RARA CRAFT DATABASE for FDA Medical Devices

Device Spotlight — DeepView AI System (DEN250028): Reading the De Novo Record for a Burn-Assessment AI

日本語版

When you open a De Novo record, read the classification name and the regulation number first. In a 510(k), the applicant places the device into a classification that already exists. In a De Novo, FDA writes a new classification for this device, and any similar device that comes later through a 510(k) will be measured against it.

On May 21, 2026, FDA granted De Novo request DEN250028 for the DeepView AI System from Spectral MD, Inc. (Dallas, TX). The grant created a Class II classification under 21 CFR 878.1840 with product code SHY. "21 CFR" refers to Title 21 of the Code of Federal Regulations, where FDA's regulations are codified. The device analyzes images taken with its own imaging system and uses AI to highlight deep burn tissue, meaning deep partial-thickness and full-thickness burns.

A device classified through De Novo may serve as a predicate for later 510(k)s where applicable. A follow-on device would still have to meet the special controls for SHY and demonstrate substantial equivalence to its predicate. Sharing the classification is where that work starts, not where it ends.

What the record says

From FDA's De Novo database, checked on October 1, 2026:

ItemRecord
Device NameDeepView AI® System
RequesterSpectral MD, Inc. (Dallas, TX)
Pathway and numberDe Novo, DEN250028
Date Received2025-06-27
Decision Date2026-05-21 (Granted)
Product CodeSHY
Regulation Number21 CFR 878.1840
Classification Namesoftware-aided adjunctive diagnostic device for use by healthcare providers in skin wound assessment
ClassII
Advisory CommitteeGeneral & Plastic Surgery
Public documentsClassification Order (PDF). No Decision Summary has been published yet

The advisory committee is General & Plastic Surgery even though the device is AI software. Panels follow the clinical area where a device is used, not the technology inside it.

The classification says "skin wound." The indication says "burn."

The classification name and the Indications for Use cover different ground.

The regulation describes a "software-aided adjunctive diagnostic device for use by healthcare providers in skin wound assessment." That wording covers skin wound assessment in general, not burns alone.

The Indications for Use for the DeepView AI System, as published in the Classification Order, are specific to burns:

The DeepView AI®-Burn software analyzes images obtained using the DeepView AI® System, using artificial intelligence to detect deep burn tissue (deep partial thickness and full thickness) and provides segmentation of specific areas in a burn wound by highlighting areas that are unlikely to heal within 21 days with conservative treatment alone. The absence of a highlighted region on an image indicates the absence of deep burn tissue. The system will calculate wound area measurement and percent total body surface area (TBSA) burn.

The input is images from the device's own imaging system. The output is a highlighted region marking tissue unlikely to heal within 21 days with conservative treatment, and the indication also gives meaning to the absence of a highlight. The system adds wound area and percent TBSA.

A company planning a follow-on device has two separate questions to answer: does the device fit the classification, and are its indications close enough to the DeepView AI System to use it as a predicate? The breadth of the classification name does not by itself make a device for non-burn wounds a straightforward 510(k).

Seven special controls

The Classification Order sets out seven special controls for 21 CFR 878.1840:

  1. Clinical performance validation testing
  2. Non-clinical performance testing, covering software compatibility and photobiological safety
  3. Precision testing (repeatability and reproducibility)
  4. Electromagnetic compatibility and electrical and thermal safety
  5. Software verification, validation and hazard analysis
  6. Human factors assessment
  7. Labeling, including summaries of performance testing, limitations, guidance on interpreting the output, a statement that the device is not for stand-alone diagnosis, and safety warnings

Photobiological safety is on the list, most likely because the software works with an imaging system that illuminates the patient. The testing burden for this category is wider than for software that runs on images from any camera.

These controls are a baseline. They name what must be addressed without prescribing test methods or acceptance criteria, and a 510(k) under SHY will need more than this list, built from the applicable FDA guidance and the predicate comparison. A test plan can start from the special controls, but it does not end with them.

Why De Novo, and why Class II

De Novo is the pathway for a low-to-moderate-risk device with no legally marketed predicate. The grant tells you the device did not fit an existing classification.

The device is Class II because FDA concluded that general controls plus special controls provide reasonable assurance of safety and effectiveness. The specific risks and the controls matched to each are normally laid out in the Decision Summary, which has not been published for this device yet.

What we can see is how the word "adjunctive" in the classification name connects to the seventh special control, which requires labeling to say the device is not for stand-alone diagnosis. Calling a device adjunctive does not by itself put it in Class II. Here, though, limiting how the output is used is written into the controls as one of the mitigations. When you draft indications and labeling for an AI device, how the output is meant to be used and how its risks are controlled have to be worked out together.

If you are building something similar

  1. Check the fit with the classification. The regulation defines the category. A device that assesses a different kind of wound, or is used by different users, may not fit it.
  2. Map your test plan to the seven special controls. Each one needs an answer in your submission, and photobiological safety applies if your device illuminates tissue.
  3. Compare your Indications for Use with the DeepView AI System. If they are close, a 510(k) under SHY is a realistic route. If they differ in ways that raise different questions of safety and effectiveness, you may need a De Novo of your own.

These questions are easier to answer before testing begins than after.

Follow SHY on this site

Open the DEN250028 device page and star (☆) it. On Search, enter SHY in the Product Code field to list the records under this classification. When a follow-on 510(k) appears, star it as well, tick Compare on both in My Workspace, and read the two Indications for Use side by side. You will see whether the category has moved beyond burns, and how far.

RA Craft Database is an independent service and is not affiliated with or endorsed by the U.S. FDA. Articles are for general information on product development and regulatory research; they are not legal or regulatory advice for a specific device. Official FDA documents remain the authoritative source.

← All articles