DICOM viewer

DICOM is the standard format and network protocol for medical images. A DICOM file holds the image pixels together with tags: who the patient is, what study and series it belongs to, which machine made it, and how to display it, such as the pixel spacing and the suggested window. A CT or MRI series is usually one file per slice.

A CT stores far more shades of gray than a screen can show, so the viewer shows only a range of them. The level (window center) picks the middle of that range and the width sets how wide it is. A narrow lung window makes the airways visible; a wide bone window shows the skeleton. Drag left and right for the width, up and down for the level.

CT values are measured in Hounsfield units (HU): water is 0, air is about -1000, fat is around -100, and bone is in the hundreds or more. Point at the CT and the lower right corner shows the value under the mouse in HU.

With the Pixel Spacing tag of the file, which says how many millimeters one pixel covers. The X-rays were scaled down for the web, and DCMTK's dcmscale rewrote their pixel spacing to match, so the measurements stay in millimeters.

It is stored as MONOCHROME1, where a higher value means darker. The viewer reads the Photometric Interpretation tag and flips it, so bone is white as usual. Press Invert to see it the other way around.

The human studies are from public research collections of The Cancer Imaging Archive (TCIA): COVID-19-AR for the chest CT and X-rays, 4D-Lung for the breathing CT, UPENN-GBM for the brain MRI and B-mode-and-CEUS-Liver for the ultrasound. The archive removes names and other identifiers before publishing: the patient ID you see is its own code. The heart is a pig's, from a published animal study. All are shared under Creative Commons Attribution licenses.

DCMTK is the open-source DICOM toolkit from OFFIS in Oldenburg, Germany: C++ libraries and command-line tools for reading, writing, converting, compressing and sending DICOM files. Here it reads every tag (dcmdump), extracts the pixels, scales the large X-rays (dcmscale) and renders the preview images (dcm2img). The browser then does the window, level, zoom and measuring itself.

Yes. Click Open files or Open a folder, or drop files on the image. They are read by your browser on your own device: nothing is uploaded or stored. Close and delete removes them from the page, and they are gone when you leave or reload it too. A folder or several files open as one series (the one with the most images); files that aren't DICOM are skipped.

Uncompressed files (implicit or explicit VR, little or big endian), deflated files, RLE lossless and baseline JPEG, in gray (8 or 16 bits) or color, single images or multi-frame clips. Files compressed as JPEG lossless or JPEG-LS still show all their tags, with the DCMTK command that decompresses them (dcmdjpeg or dcmdjpls); JPEG 2000 isn't supported. A DICOMDIR is only the index of a CD: open the image files in its folders instead.

They are cine studies: several images of the same place at different moments, which the viewer plays in a loop at the rate stored in the file. The breathing CT is a 4D CT, sorted into ten phases of one breath and shown here as a front view rebuilt from each phase's slices. The heart MRI keeps two moments of the heartbeat, squeezed and relaxed. The ultrasound clip is 38 frames in one file.

No. This is a technology demonstration, not a medical device, and it is not for diagnosis. Ask the clinic or hospital that made your scan for the report from a radiologist.