A photo x-ray app, in the sense people search, is a phone that photographs an existing film or a monitor and runs a model on that JPEG. It is not a phone that emits X-rays. It is not a CAD that reads a DICOM from PACS. It is not a PYCAD product.
If you meant train / infer / review on a real study → AI for medical diagnosis (the URL /blog/how-to-xray-images/ already 301s there). If you meant how a clinician reads a head MRI → how to read head MRI scans. If you meant the file format → what is DICOM.
PYCAD builds custom web DICOM viewers and imaging models. It does not ship a photo-x-ray app, a store listing, or a “see through skin” filter.
Three things that share the search
| What someone meant | What it actually is | What it is not |
|---|---|---|
| Entertainment filter | A camera effect that paints a skeleton on a selfie | A medical device. There is no tube, no dose, no anatomy |
| Photo of a film / monitor | JPEG of a printed radiograph or a workstation screen, then a classifier or detector | A DICOM retrieve. No tags, no window that the PACS used, no series |
| Portable X-ray hardware | A real tube on a cart or a handheld generator | An App Store listing. Different market, different clearance |
The first is a toy. The third is equipment. This URL is the middle one — the thing a search for “photo xray app” usually wants once the toy is ruled out.
A phone camera does not produce X-rays. Any listing that says “live X-ray” with the rear camera is the toy.
What the middle app can honestly do
Someone already made a radiograph. You photograph it because you do not have the DICOM: a film on a light box, a PNG on a WhatsApp thread, a screen in a room you do not admin. A model that saw a lot of chest or bone radiographs may highlight a region or emit a class.
That is a degraded input:
- The file is a photo. Glare, perspective, a thumb, a window the phone chose, JPEG blocking. The network was probably trained on DICOM pixels or challenge PNGs, not on your light-box snapshot.
- There are no tags. No laterality that you can trust, no technique, no later series. “Left” is a guess from the letter marker if it is even in frame.
- It is not the signed report. A class score is not a diagnosis. A box is not a finding until a person says so.
Triage theatre (ER snaps a film, AI “flags a fracture in seconds”) is a demo script. If the site has a PACS, send the DICOM. If it does not, the honest use is “we are stuck with a picture of a picture.” Rural-clinic marketing that skips that sentence is the Outrank version of this page.
When it becomes a device
If the app claims to detect a fracture, a pneumothorax, or a cavity for clinical use, it is making a device claim. In the US that is FDA; in the EU that is a CE-marked medical device. A privacy policy and a HIPAA paragraph do not replace that. Entertainment listings do not need a clearance because they do not claim a finding.
Evaluation, if you are actually buying one: a named intended use, a study on photos of films (not only on DICOM), sensitivity / specificity with a denominator, and what happens to the image after you upload it. Store-star counts and “millions of downloads” are the toy side of the aisle.
What this page will not do
- Not a market CAGR. Mobile X-ray hardware numbers are a different industry.
- Not a list of store apps. They turn over. A named cleared product belongs on AI medical devices if it is a real device, not here as a costume.
- Not a PYCAD photo-x-ray SKU. A viewer that opens DICOM is the opposite job: keep the file, do not rephotograph the monitor.
If you have the study, use the study. The diagnosis URL is AI for medical diagnosis. This URL is only for the case where the input really is a phone photo of a film.
Keep Reading
Related Articles
Explore the full DICOM HubAll services, tools, and guides on this topic — in one place.
Visit Hub →