20 November 2026 · 14 min

Company Spotlight: Ever Fortune.AI ETT-XR-100 - AI Triage for Misplaced Endotracheal Tubes

Radiology & ImagingCritical & Emergency CareFDA Clearances

This episode unpacks the FDA 510(k) clearance summary for Ever Fortune.AI's EFAI ETTXR system, an AI tool designed to flag malpositioned endotracheal tubes on adult chest X-rays. We break down the clinical testing data, the algorithm's specific measurement thresholds, and crucial limitations regarding esophageal intubation and double lumen tubes.

Key points

Source: FDA 510(k) summary K242821: EFAI Chestsuite XR Malpositioned ETT Assessment System (ETT-XR-100) (Ever Fortune.AI) - U.S. Food and Drug Administration, 2025

This episode is an AI-generated conversation summarising a public document; the hosts' voices are synthetic. It is for information only and is not medical advice. Always refer to the original source.

Company: Ever Fortune.AI · Product: EFAI Chestsuite XR Malpositioned ETT Assessment System (ETT-XR-100) · FDA 510(k) K242821

Your company here. This podcast is looking for its first sponsors: reach clinicians, health-system leaders and medtech and pharma teams following AI in medicine. Sponsorship options and rates →

Transcript

Sam: Welcome to AI in Medicine - Smart Summaries. I am Sam, and today we are looking at a new algorithm that scans portable bedside chest X-rays to verify endotracheal tube depth, and the testing data shows it caught malpositions with a sensitivity of 0.890.

Maya: I am Maya. First things first: our voices are AI-generated, and this is a summary of a public document. Today, we are unpacking an FDA 510(k) summary published on February 20, 2025, for a product from Ever Fortune.AI.

Sam: And since this is a Company Spotlight episode, we need to set the ground rules early. This is not an advert. We are completely independent, and our conversation is based strictly on the FDA's public 510(k) summary. We will not be pulling in any external marketing material, press releases, or company history that is not printed right here in this specific document.

Maya: Absolutely. The official trade name of the product is the EFAI CHESTSUITE XR MALPOSITIONED ETT ASSESSMENT SYSTEM. But the summary notes the common name is EFAI ETTXR. We will stick with that for the sake of our sanity today.

Sam: I appreciate that. So, let us get into the problem this solves. Unrecognized endotracheal tube malpositioning into the bronchus or esophagus causes immediate, life-threatening complications. Instant algorithmic verification of these tubes could theoretically relieve some serious bottlenecks in ED and ICU workflows.

Maya: That is the exact clinical stake. The EFAI ETTXR is a radiological computer-aided triage and notification software. It is indicated for use in the analysis of chest X-ray images in adults. The intended users are hospital networks and appropriately trained medical specialists who need help with workflow triage.

Sam: Workflow triage. That is an important distinction. This software is not making a primary diagnosis or replacing the doctor, correct?

Maya: Correct. It analyzes cases using deep learning algorithms to identify suspected positive cases of a vertically malpositioned endotracheal tube, or ETT, in relation to the carina. When it finds a suspected issue, it makes case-level output available to a workstation to flag and communicate the finding.

Sam: Let us define the anatomy really quickly for our business listeners. The endotracheal tube is the breathing tube placed down a patient's airway to help them breathe. The carina is the ridge at the base of the windpipe where it splits into the left and right lungs. Getting the depth of that tube right is critical. How exactly is the AI measuring this?

Maya: The document gives us the precise rules. Findings are flagged when the distal tip of the tube is assessed as being more than 7 cm above the carina, less than 3 cm above the carina, or when it is below the carina altogether, which means it has slipped into the right or left mainstem bronchus.

Sam: Wait, really? More than 7 cm or less than 3 cm above the carina. That is a tight window. If it is 4 cm or 5 cm above, the AI stays quiet.

Maya: Yes, and this brings up a fascinating part of the 510(k) process. To get clearance, Ever Fortune.AI had to prove substantial equivalence to a legally marketed predicate device. The predicate they used is called BriefCase, which was cleared under a different submission.

Sam: I see the side-by-side comparison table in the document. And there are some notable differences in those measurement rules we just talked about.

