How to Choose a PACS

A procurement guide for radiology directors, biomedical engineers, and imaging-center managers navigating software, storage, and standards in 2024–2025.

---

What this is and who buys it

A Picture Archiving and Communication System (PACS) is the clinical software infrastructure that acquires, stores, distributes, and displays digital medical images in DICOM format — covering modalities from CR and CT to MRI, ultrasound, and mammography. Unlike a traditional film library, a PACS makes studies simultaneously available to radiologists, referring physicians, and technologists across a network, enabling filmless workflow and teleradiology. The DICOM standard (published as NEMA PS3 and maintained by the ACR/NEMA DICOM Standards Committee) is the universal language that makes this interoperability possible [S3].

Primary buyers span hospital radiology and cardiology departments, multi-site integrated delivery networks (IDNs), independent imaging centers, ambulatory surgery centers, and specialty clinics in ophthalmology, oncology, and orthopedics. Procurement is almost always triggered by one of four scenarios: a legacy system approaching end-of-life, a new facility build-out, a transition from film-based to digital imaging, or a strategic shift from on-premise to cloud or hybrid architecture. Each scenario carries meaningfully different requirements, which is why a PACS RFP that works for a community hospital may be wholly inadequate for a 15-site IDN.

The market has also grown more complex. Enterprise PACS platforms now routinely embed AI orchestration layers, vendor-neutral archive (VNA) components, and web-based zero-footprint viewers — capabilities that were add-on modules a decade ago. Understanding where those components begin and end, who holds FDA clearance for each, and how they price is the central challenge of a modern PACS evaluation.

---

Key decision factors

DICOM conformance and interoperability should be the first document you request from any vendor — their current DICOM Conformance Statement per NEMA PS3.2 [S3]. This isn't a formality. Specifically verify support for DICOMweb protocols: WADO-RS, STOW-RS, and the web services defined in PS3.18. Without these, cloud-hosted deployments and zero-footprint browser viewers are effectively blocked. A vendor who hesitates to produce this document is telling you something important before the contract is signed.

Deployment architecture — on-premise, cloud, or hybrid — is a structural decision, not a preference. On-premise systems deliver lower per-study costs at high volumes (roughly above 5,000 studies per month is a common inflection point) and give the institution full data sovereignty, but they require capital for servers and NVMe/RAID storage, and an IT staffing overhead estimated at approximately 1/20 of an FTE per system. Cloud and hybrid models convert that capital expenditure to a predictable operating cost, shift disaster-recovery responsibility to the vendor, and compress deployment timelines from 6–18 months down to 4–12 weeks — but they require close scrutiny of SLA uptime guarantees (target ≥99.9%), data-residency jurisdiction, and vendor financial stability.

VNA vs. integrated storage is a distinction that procurement teams frequently overlook until a migration crisis makes it unavoidable. A PACS manages clinical workflow and viewing; a Vendor Neutral Archive handles long-term retention in a vendor-neutral format. Legacy PACS platforms that store images in proprietary wrappers create data-silo lock-in that can cost as much as the original system to unwind. Insist on native DICOM storage, and ask for a written export procedure and test export of 100 studies before any contract is executed.

RIS/HIS/EHR integration is not a nice-to-have — it determines whether a radiologist's worklist is populated, whether orders and reports flow correctly, and whether the PACS participates in cross-enterprise image sharing. Validate support for HL7 v2.x, HL7 FHIR R4, and IHE XDS.b/XCA profiles. If AI diagnostic tools are on the roadmap, factor in additional API licensing costs, which vendors rarely surface in headline pricing.

Concurrent user and volume licensing is the area most likely to generate cost surprises at renewal. Most vendors price on a combination of concurrent radiologist seats and annual study volume. Map your current volumes and project 5-year growth by modality before issuing an RFP; a mid-sized imaging center that grows from 40,000 to 70,000 studies per year can trigger a pricing tier jump that was nowhere in the initial proposal.

Cybersecurity deserves its own paragraph because PACS archives are high-value ransomware targets containing protected health information. Under the HIPAA Security Rule (45 CFR §164.312), the covered entity — not the vendor — is ultimately responsible for access controls, audit controls, integrity controls, and transmission security. Require AES-256 encryption at rest and in transit, role-based access controls, multi-factor authentication, and a documented vulnerability disclosure program. For cloud vendors, a signed Business Associate Agreement (BAA) is non-negotiable before any ePHI enters the system.

