and clause 8.3 for nonconforming product.

Maya: The letter also mandates compliance with the Unique Device Identification System rule, meaning the device must bear a unique device identifier on its label and package, unless an exception applies. Information must also be submitted to the Global Unique Device Identification Database under 21 CFR Part 830 Subpart E.

Sam: The FDA letter also mentions medical device reporting requirements under 21 CFR Part 803 for adverse events, and provides contact information for the Division of Industry and Consumer Education if the company has questions. It is a very thorough regulatory framework.

Maya: Okay, so that covers the administrative background. But what exactly is the EFAI Ersuite CT Appendicitis Assessment System designed to do for patients and clinicians on the ground?

Sam: According to the indications for use, it is a software system indicated for use in the analysis of contrast-enhanced abdominal CT images. Its intended use is to assist hospital networks and trained clinicians in workflow triage by flagging and communicating suspected positive findings of appendicitis.

Maya: So a patient comes into the emergency department with severe abdominal pain. They get a contrast-enhanced CT scan. The software uses deep learning techniques to automatically analyze that scan.

Sam: Precisely. It uses an artificial intelligence algorithm with a database of images to analyze the current scan in parallel to the ongoing standard of care image interpretation.

Maya: Operating in parallel is a key concept here. The document explicitly states that the device does not remove cases from the worklist queue. It does not deprioritize anything.

Sam: That is correct. Standard of care remains the default option for all cases. The device provides a passive notification through the PACS or RIS workstation to trained clinicians when it identifies features suggestive of appendicitis.

Maya: The goal, as the summary states, is that through the use of EFAI APPENDICITIS, a trained clinician is able to review studies with features suggestive of appendicitis earlier than in the standard of care workflow.

Sam: But the document is very strict about the limitations of this software. It states repeatedly that the device does not alter the original medical image and is not intended to be used as a diagnostic device.

Maya: It also notes that the device does not mark, highlight, or direct users attention to a specific location on the original contrast-enhanced abdominal CT.

Sam: The fact that it does not mark the image is important for the boardroom. Hospital IT leaders evaluating this system should know it is a passive alert system, not an image annotation tool.

Maya: Right. The results are intended to be used in conjunction with other patient information and based on the clinician's professional judgment. The document makes it clear: notified clinicians are responsible for viewing full images per the standard of care.

Sam: To get this FDA clearance, Ever Fortune.AI had to prove substantial equivalence to a predicate device. They used a product called BriefCase, which was cleared under a previous premarket notification, K193298.

Maya: BriefCase is also a Class II radiological computer-assisted triage and notification software, sharing the same regulation number and product code.

Sam: But they do not look for the same thing. The subject device processes contrast-enhanced abdominal CT images to detect appendicitis. BriefCase processes abdominal CT images without restriction on contrast use and is designed to detect Intra-abdominal Free Gas.

Maya: Despite these differences in clinical findings and contrast use, the FDA determined that both devices analyze the same abdominal CT region, and these differences do not impact the underlying technology or raise new performance concerns.

Sam: They both accept DICOM format images, use artificial intelligence algorithms to identify suspected positive findings, and generate case-level notifications. They both target the abdomen as the anatomical region of interest.

Maya: There are a few subtle differences in their user population, though. EFAI APPENDICITIS is intended for a trained clinician, whereas BriefCase specifies a trained radiologist.

Sam: That is an interesting distinction. A trained clinician could theoretically include an emergency department physician, not just a radiologist in the reading room.

Maya: The document also points out a notable technical difference in how they deliver their outputs. The predicate device, BriefCase, provides notifications that include compressed preview images.

Sam: The summary notes that BriefCase presents a small, compressed, black and white preview image labeled as not for diagnostic use.

Maya: In contrast, the proposed device, EFAI APPENDICITIS, does not provide preview images. It provides notification as a text-based JSON file.

Sam: A text-based JSON file. That means the clinician gets a pure text alert in the system rather than a visual thumbnail.

Maya: Correct. But the FDA concluded this difference does not change the fundamental technology of using AI to generate case-level notifications for triage. Both operate without altering the original image or removing cases from the worklist.

Sam: Let us move to the performance data. How did they prove the system actually works in practice?

Maya: The document details software verification and validation testing performed in accordance with IEC 62304:2006/A1:2015 for medical device software life cycle processes.

Sam: They also followed FDA guidance documents regarding the content of premarket submissions for device software functions, as well as cybersecurity in medical devices.

Maya: Beyond the software standards, they conducted a retrospective, blinded, multisite validation study using pre-specified primary and secondary endpoints.

