DICOM
Also known as: Digital Imaging and Communications in Medicine, DICOM standard
DICOM (Digital Imaging and Communications in Medicine) is the international standard for medical images and related information.
It defines both the file format, which bundles the image with details like the patient, scanner, and settings, and the network rules that let scanners, viewers, and archives exchange images. While patients rarely hear of it, nearly every X-ray, CT, MRI, and ultrasound taken in a hospital today travels as DICOM.
DICOM, explained at three levels
At home
Think of a DICOM file as a photo in an envelope with a detailed label. The photo is your X-ray or MRI image. The label records who you are, which machine took the scan, and its settings. DICOM is the shared standard that defines both the envelope and the delivery rules, so a scan made on one company's machine can be opened and sent by another company's system.
At work
DICOM is the international standard for medical images and the information attached to them. It defines the file format and the network services imaging devices use to send, find, and retrieve studies. For hospitals, it means scanners, archives, and viewers from different vendors work together without custom integration, which protects investments and makes it possible to switch vendors.
At home
A DICOM file has two parts: a header and the image data. The header holds tagged fields, each with a number and a value, such as patient name, birth date, scan date, body part, and machine settings. The image data may hold one picture or hundreds of slices from a CT scan. When you request your imaging records, you often get a CD or download containing these files plus a small viewer, because ordinary photo apps cannot read the format.
At work
DICOM organizes everything around information objects, such as CT Image or Structured Report, and services that act on them. C-STORE sends images, C-FIND and C-MOVE query and retrieve them, and Modality Worklist sends scheduled patient details to the scanner so names are not typed twice. Data is grouped as patient, study, series, and instance, each with a unique identifier. Vendors publish conformance statements listing what they support, and DICOMweb offers the same functions over RESTful web APIs.
At home
Removing personal information from DICOM is harder than it looks. Names and IDs sit in standard header fields, but also in free-text descriptions, private vendor tags, and sometimes text burned into the image pixels, like an ultrasound's on-screen label. Faces can even be rebuilt from head CT and MRI scans. While a quick script can blank the obvious fields, in my view de-identification deserves careful, tested tooling, because one missed field exposes a patient.
At work
Classic DICOM networking was designed for trusted hospital networks and often runs unencrypted, with devices identified only by application entity titles. Researchers have repeatedly found internet-exposed imaging servers leaking millions of images. The 128-byte file preamble can hide executable code, a technique demonstrated in 2019. Use TLS, network segmentation, and access controls. For AI, DICOM Structured Reports and Segmentation objects return results into radiology workflows in a standard form.
Real-world examples
At home
Nadia needed a second opinion on a knee MRI. The clinic gave her a download link with DICOM files, which the specialist's office opened directly in its own viewing system.
At work
Jerome is an imaging IT manager at a Saskatchewan hospital. When adding an AI tool that flags lung nodules, his team used DICOM routing so CT scans flowed to the AI automatically and its results returned as DICOM reports radiologists could see.
Common confusions
- DICOM is just an image format like JPEG. — It's also a network protocol and stores detailed patient and scan metadata with the image.
- DICOM files are anonymous. — They usually contain identifying details unless carefully de-identified.
- DICOM and PACS are the same thing. — DICOM is the standard. PACS is the system that stores and displays images using it.
For analysts
Imaging data is valuable for AI projects but full of hidden personal information. I'd require a documented, tested de-identification process before any DICOM images leave clinical systems.
History & origin
In the early 1980s, each imaging machine maker used its own format, so images couldn't easily move between systems. The American College of Radiology and the National Electrical Manufacturers Association published the ACR-NEMA standard in 1985. Its third version, released in 1993, was renamed DICOM and added network communication. DICOM has been continuously extended since, covering new modalities such as digital pathology and ophthalmology, and DICOMweb added web-based APIs in the 2010s. Cloud providers now offer DICOM-compatible imaging services. In my opinion, DICOM is a model of a standard that quietly worked, though its age shows in privacy-sensitive metadata scattered through every file.