---

What it costs

PACS pricing varies enormously depending on deployment model, modality count, storage volume, and user seats. Published list prices are rare; most enterprise quotes require a formal RFP [S4]. Expect these broad bands:

• Entry tier ($5,000–$40,000): Mini-PACS for dental practices, small ASCs, or single-modality departmental deployments. Often SaaS-based with per-study pricing.
• Mid tier ($50,000–$300,000): Community hospital radiology departments, independent imaging centers, or hybrid-cloud deployments with 2–10 radiologist workstations and full RIS integration.
• Premium tier ($500,000–$10,000,000+): Enterprise imaging platforms for multi-site health systems or IDNs, including VNA, AI orchestration, cardiology and specialty modules, and phased data migration from legacy systems.

These figures reflect software licensing and initial implementation. Annual software maintenance contracts typically add 15–20% of the license cost per year for on-premise systems. Storage hardware, workstations, and data migration from a legacy archive are almost always quoted separately.

---

Common use cases

The correct PACS configuration differs substantially by clinical context, and applying an enterprise radiology platform where a departmental system would suffice — or vice versa — creates either unnecessary cost or functional gaps.

• Community hospital radiology department: Single-site, multi-modality archive (CR/DR, CT, MRI, ultrasound) with RIS integration and 3–10 concurrent workstations; typically mid-tier on-premise or hybrid cloud.
• Multi-site health system or IDN: Enterprise imaging platform with cross-site worklist federation, unified hanging protocols, and centralized VNA — premium-tier investment with a phased rollout measured in months, not weeks.
• Independent imaging center or teleradiology group: Cloud-native or SaaS PACS with a zero-footprint browser viewer and per-study pricing; lower upfront cost with variable opex tied to volume.
• Cardiology department: Requires hemodynamics integration, cath-lab DICOM support, ECG correlation, and structured reporting — a general radiology PACS is typically insufficient, and a specialty cardiology PACS should be evaluated separately.

---

Regulatory and compliance

PACS platforms with image processing, manipulation, or quantification functions — including multiplanar reconstruction (MPR) or 3D rendering — are classified as Class II Medical Image Management and Processing Systems (MIMPS) under 21 CFR 892.2050 (product code LLZ) and require 510(k) premarket clearance [S1]. Verify the vendor's 510(k) clearance number directly in the FDA CDRH database; do not rely on the vendor's word. Storage-only or display-only configurations may qualify as Class I under 21 CFR 892.2010/892.2020 and be exempt from 510(k), but any platform with quantification or processing tools does not qualify for that exemption [S2].

Following the 21st Century Cures Act, the FDA renamed the 892.2050 classification to MIMPS and clarified that certain pure storage and display functions may fall outside the device definition. The practical implication for procurement: confirm which specific software modules in the proposed solution hold 510(k) clearance and which are excluded. Bundled AI diagnostic algorithms — mammography CAD, lung nodule analysis, stroke detection — require their own separate 510(k) or De Novo clearance; the PACS clearance does not extend to them. Vendor software development should also comply with IEC 62304:2006/AMD1:2015 (Medical Device Software — Software Lifecycle Processes), an FDA-recognized consensus standard; request a Declaration of Conformity and safety class assignment. Separately, diagnostic display monitors used with any PACS must be calibrated to DICOM Part 14 (GSDF) and tested per AAPM TG-270 protocols — typically on an annual basis.

---

Service, training, and total cost of ownership

Implementation timelines diverge sharply by deployment model. Cloud and SaaS deployments at a single site commonly go live in 4–12 weeks, assuming DICOM routing and EHR integration are straightforward. On-premise enterprise deployments — with modality routing, RIS integration, workstation configuration, and legacy data migration — routinely run 6–18 months and require dedicated project management on both the vendor and buyer sides.

Training is frequently underbudgeted. A realistic estimate is $200–$700 per user for initial training across radiologist, technologist, and IT administrator roles. Before signing, confirm whether training is delivered on-site or virtually, whether it covers all user roles, and whether re-training during major version upgrades is included in the base contract or invoiced separately.

