Why Device Images Are Treated as Claims, Not Decoration
FDA's promotional-materials framework does not distinguish between a printed brochure and a product listing image. If a picture communicates something about what the device is, how it is used, or what it achieves, it is a claim — and claims must be truthful, non-misleading, and consistent with the cleared indications for use.
This has three practical consequences for anyone building a product catalog:
- Implied use claims count. An image showing a device on a patient's chest implies an indication. If your clearance covers a different anatomical site or population, the image is off-label promotion regardless of what the caption says.
- Before-and-after imagery carries an evidentiary burden. Any visual comparison implying an outcome needs clinical support behind it. This is where aesthetic and wound-care device brands get into trouble fastest.
- Images can become controlled records. Photography used in submissions or labeling falls under design-control and record-keeping expectations in 21 CFR Part 820 — meaning version control and documented change history, not a folder of files named final_v3_REAL.jpg.
Marketing commissions a lifestyle shoot, the agency stages a clinician using the device in a plausible-looking scenario, and nobody routes the shot list past regulatory. The staging itself becomes the claim. Approve the shot list, not just the final images.
Note that scope varies. A Class I exempt product like a basic bandage carries far lighter promotional risk than a Class II or III device. The discipline below scales — apply it in proportion.
| Product type | Image risk level | Review needed |
|---|---|---|
| Class I general wellness accessory | Low | Marketing self-check |
| Class II diagnostic or monitoring device | Moderate | Regulatory sign-off on shot list + finals |
| Class III / implantable | High | Full promotional review, documented |
| Any before/after or outcome imagery | High | Clinical evidence on file |
The Edit Line: What Retouching Is Safe and What Is Not
The most useful mental model is this: you may correct how the camera lied. You may not correct how the device looks.
Cameras introduce artifacts that are not properties of the product — sensor dust, a color cast from mixed lighting, a reflection of the softbox, uneven exposure across a long housing. Removing those brings the image closer to reality. Removing a mold line, evening out a textured grip, or making a matte polymer look glossy moves the image away from reality, and those are exactly the details a reviewer can compare against a physical sample.
Generally safe
- White balance and color calibration to a measured target
- Exposure and contrast normalization
- Dust, lint, and sensor-spot removal
- Background replacement to plain white or a neutral field
- Removing studio reflections and stand/rig hardware
- Cropping that preserves all required markings
- Consistent scaling and centering across a catalog
Risky or prohibited
- Smoothing seams, parting lines, or surface texture
- Altering proportions, port placement, or tubing gauge
- Brightening indicators or screens that were not active
- Removing or obscuring UDI, lot fields, or regulatory marks
- Generating a device or accessory that was not photographed
- Compositing a device into a clinical scene it is not cleared for
- Enhancing before/after contrast in outcome imagery
Color deserves its own paragraph. In healthcare catalogs, color is frequently a functional attribute: sterile versus non-sterile packaging, gauge or size coding on connectors, dosage variants within a family. A retoucher who nudges saturation to make a listing pop can make two SKUs visually indistinguishable — a genuine patient-safety issue, not just a compliance one. Shoot with a color reference target in a throwaway frame and calibrate to it rather than to taste.
Keep an unedited master alongside every published asset, and keep the two linked by SKU. When a reviewer asks whether an image was altered, the answer should be a two-file comparison, not a memory test.
Using AI Editing Tools Without Creating an Audit Problem
AI editing is now standard in e-commerce production, and regulated categories are not exempt from the efficiency pressure — a mid-size device distributor may carry thousands of SKUs across dozens of accessory variants. The question is not whether to use these tools but which operations you let them perform.
The split follows the same logic as manual retouching, with one addition: generative operations require more scrutiny than corrective ones, because a generative model invents pixels rather than adjusting them. Background removal is corrective in effect — the product pixels are preserved and the field behind them is replaced. Generative fill inside the product silhouette is not, because the model is producing device geometry that no camera recorded.
