// blog
Count DICOM Files Opened Without Tracking Patients
When a study opens, the viewer sends one small count: that it opened, how many files it had, their total size, the modality, the transfer syntax UID, how it was opened, and, inside another website, that site’s name. The server checks every field against a short list and adds it to a daily total, with no IP address, file name or anything from inside the file. That’s enough to see use and decide which compressions to support next.

Our free DICOM viewer reads files in the visitor’s browser and never uploads them. Now it’s also on other websites. We wanted to know one thing: how many files does it actually read? Page views don’t say. A visit can be someone glancing at a sample and leaving.
Medical files are the worst place to be careless with analytics. A DICOM file carries the patient’s name, ID, birth date and the clinic in its tags, and the file name often repeats some of it. So we started from what must never be sent and worked back.
Never sent
- The file, or any part of it.
- The file name. People name files after patients.
- Any tag value: no names, IDs, dates, institutions or descriptions.
- The visitor’s IP address, an ID or a cookie.
Sent, once per study
| Field | Example | Why |
|---|---|---|
| What happened | opened, no image, failed, 3D | Did it work? |
| Where | our page, another website, our imaging example | Where it’s used |
| How | Open button, dropped, handed over by the site | Which way people use it |
| Modality | CT, MR, US | What kind of scan, as a two-to-four-letter code |
| Transfer syntax UID | 1.2.840.10008.1.2.4.70 | Which compression, so we know which decoder to add |
| Files, megabytes | 312, 160 | How much it actually read |
| Website | clinic.example.com | Only inside another website |
Seven fields, each one a number or a word from a short list.
The server checks every field
The browser could send anything; the server keeps only what fits:
- The event, place and how must each be one of a few fixed words.
- The modality must be two to four capital letters. A patient’s name, such as
DOE^JOHN, has a caret and is refused. - The transfer syntax must be digits and dots. Its readable name is refused.
- File counts and sizes must be in range.
- A website’s name is reduced to its plain form (
www.Clinic.Example.com/viewerbecomesclinic.example.com). Anything that isn’t a site’s name is left out, and the count still counts.
Anything else, and the whole count is refused. These rules have their own tests.
Daily totals, not events
Each count adds one to a row for today with the same fields, and adds its files and megabytes. There’s no row per visit, so there’s nothing to trace back to a person, even for us. Like any request, the call appears in the server’s own access log, which keeps no file data either and is described in our privacy policy.
What it tells us
The admin dashboard shows the last 30 days:
- Studies opened and DICOM files read, with their megabytes.
- How many were read but showed no image, and their compression.
- Our page, other websites and the imaging example, side by side.
- Opened with the button, dropped, or handed over by a site.
- Which websites embed it, how often it loads there and how many files it reads.
The no-image count by transfer syntax is the useful one. It’s a ranked list of the decoders people need.
The same idea for your product
Most products need to know how they’re used, not who used them. Decide what must never leave first, send a few fixed fields, check them on the server, and keep totals. It’s less to protect, less to explain and nothing to leak. If you want this in your own software, tell us what you need.
Frequently asked questions
Is a file name ever sent?
No. File names often hold a patient's name or ID, so they never leave the browser. Only the number of files and their total size are counted.
Why the transfer syntax?
It says how the pixels are compressed. Files the viewer can't decode yet are counted with it, so we can see which decoder to add next. That is how JPEG lossless and JPEG 2000 got theirs.
Why keep the embedding site's name?
To know which websites use the viewer, and how much. A website's name isn't personal data. Nothing about the visitor goes with it.
Is our own testing counted?
Mostly not. A browser that has signed in to our admin stops sending counts from our own pages, so our testing doesn't look like use. Inside other websites, browsers keep that setting apart, so a test there can still count.

































































































