Sam: So this was a retrospective study, meaning they used historical data rather than testing it live in a prospective clinical trial.

Maya: Yes. The study evaluated 300 consecutively collected contrast-enhanced abdominal CT cases from multiple U.S. clinical sites. They ensured there was only one case per patient, and importantly, none of these cases were used in model development or internal testing.

Sam: That separation of training data and testing data is a fundamental rule for AI validation, and the FDA summary specifically confirms they followed it. But how did they determine the actual ground truth for those 300 scans? Did they just use the original radiology reports?

Maya: No, they established a rigorous reference standard. The appendicitis status of each case was independently assessed by three U.S. board-certified radiologists. The final ground truth was determined by majority agreement among those three.

Sam: Using three board-certified radiologists for a 300-case study is a significant investment in establishing ground truth. By using a majority agreement, Ever Fortune.AI ensured they were not just matching a single doctor's potentially subjective opinion.

Maya: Out of the 300 cases, the majority agreement yielded 139 positive cases and 161 negative cases. That is a fairly balanced dataset for evaluating the algorithm.

Sam: The primary endpoints for the study were sensitivity and specificity. The pre-specified performance goal for both metrics was 0.8.

Maya: And it exceeded the goals. The device demonstrated a sensitivity of 0.928, with a 95% confidence interval of 0.873 to 0.960.

Sam: And for specificity?

Maya: The specificity was 0.932, with a 95% confidence interval of 0.882 to 0.961.

Sam: Those are robust numbers. To put that in context, the document compares these results directly to the predicate device, BriefCase.

Maya: Yes, the predicate device had a reported sensitivity of 0.910 and a specificity of 0.889. EFAI APPENDICITIS demonstrated substantial equivalence while exceeding its own targets.

Sam: But accuracy is only half the battle in triage. If the AI takes too long to process the scan, the notification is useless to the emergency room workflow.

Maya: The validation study addressed this with a secondary endpoint focused on mean system processing time per case.

Sam: And what was the processing time?

Maya: The mean system processing time was 3.37 minutes, with a 95% confidence interval between 3.34 and 3.40 minutes. The summary notes this was significantly less than their pre-specified goal.

Sam: A turnaround time of 3.37 minutes is very fast. It means the text-based JSON alert could be sitting in the PACS system just a few minutes after the patient gets off the CT table.

Maya: The validation did not stop at the top-line numbers. They performed subgroup analyses to ensure the software worked reliably across different variables.

Sam: The document says they looked at performance across sex, age, race or ethnicity, CT manufacturer, and CT slice thickness.

Maya: According to the summary, these subgroup analyses showed consistent performance, underscoring reliability across diverse populations and imaging conditions.

Sam: They also tested the system against a long list of potential confounding factors. Diagnosing appendicitis can be tricky because so many other abdominal issues can mimic it on a scan.

Maya: The document provides a very specific list of what they evaluated. They looked at cases with different levels of appendiceal thickening, as well as other appendiceal findings.

Sam: They also tested it against inflammatory or infectious conditions, diverticulitis, urinary findings, and gynecological findings.

Maya: Furthermore, they evaluated vascular or ischemic conditions, fluid collections, neoplastic or metastatic lesions, and postoperative findings. They even evaluated cases with suboptimal image quality issues.

Sam: That is an extensive list of real-world scenarios that an ER scanner sees every day. Based on the 510(k) summary, the device held up well enough across these conditions to support the substantial equivalence claim.

Maya: Yes. The summary concludes that based on the indications for use, technological characteristics, and performance testing, EFAI APPENDICITIS is substantially equivalent to the predicate device in terms of safety, efficacy, and performance.

Sam: It is a fascinating step forward for acute-care triage. We are moving beyond head CTs for stroke and chest scans for pneumothorax, and bringing targeted AI directly into the acute abdomen.

Maya: However, it is vital to remember the limitations stated in the document. This clearance is based on a retrospective dataset of 300 cases, not a prospective trial. And it relies on passive, text-based notifications without marking the image itself.

Sam: Exactly. The clinician still has to open the scan, find the appendix, and make the diagnosis based on the standard of care.

Maya: Here is a crisp recap: The FDA has cleared the EFAI Ersuite CT Appendicitis Assessment System as a triage and notification tool for contrast-enhanced abdominal CTs. In a 300-case retrospective validation study, it achieved a sensitivity of 0.928 and a specificity of 0.932, with a mean processing time of 3.37 minutes.

Sam: And as always, a reminder that the source document for this episode is linked in the show notes, and that this podcast is for informational purposes only and is not medical advice.