DICOM vs JPG: what you lose when you convert

What a JPG actually holds

Open a chest X-ray in a DICOM viewer, export it as a JPG, and put the two side by side. They look the same. That similarity is misleading: the JPG is a photograph of a measurement, and the DICOM is the measurement.

A JPG stores 256 levels of grey per channel — one byte per pixel. That is plenty for a holiday photo, because your eye cannot tell apart more than about a hundred shades in a single image.

A CT slice stores 4,096 levels, sometimes 65,536. Those numbers are not brightness: they are Hounsfield units, a physical scale where water is 0, air is around −1,000 and dense bone is above +1,000.

That window is not an abstraction: it is two specific fields in the DICOM standard itself. They are called Window Center (0028,1050) and Window Width (0028,1051), and the standard defines them as the linear conversion between the stored values and what ends up on screen. Before that, Rescale Slope and Intercept turn the sensor's raw numbers into Hounsfield units.

When you export to JPG, software applies that window and flattens the result to the 256 greys the format allows. Everything outside the window becomes pure black or pure white, permanently.

You can check it with one of your own files right now:

Why that matters in practice

A radiologist looking at a chest CT does not view it once. They look at it with a lung window, then a mediastinal window, then a bone window. Same data, three completely different images, because each window brings out a different range of densities.

A JPG only has the window that was applied when it was exported. If the export used a lung window, the soft tissue is gone. Not hidden — gone. There is nothing left to recover it from.

It makes more sense once you move the window yourself on a real file:

That is the single most important thing to understand: converting is not compressing, it is discarding.

What else disappears

• Scale. DICOM records pixel spacing, so software knows a lesion is 8 mm across. A JPG has pixels and no idea how big they are.
• Position. Slice location and thickness are what let a viewer reconstruct other planes from the same series.
• Context. Which study and series the image belongs to, the equipment, the acquisition settings.
• The patient's data. This one you usually want gone — see below.

When a JPG is exactly right

Most of the time, honestly. If you are emailing an image to a relative, attaching it to an insurance claim, putting it in a report or printing it for an appointment, a JPG does the job and a DICOM does not: no ordinary program opens one.

The rule is simple. Share a JPG, keep the DICOM. Never send the JPG as a replacement for the original when a professional is going to look at it.

The privacy side

A DICOM file normally carries the patient's name, identifier, date of birth, the hospital and the exact date of the scan, all in its metadata. A JPG produced from the rendered pixels carries none of that.

So converting is a rough way of stripping identifying data. But be careful with one thing: ultrasound, angiography and secondary capture images often have the patient name printed into the image itself by the scanner. That text is part of the picture. Converting does not touch it, and neither does removing metadata.

If you need to share a study properly de-identified, use the DICOM anonymiser and then look at the image before sending it.

Doing it yourself

Everything above runs in your browser, with no upload:

• Online DICOM viewer — open the file and adjust the window until the structures you care about are visible.
• DICOM to JPG converter — export what you see, with the window already applied.
• DICOM metadata viewer — check exactly what your file says about you before sharing it.

And if what you actually want is to understand what the report means rather than what format it is in, that is what ExplainMyScan does.