What Is DICOM? A Guide to Medical Imaging Standards
DICOM for orthopaedic surgeons — the universal standard behind every CT, MRI, and X-ray, and what clinicians need to know.
Why DICOM Matters
Every CT scan, MRI, and digital X-ray you have ever viewed on a hospital workstation was stored and transmitted using DICOM — Digital Imaging and Communications in Medicine. It is the universal standard that allows imaging equipment from different manufacturers (Siemens, GE, Philips) to produce files that any compliant viewer can display.
For orthopaedic surgeons, DICOM is not just a technical detail — it is the foundation that makes cross-institutional image sharing, AI analysis, and 3D surgical planning possible.
The DICOM Hierarchy
Every DICOM dataset follows a four-level hierarchy:
Patient is the top level, identified by Patient ID and Patient Name (DICOM tags 0010,0020 and 0010,0010).
Study represents a single imaging session — for example, "CT Knee Left, 09 March 2022." Each study has a unique Study Instance UID and a Study Date.
Series groups related images within a study. A knee CT might contain a bone-window series, a soft-tissue series, and a 3D reconstruction series — all from the same scan session but with different reconstruction parameters.
Instance (or Image) is a single slice or frame. A CT series of the knee might contain 100–600 instances, each representing one axial slice.
This hierarchy is why a surgeon can open a study and navigate between different series and views without manually sorting through hundreds of individual files.
Key DICOM Tags Every Surgeon Should Recognise
You do not need to memorise DICOM tags, but knowing a few helps when troubleshooting viewer issues or communicating with your IT department:
(0010,0010) Patient Name — the patient identifier displayed in the viewer header.
(0008,0060) Modality — CR (computed radiography), CT, MR, DX (digital X-ray). Determines how the viewer renders the image.
(0028,1050) Window Center and (0028,1051) Window Width — the default brightness and contrast settings. Different windowing presets (bone window WC:300/WW:1500, soft tissue WC:40/WW:400, lung WC:-600/WW:1500) each reveal different anatomical structures from the same underlying data.
(0020,0032) Image Position Patient — the 3D coordinate of each slice. This is how viewers reconstruct a stack of 2D slices into coronal, sagittal, and 3D views.
(0018,0015) Body Part Examined — identifies the anatomical region (KNEE, HIP, SHOULDER). AI tools use this tag to select the appropriate analysis model.
(0028,0004) Photometric Interpretation — tells the viewer whether pixel values represent brightness directly (MONOCHROME2, where higher = brighter) or inversely (MONOCHROME1). Getting this wrong inverts the image — bones appear dark instead of bright.
More Than Pixels: Metadata Is the Real Value
A JPEG export of a radiograph contains only pixel data. A DICOM file contains the pixels plus hundreds of metadata tags: patient demographics, acquisition parameters (kVp, mAs, slice thickness), spatial information, and calibration data.
This metadata is why measurements taken in a DICOM viewer are physically accurate — the Pixel Spacing tag (0028,0030) tells the software exactly how many millimetres each pixel represents. A joint space width measurement of "4.2 mm" in a DICOM viewer is a calibrated measurement; the same measurement on a JPEG screenshot is an estimate at best.
For AI applications, this metadata is equally critical. Models trained on DICOM data can normalise for differences in acquisition equipment, patient positioning, and exposure settings — producing more reliable results than models that work from pixel data alone.
From PACS to Cloud: The Shift in Medical Imaging
Traditional PACS (Picture Archiving and Communication System) infrastructure requires dedicated on-premise servers, thick-client viewer software installed on hospital workstations, and VPN connections for remote access. This model has served hospitals well for two decades but creates friction for surgeons who need to review images from a clinic, from home, or during a conference.
Cloud-based DICOM viewers represent a new approach: upload DICOM files through a browser, and the viewer handles parsing, rendering, multiplanar reconstruction, and measurement — all client-side, without sending pixel data to a server.
The Salnus Surgeon Portal at app.salnus.com is built on this model. It uses Cornerstone3D for GPU-accelerated rendering, supports MPR viewing (axial, coronal, sagittal), and includes diagnostic tools (window/level presets, measurement overlays) — all running in the browser with zero server-side image processing.
DICOM and AI: The Connection
When an AI model analyses a knee radiograph, the DICOM file provides the input pipeline. The pixel data is extracted using the Pixel Data tag (7FE0,0010), normalised using the Window Center/Width values, and fed to the neural network. Spatial calibration from Pixel Spacing ensures that any measurements the AI produces (joint space width, osteophyte dimensions) are in real-world units.
For volumetric AI analysis (3D bone segmentation, tumour detection), a stack of DICOM slices sorted by Image Position Patient (tag 0020,0032) becomes a 3D volume that can be fed to segmentation models (nnU-Net, SAM-Med3D) for bone segmentation, cartilage analysis, or 3D mesh generation.
Practical Takeaways
If you are a surgeon working with medical imaging in 2026, these points matter: always request DICOM files rather than JPEG screenshots when receiving referral imaging — the metadata is essential for accurate measurement and AI analysis. When evaluating cloud viewers, ask whether processing happens client-side (your browser) or server-side (their cloud) — this has direct implications for patient data privacy and KVKK/HIPAA compliance. And if you are exploring AI tools for your practice, ensure they work directly with DICOM rather than requiring manual image conversion.
Disclaimer: This article is for educational purposes only. Clinical decisions should be made by qualified healthcare professionals.
Reviewed by the Salnus biomedical engineering team.