Maya: There are. The predicate device, BriefCase, flags cases when the tip is more than 5 cm above the carina, or less than 2 cm above the carina. The EFAI ETTXR uses the 7 cm and 3 cm thresholds.

Sam: That is a significant difference for a hospital buyer to note. Different algorithms might have entirely different rules for what they consider out of bounds. Does the document explain why they shifted to 7 cm and 3 cm?

Maya: It does not. It simply presents the comparison and later concludes that despite minor technical differences, the new device is substantially equivalent in terms of safety and effectiveness.

Sam: There is another difference in that table, too. The user population.

Maya: Right. While the EFAI ETTXR is indicated for adults, the predicate BriefCase was indicated for adults or transitional adolescents aged 18 and older. The new device does not mention transitional adolescents at all.

Sam: Let us talk about what actually happens on the screen when the AI flags an image. Does it draw a bright red circle around the misplaced tube so the doctor sees it immediately?

Maya: Absolutely not. The summary is extremely firm on this point. It states the software is not intended to direct attention to specific portions of an image or to anomalies of an image. It does not mark or highlight the original chest radiographs.

Sam: So how does the radiologist actually get the alert?

Maya: It provides a passive notification through the PACS or workstation. And here is another technical difference from the predicate device. The older BriefCase device provides compressed preview images within their notification. But the EFAI ETTXR provides its notification as a text based JSON file.

Sam: A text-based JSON file? That is really interesting from an IT integration perspective. So the radiologist gets a system text notification that basically says there is a suspected malposition, but they still have to open the original image and find the tube themselves.

Maya: Exactly. The results are not intended to be used on a stand-alone basis for clinical decision-making. The notified clinicians are entirely responsible for viewing the full images per the standard of care.

Sam: And the summary is clear that no cases are removed from the worklist or deprioritized. It only bumps suspected issues up. It does not hide the scans it thinks are fine.

Maya: Correct. It explicitly says it is not intended to rule out malpositioned ETT or otherwise preclude clinical assessment of chest radiographs.

Sam: Okay, let us get into the proof. How did Ever Fortune.AI validate that this algorithm actually works in the real world?

Maya: They conducted a retrospective, blinded, multisite clinical validation study. The dataset consisted of 940 studies consecutively collected from multiple clinical sites across the United States. And they made sure that each patient included only one study.

Sam: Retrospective is the key word there. This was run on historical data, not running live in an active emergency room or ICU during the trial. What did the demographics of those 940 studies look like?

Maya: The mean age of the cases was 57.9 years. The population split was 49.57% females and 50.43% males. The study included cases from White, Black or African American, Hispanic, Asian, Multiracial, and other races or ethnicities.

Sam: And how did they determine the ground truth? Because an AI is only as good as the answer key it is graded against. Who decided if the tube was actually misplaced in those 940 images?

Maya: The reference standard was based on the majority agreement among three U.S. board-certified radiologists. Out of the 940 studies, this panel classified 259 cases as positive for a malpositioned tube, and 681 cases as negative.

Sam: Let us break down those 681 negative cases. Were they all just perfectly placed tubes?

Maya: Not exactly. The 681 negative cases included 316 correctly positioned tubes, and 365 cases with no endotracheal tube present at all.

Sam: That is a very smart way to test it. The AI has to prove it does not hallucinate a misplaced tube when there is not even a tube in the patient's airway to begin with. What hardware were these X-rays taken on?

Maya: The summary lists several manufacturers to demonstrate variety. The images were obtained from Philips, Canon, Carestream Health, Samsung Electronics, Siemens, and others.

Sam: Okay, the moment of truth. What were the actual performance numbers? The document says they set an 80% performance goal for their primary endpoints. Did they clear that bar?

Maya: They did. Evaluating exclusively on the AP view, which is the standard frontal chest X-ray, the device demonstrated a sensitivity of 0.890 and a specificity of 0.935.

Sam: Those are solid numbers. A sensitivity of 0.890 means it is catching a strong majority of the actual misplacements. And a specificity of 0.935 means it correctly identifies the vast majority of properly placed or absent tubes. It is not constantly crying wolf.