Upscaling is the underrated hazard. AI upscalers hallucinate plausible detail — they will happily invent crisp text on a label that was three pixels tall in the source, and that text may not match the actual labeling. If a device image needs to support zoom, shoot it at the resolution you need rather than manufacturing that resolution afterward.
Whatever tooling you use, write down the protocol. A workable one fits on a page: which operations are approved for unattended batch processing, which require named human review, who signs off, and where masters and outputs are stored. Tools like Retouchable handle the corrective and background work across a large catalog while leaving product pixels intact — but the governance around which operations you enable is yours to define, and that document is what an auditor will ask for.
The three artifacts: the unedited RAW master, the published output, and a record of the operations applied between them. If your pipeline cannot produce all three for any given listing image, it is not audit-ready.
Shooting Devices So They Need Less Retouching
Every compliance argument above becomes easier if the capture is good enough that heavy editing is never tempting. Medical devices are unusually hard to shoot well: they combine polished polymer, brushed metal, clear tubing, printed labeling, and display glass — four different reflectance problems in one frame.
Light for the hardest surface first. Clear tubing and display glass dictate the setup; matte housings tolerate almost anything. Large, close, heavily diffused sources — a big softbox or a light tent — give you broad soft reflections instead of specular hotspots. Add a black flag or negative fill on one side so cylindrical parts show an edge and read as three-dimensional rather than as a flat white blob.
Kill screen glare in camera. A cross-polarized setup (polarizing gel on the light, circular polarizer on the lens) removes reflections from display glass and glossy labeling far more cleanly than any post-processing, and it does not risk altering what the screen shows. If the device has an active display, photograph it in a documented, on-label state.
Keep every required marking legible. Frame so that UDI carriers, lot and serial fields, sterility indicators, and regulatory marks are visible and in focus in at least one catalog image. Focus-stack if depth of field will not cover a deep device at the aperture you need. This is also good merchandising — buyers in healthcare procurement look for exactly these details.
| Surface | Main problem | In-camera fix |
|---|---|---|
| Display glass | Specular glare | Cross-polarization |
| Clear tubing | Disappears on white | Gradient background + edge fill |
| Brushed metal | Blown highlights | Large diffused source, angled |
| Printed labeling | Illegible at catalog size | Focus stack, dedicated detail shot |
| Sterile packaging film | Wrinkle reflections | Tent lighting, polarizer |
Shoot a color-reference target and a scale reference in a throwaway frame at the start of every device setup. Both cost about ten seconds and settle later arguments about whether the published color and size representation are accurate.
Building a Catalog Workflow That Scales
A single hero shot can be handled carefully by one person. A catalog of two thousand SKUs, refreshed as packaging revisions roll through, needs a system. The structure that works looks like this:
- Classify SKUs by risk tier using something like the table in the first section. Most catalogs are heavily weighted toward low-risk accessories and consumables, and those can move through an automated pipeline quickly.
- Approve the shot list before the shoot, not the images after it. Staging, models, anatomical placement, and any implied-use scenario should be signed off while changing them is still cheap.
- Define an approved operation set per tier. Low-risk SKUs: automated background, crop, color-calibrate. High-risk SKUs: same operations, plus named human review against the physical sample.
- Version images alongside the product record. When a device revision changes the housing or the labeling, the image is stale — and a stale image of a regulated product is a misrepresentation, not just a cosmetic issue.
- Re-audit on a fixed cadence. Quarterly is reasonable. Check that every published image matches the current physical product and the current cleared labeling.
Step four is the one most teams skip, and it is the one that quietly accumulates risk. A device gets a running change, engineering documents it properly, and the product page keeps showing last year's housing for eighteen months because nobody owns the link between the change order and the image library. Make that ownership explicit.
The payoff for the discipline is not only regulatory. Healthcare buyers — procurement managers, clinic administrators, distributors — inspect imagery more closely than consumer shoppers do, because they are matching a photo against a spec sheet and a purchase order. Accurate, high-resolution, well-lit device imagery with legible markings is a genuine competitive advantage in a category where most listings are lit badly and shot from one angle.