On-premise storage hardware has a practical useful life of 5–7 years before capacity, performance, or vendor support becomes a constraint. Budget a comparable capital reinvestment at that interval. Current best practice is a tiered storage architecture: NVMe or SSD for active and recent studies, with automated migration to lower-cost object storage for long-term archive. On-premise PACS software platforms, absent hardware constraints, typically have a functional lifespan of 7–12 years before AI integration gaps, OS compatibility issues, or vendor end-of-life make replacement necessary. Cloud/SaaS platforms are continuously updated, but vendor financial stability and contract portability carry their own longevity risks.

Data migration from a legacy system deserves explicit contract language — not a line item. Migrations of 10–100 TB require DICOM re-routing, study reconciliation against the RIS worklist, and formal validation; plan 3–12 months and budget for the possibility of a third-party migration service for archives above 50 TB. The parallel-run period — running old and new systems simultaneously for at least two weeks — is a patient-safety requirement, not an optional phase.

---

Red flags to watch for

Any vendor that stores images in a proprietary format rather than native DICOM is creating long-term lock-in that will cost you significantly at the next replacement cycle. Require written confirmation of native DICOM storage and conduct a test export of 100 studies before signing. Similarly, a vendor that cannot produce a current, version-specific DICOM Conformance Statement should not advance past the initial screening stage — this document is fundamental to evaluating interoperability, and its absence signals either a technically immature product or a vendor unwilling to be held to published standards.

Watch carefully for bundled AI tools presented as part of the PACS clearance. Each AI diagnostic module requires independent FDA review; verify clearance status for every algorithm individually in the 510(k) database. Finally, per-study pricing models without contractual volume caps are a structural budget risk: model your 5-year study growth curve against the pricing schedule before signing, and ensure any cloud vendor is willing to execute a HIPAA BAA — refusal is an automatic disqualifier.

---

Questions to ask vendors

1. Provide your current DICOM Conformance Statement. Does the system support DICOMweb (WADO-RS, STOW-RS per PS3.18), IHE Scheduled Workflow, and HL7 FHIR R4? What IHE Integration Profiles have been demonstrated at an IHE Connectathon?
2. What is your 510(k) clearance number under 21 CFR 892.2050, and which specific software modules are covered? Are any AI or CAD modules separately cleared, pending clearance, or explicitly excluded from the device classification?
3. Are images stored in native DICOM format with no proprietary wrapper? If we terminate the contract, what is the exact process and timeline to export our full archive, and is there a fee?
4. Provide an itemized 5-year TCO including software licensing, maintenance, storage (by projected study volume), user seats, training, and hardware if on-premise. What triggers a price increase at renewal?
5. What are your contractually guaranteed uptime SLA, RPO, and RTO targets? How frequently is failover tested, and where are data centers located?
6. Describe your migration methodology for our existing archive. What are the acceptance criteria for migration completeness, who resolves RIS reconciliation errors, and is migration included in the base contract price?

---

Alternatives

The on-premise versus cloud decision is not binary, and the right answer depends on study volume, IT staffing, and institutional risk tolerance. On-premise systems remain cost-competitive at high volumes and offer full data control, but hybrid architectures — a local edge node for active studies combined with cloud object storage for long-term archive — are increasingly the operational middle ground. For institutions with limited IT resources, fully managed PACS-as-a-service reduces internal burden but reduces visibility into the system stack and typically costs more over a 7–10 year horizon than a capital purchase with a maintenance contract.

On the capital-versus-subscription question: per-study or SaaS pricing eliminates upfront capex and works well at low-to-moderate volumes, but total cumulative spend over 7–10 years generally exceeds an equivalent capital purchase at high volumes. Fee-per-use contracts commonly run 60-month terms with minimum payment thresholds around 80% of contracted volume — model this against your actual utilization before committing. For associated hardware (workstations, servers), refurbished equipment can reduce capital cost, but verify OS compatibility, end-of-support dates, and impact on the vendor's support agreement before purchasing anything used. Finally, evaluate whether a departmental radiology PACS serves your actual scope, or whether an enterprise imaging platform that also consolidates cardiology and specialty imaging would reduce per-study storage cost and IT overhead across the organization — a more complex procurement, but one that avoids rebuilding the infrastructure in five years.

---

Sources

• 21 CFR 892.2050 FDA PACS/MIMPS Classification and Requirements — LegalClarity
• Where Do PACS Fit Into FDA Rules on Device Approvals — AuntMinnie
• DICOM Standard PS3.1 Overview — NEMA/DICOM Standards Committee
• How Much Does a New PACS Cost — Purview