Maya: Exactly. They also tracked a secondary endpoint, which was the system processing time. The observed processing time per study was 2.49 minutes on average.

Sam: 2.49 minutes from the time the algorithm gets the X-ray to the time the JSON text file hits the worklist. In a busy hospital, getting a heads-up in under three minutes about a tube that is pushed down the right mainstem bronchus could certainly alter a clinical workflow. But I want to push back a bit. What are the blind spots here?

Maya: The document is very transparent about its limitations, and these are crucial for any clinical team to understand. First, the device assesses solely the vertical position of the tube tip relative to the carina. It does not factor patient positioning.

Sam: Right. So if the patient's neck is severely flexed or extended when the portable X-ray is taken, which we know can physically shift the tube up or down in the airway, the algorithm is not accounting for that anatomical shift. It is just measuring the straight distance on the 2D image it sees.

Maya: Exactly. The second major limitation is explicitly stated in the indications for use: it cannot detect esophageal intubation.

Sam: Wow, okay. Esophageal intubation is when the tube accidentally goes down the swallowing tube to the stomach instead of the windpipe to the lungs. That is a massive clinical emergency. And this AI is completely blind to it.

Maya: Yes. If the tube is in the esophagus, this software is not designed to alert you. This reinforces why it cannot be used on a stand-alone basis for clinical decision-making. The radiologist still has to evaluate the entire clinical picture.

Sam: Are there any issues with the types of tubes themselves?

Maya: There is a specific warning about that. The device was tested in the single lumen endotracheal tube. The document states it may trigger a false prioritization alert in the case of a properly positioned double lumen ETT.

Sam: A double lumen tube has a different shape and depth requirement, often used in thoracic surgeries for separating the ventilation of the right and left lungs. So if a hospital uses a lot of double lumen tubes, this system is going to throw false flags. That is something a purchasing committee definitely needs to know.

Maya: It is. Now, beyond the main performance numbers, they also ran subgroup analyses to see if the AI stumbled over complex medical realities.

Sam: Chest X-rays are notoriously messy. Patients have all sorts of hardware and underlying diseases. What specifically did they test for confounding factors?

Maya: They tested cases with other non-ETT tubes, lines, and life support devices. They also tested cases with various lung diseases. The document lists cavitary lung lesions, lung abscesses, lung masses, nodules, pleural effusion, pneumonia, pneumothorax, and pulmonary edema.

Sam: That is a comprehensive list of things that make a chest X-ray cloudy or difficult to read. Did they test for bad image quality too?

Maya: Yes, they evaluated performance against poor patient's position, artifact present, field of view issues, and missing anatomy, as well as post-op cases with surgery or implants.

Sam: And how did the algorithm handle all that noise?

Maya: The summary concludes that the device consistently performed reliably across these circumstances.

Sam: Alright, looking at the big picture, this 510(k) summary tells a clear story. Ever Fortune.AI has built a triage tool that looks at adult chest X-rays, does some quick math to see if a single-lumen breathing tube is out of bounds, and if so, drops a JSON file into the PACS system to push that case up the radiologist's list. It does all of this in an average of 2.49 minutes.

Maya: That is the core function. But any hospital adopting this must train their staff on what it does not do. It does not circle the tube on the image, it does not factor in how the patient is positioned, it will likely flag a perfectly good double lumen tube, and most importantly, it cannot warn you if the tube is in the esophagus.

Sam: It is a safety net for a specific type of malposition, not a replacement for a doctor's eyes. The numbers are impressive—a sensitivity of 0.890 and specificity of 0.935—but the boundaries of what it is allowed to do are just as important as how well it does it.

Maya: Absolutely. And based on these indications for use, technological characteristics, and the retrospective performance testing on those 940 studies, the FDA agreed that the EFAI ETTXR raises no new questions of safety and effectiveness compared to its predicate.

Sam: Which earned them their Class II clearance on February 20, 2025. That is all for this Company Spotlight on AI in Medicine - Smart Summaries. A quick reminder that the source document is linked in the show notes so you can read the data yourself.

Maya: And remember, this is not medical advice. Thank you for listening, and we will catch you on the next deep dive.