Production formats for medical imaging: DICOM originals, exports, and what to specify

What a DICOM original is, and what an export throws away

A DICOM study is not a folder of pictures. It is a structured record in which every image carries, inside the same file, the circumstances of its acquisition: patient identity as registered, study and series identifiers, acquisition date and time, modality, institution, equipment, protocol, slice position and orientation, pixel spacing — the calibration that makes an on-screen measurement a measurement. The pixel data itself is typically stored at diagnostic bit depth, well beyond what a monitor shows at once; the reader chooses a window into that range (bone, lung, soft tissue) at review time.

Exporting to JPEG or PDF discards nearly all of it. The metadata is gone or reduced to a burned-in caption. The bit depth collapses to what the exporter’s window happened to show — window/level becomes a decision someone else already made and cannot be revisited. Series structure flattens into loose files; spatial calibration is lost, so nothing measured on the export is reliable; and the selection itself is an edit, because someone chose which slices to export. An expert working from exports is reviewing a curated derivative. That can be fine for a demonstrative; it is not the record.

Lossy compression is an alteration someone chose

JPEG’s ordinary mode is lossy: it discards pixel information permanently to shrink the file, and re-saving compounds the loss. For holiday photographs this is invisible and free; for a study whose evidentiary weight may turn on a subtle finding, it is an unforced error. The moment imaging is produced in a lossily recompressed form while the diagnostic original exists elsewhere, the producing party has handed its opponent a genuine line of attack — the exhibit is measurably not what the radiologist read, and the difference is the producing party’s doing. The authentication framework in FRE 902(13) and 902(14) makes the contrast sharp: a hash-verified copy of the original is arithmetically identical to what the machine produced; a recompressed export is arithmetically guaranteed not to be. (DICOM does permit lossless compression internally — the disqualifying word is lossy, not compressed.)

“Produce what was received”

The cleanest production position is also the simplest to defend: the producing party hands over the bytes it received, unmodified — the disc image, the records-vendor ZIP, the portal download — with a hash manifest, and lets the requesting party do its own viewing with tools of its choice. Every transcoding step a producing party inserts “for convenience” is a step it may later have to explain. This is a discipline that has to be practiced before it is needed: if the working copy is the only copy, the original was destroyed by the workflow. Slicefield takes the position structurally — studies are immutable once ingested, and alongside the viewer’s working copy the platform keeps the original files byte-for-byte as received, so a production ships what arrived, and the append-only ledger and printable chain-of-custody report say so. The full custody story is on the security & chain of custody page.

When imaging meets e-discovery tooling

Standard e-discovery production — Bates-stamped TIFF/PDF images plus load files carrying the metadata — was designed for documents, and imaging breaks its assumptions. A single study can be thousands of files that only mean anything as a set; an imaging file’s “page” is not a page; and Bates-stamping the pixels of a medical image is exactly the alteration everything above says to avoid. The workable pattern is the one already used for other native productions: produce the study natively, as received, on labeled media or a secure transfer; assign each study a Bates number or range on the media, the folder and the manifest rather than burned into pixel data; and place a slip-sheet or placeholder record in the review set that carries the Bates number and points at the native production. Review platforms tolerate imaging as opaque attachments at best; nobody windows a CT in a document reviewer, and pretending otherwise produces a review nobody actually performed.

What to specify in the request for production

Most bad imaging productions are answers to requests that never said otherwise. A request that asks for “copies of imaging” invites a PDF. Ask instead for, in substance: all diagnostic imaging in original DICOM format as maintained or received, including complete studies and all series — not selected images; original metadata intact, with no conversion, recompression or other transcoding of pixel data; any DICOMDIR or directory structure as it exists; prior and comparison studies for the same patient where they exist; a hash manifest of the produced files; and identification of the source system and export method. Two things do not belong in the request: a demand for any particular vendor’s viewer (DICOM is the point — any conformant viewer reads it), and silence on radiology reports, which are records in their own right and travel with the imaging under the usual business-records mechanics. On the receiving side, the same list is the checklist for what arrived — and a matter’s expected records list is where the gap between requested and received stays visible instead of surfacing at deposition.

This article is general background for litigators, not legal advice, and reading it creates no attorney–client relationship. How any rule applies turns on the jurisdiction, the court and the facts of the particular case — those judgments belong to counsel.

All articles · Security & chain of custody · Open the first disc free