hjLabs AI Playground
Starting the on-device runtime…
Every DICOM file carries the patient’s identity in its header — name, patient ID, date of birth, accession number, the institution, the exact date and time — and almost nobody who shares one knows that. This page opens a .dcm file with the File API, parses it with dicom-parser (MIT, 0.55 MB), draws the pixel data to a canvas with window and level controls, lists every identifier it found in the header, and lets you blank the ones you choose and download the result. It runs no model and makes no claim whatsoever about what the image shows. The file never leaves the tab: there is no upload endpoint here, which your browser’s own network panel will confirm.
dicom phi vieweranonymize dicom onlinedicom viewer browserremove patient name from dicomdicom metadata viewerdicom anonymiser freeview dcm file onlinedicom header viewer
No. It is read with the File API and parsed in this tab. hjLabs.in has no upload endpoint and no storage for it — closing the tab is the whole deletion process. Open your browser’s network panel while you use the page and you will see model and asset requests only, never your file.
No, and it never will. It reads the header and windows the pixels; it runs no model and makes no claim about the image. It is a file inspector, not a diagnostic tool, and nothing here is medical advice — questions about a scan belong with the clinician who ordered it.
Uncompressed Part 10 files — Implicit and Explicit VR Little Endian, and Explicit VR Big Endian — at 8 or 16 bits per pixel, greyscale or RGB, single- or multi-frame. JPEG, JPEG-2000, JPEG-LS and RLE files parse their headers fully and every de-identification control works on them, but the picture needs a codec this build does not ship, and the page says so rather than drawing noise.
Selected element values are overwritten in place, at their existing byte offsets and lengths, with a valid replacement for their value representation — a blank name, 19000101 for a date, a fresh 2.25-arc UID. Everything else in the file, pixel data included, is byte-identical to the original. Tags are emptied rather than removed, and identifiers burned into the pixels themselves are not touched, so this is a redaction pass to review, not a certified anonymiser.
Use one if you have one. This exists for the case where installing software is not an option — a borrowed machine, a locked-down hospital desktop, a patient who has been handed a disc of their own scans — and where uploading the file to a web service is exactly the thing you are trying to avoid.
Yes, once a model is cached. Weights are stored in the Cache Storage API under the key "transformers-cache", so a reload with no network still runs the model. Call "Make storage persistent" first — browsers may evict non-persistent caches under disk pressure.