How a PACS System Works in Radiology

One study, end to end

1. 1

   The order. A referring physician orders an exam. The RIS receives it, schedules it, and registers the patient. The study now has an accession number, which is the identifier everything downstream depends on.

2. 2

   The worklist. The RIS publishes the scheduled study to a modality worklist, over DICOM. The technologist selects the patient at the scanner rather than typing their details, which is what keeps the identifiers consistent.

3. 3

   Acquisition. The modality produces the images and sends them to the PACS as DICOM objects, carrying both pixel data and metadata.

4. 4

   Archiving. The PACS stores the study. Most archives tier their storage: recent studies on fast media, older ones moved to cheaper and slower storage.

5. 5

   The read. The radiologist opens the study in a diagnostic viewer, with priors retrieved alongside it and the clinical context supplied by the RIS.

6. 6

   The report. The radiologist dictates. The report is stored and distributed by the RIS, and the referring physician receives it.

Step two is the one that decides whether everything else works. If the patient is keyed in by hand at the modality rather than selected from the worklist, the study can arrive at the archive with the wrong identifier. It no longer matches the order. It is stored correctly and still cannot be found.

The four parts of a PACS system

Whatever the vendor calls its product, a PACS is four things working together. Naming them separately is useful, because when something breaks it is almost always one of the four rather than "the PACS".

• The modalities. The CT, MRI, X-ray, ultrasound and mammography units that produce the images. They are not part of the PACS, but they are the source everything else depends on.

• The network. The path between the modalities, the archive and the workstations. Most "the PACS is slow" complaints resolve here.

• The archive. Storage plus a database of what is stored, which is what makes retrieval possible at all.

• The workstations and viewers. Diagnostic displays for radiologists, and lighter web viewers for everyone else.

How an image actually moves

The three operations below are named services in the DICOM standard, not vendor features, which is why a modality from one manufacturer can talk to an archive from another.

• Sending. When the scanner finishes, it pushes the study to the archive with C-STORE, defined in the DICOM Storage Service Class. Each image arrives with the metadata identifying the patient, the study and the equipment that produced it.

• Finding. When a radiologist opens a worklist, or the system goes looking for prior studies, it asks the archive what it holds using C-FIND.

• Retrieving. The images themselves come back with C-MOVE or C-GET, both defined in the DICOM Query/Retrieve Service Class.

Newer systems do the same three jobs over HTTP instead, using DICOMweb: STOW-RS to store, QIDO-RS to query, WADO-RS to retrieve. That matters practically, because a web-based service can reach a browser without a VPN or a client install.

Prefetching, and why priors appear before you ask

A radiologist reading a chest CT wants the last one to compare against. Waiting for a retrieval mid-read is expensive, so most systems prefetch. When a study is scheduled, the archive is asked in advance for that patient's relevant priors. They are pulled into fast storage before the reading session begins.

Prefetch depends entirely on the patient being matched correctly. If the prior sits under a different medical record number, or was acquired at a site whose archive is not reachable, the prefetch finds nothing and the radiologist reads without comparison. That is the single most common cause of a "missing prior", and it is usually an identity problem rather than a storage one.

Where studies live: online, nearline and offline

An archive is not one pool of storage. Studies move through tiers as they age, and the tier decides how fast they come back.

• Online. Recent studies on fast storage, retrievable in seconds. Typically the most recent weeks or months.

• Nearline. Older studies on slower, cheaper storage. Retrievable in minutes without anyone intervening.

• Offline. Deep archive for long-term retention. Retrieval can take materially longer and may need a request.

Retention is a legal question before it is a storage one. Requirements vary by state and by study type, paediatric records are generally kept longer than adult ones, and mammography carries its own rules under MQSA. Check what applies to you rather than accepting a vendor's default, because the default is set for storage cost rather than for your obligations.

How a PACS works in a hospital, and how that differs from an imaging center

The mechanism above is the same wherever the PACS sits, but two things change in a hospital and they change what the system has to do.

The PACS is not the only system holding the patient. In a hospital the EHR owns the patient record and the order usually originates there, so the PACS receives work that has already passed through another system. In an imaging center the RIS is normally the origin, which makes the RIS-to-PACS handoff the one that matters most.

Imaging is not only radiology. A hospital PACS is frequently asked to hold cardiology, endoscopy, pathology and wound imaging as well. That is where the pressure toward enterprise imaging comes from, and it is why hospital PACS procurement looks different from an imaging center's. The buyer is solving for several departments with incompatible expectations rather than for one reading workflow.

Everything else is the same. The modality sends, the archive stores, the viewer retrieves, and the report goes back to whoever ordered the study.

Why the PACS industry is consolidating archives

Multi-site groups often inherit a PACS per site. Each holds its own studies, which means a patient imaged at one location arrives at another with no visible history.

The usual answer is a vendor neutral archive: one store behind several PACS, so priors follow the patient rather than the building. Whether that is worth the project depends on how often your patients move between your sites.