DICOM viewer and de-identification check

Open files… takes one or more .dcm files; Open folder… (or dropping a folder) takes a whole series. Slices are grouped by Series Instance UID and ordered by their position along the slice normal (Image Position and Orientation), falling back to Slice Location or Instance Number.

Load a sample (synthetic) builds a five-slice CT-like phantom in your browser: air, a soft-tissue body (+40 HU) with a fat layer (−100 HU), a lung-like region (−800 HU), a bone rod exactly 20.0 mm across (cortical +1200 HU around cancellous +400 HU) and a water vial (0 HU) in the middle slices, at 0.75 mm pixels and 5 mm slices. Its identifiers are fictional — and it has a fictional ID burned into the pixels while its header says Burned In Annotation = NO, because that is exactly the trap the book warns about.

Window — the file’s own window, the four CT presets from the book (brain 40/80, soft tissue 40/400, lung −600/1500, bone 400/2000; level/width in HU), the full range, or your own level and width. The mapping is the DICOM linear VOI function (PS3.3 C.11.2.1.2). The window is a display transform, not the data.

Probe shows the stored value and the value after Rescale Slope and Intercept — Hounsfield units for CT. Distance (drag) gives the length in mm from Pixel Spacing (row and column spacing separately); with only Imager Pixel Spacing it says the length is at the detector. Region (drag a rectangle) gives the pixel count, area, mean, SD, minimum and maximum — the exercise in ch16 of measuring air, fat, muscle and bone.

What it can support shows the pixel size, slice thickness and the spacing between slices (from their positions, so gaps and overlaps show), and checks a size you want to claim against them. A 5 mm slice cannot support a 0.5 mm claim.

Every tag lists the file meta header and every element, nested sequences included, with its VR, name and value. Private tags (odd groups) are shown with their raw bytes or text.

De-identification check flags each attribute that PS3.15 Annex E Table E.1-1 lists for the Basic Application Level Confidentiality Profile, with the profile’s action (X remove, Z empty, D dummy, U new UID, C clean), plus every private tag, and adds warnings about burned-in text and head volumes. Export anonymised copy applies the profile and lists every change; a series exports as a .zip with UIDs replaced consistently across the files. Open the anonymised copy here re-reads the result so you can check what remains.

Not a diagnostic viewer: no calibrated display, no reporting, no clinical use.

The anonymised copy follows the basic profile only. It does not remove burned-in text, deface head CT or MR (a face can be reconstructed from them), clean free text in attributes it keeps, or shift dates (it empties them). Where Annex E offers a choice that depends on the IOD (X/Z, X/D, Z/D, X/Z/D) it keeps the attribute with an empty or dummy value. Treat the result as a draft for someone qualified to verify; de-identification is a verified process, not a best effort (ch15).

“Direct identifier” versus “quasi-identifier” is this tool’s reading of each attribute name; Annex E gives only the action.

JPEG Baseline pixels are lossy, so values read from them are approximate. Palette colour and YBR uncompressed images are not displayed.

Sampling is only the floor of accuracy. Segmentation threshold, partial volume, smoothing and decimation each move a surface further (ch15).