The most important design feature of the framework is not the number five. It is the relationship among the pillars. Enterprise governance determines who can decide. Standardization determines the default method. Capital planning determines where scarce financial resources are placed. Digital informatics makes the operating model executable and measurable. Workforce strategy supplies the human capacity and expertise to deliver care. Reduction of unwarranted variation emerges when these five controls act on the same workflows using shared definitions and transparent data.

This system view is consistent with the development of enterprise imaging over the past decade. HIMSS-SIIM work has repeatedly emphasized governance, information architecture, clinical workflows, financial stewardship, and technology integration as foundational enterprise concerns rather than isolated PACS tasks.[1-4] More recently, a peer-reviewed enterprise-wide modality-based management transformation reported improvements across staffing, turnover, engagement, access, efficiency, volume, policy review, and safety review, illustrating that organizational structure itself can be a performance intervention.[5]

Figure 1. The strategic shift from department-centric management to an enterprise radiology operating model. Transformation changes the unit of management from the department to the enterprise imaging system.

Table 2. Interdependencies that determine whether the five pillars create enterprise value

Table 3. Recommended tiered governance model

Select a tier to see its membership, decisions, and cadence.

How to monitor governance

Sustainability and future improvement

Governance becomes sustainable when it is embedded in management routines rather than dependent on individual leaders. Charters should survive leadership transitions; decision-rights matrices should be reviewed annually; council membership should rotate without losing domain expertise; and unresolved exceptions should be visible in executive performance reviews.

Figure 2 rebuilt: imaging stewardship changed yield, time and cost together

Values reproduced from Turpin et al., Journal of Pediatric Surgery, 2026, as printed in Figure 2. Each relative change was recomputed here from the figure’s own printed pre and post values and matches the printed percentage exactly: 27.1 to 56.0 percent gives +106.6 percent, 347 to 299 minutes gives -13.8 percent, and $6,177 to $4,812 gives -22.1 percent.

View the published figure

Figure 2 as published. Example of the operational value of standardized imaging stewardship. In a pediatric appendicitis pathway, improved ultrasound performance was associated with higher diagnostic ultrasound yield and lower clinical team time and imaging cost per patient.[8]

Table 4. The enterprise standardization cycle

Select a stage to see its required action and control question.

What should be standardized first?

Monitoring and sustainability

The central monitoring concept is a variation registry. Enterprise leaders should be able to see where performance differs, whether the difference is expected, and whether it is stable over time. Useful measures include protocol concordance, repeat and reject rates, nondiagnostic examination rates, contrast protocol variance, dose indices, exam duration, patient preparation failures, add-on rates, cancellation reasons, report template adoption, critical-result closure, and access by site.

Sustainability requires both technical and social infrastructure. Holdsworth and colleagues’ evaluation of radiology learning collaboratives found that structured quality-improvement education, expert-guided improvement, deliberate participant interaction, accountability, and credible leadership supported sustained improvement.[22] An enterprise standard therefore needs more than publication. It needs a community of practice, visible metrics, periodic review, and a credible clinical leader who can explain why the standard matters.

Why radiology capital planning is different

Radiology is unusually capital-intensive. CT, MRI, PET/CT, interventional systems, mammography, ultrasound fleets, PACS and VNA infrastructure, advanced workstations, coils, injectors, shielding, construction, cooling, electrical systems, service contracts, cybersecurity, and software licensing create a portfolio whose costs extend far beyond purchase price. Asset decisions also shape workflow for years. A scanner purchased without enterprise protocol, interoperability, staffing, service, and capacity assumptions can lock the organization into operating variation long after the capital committee has adjourned.

Lifecycle evidence supports a forward-looking approach. The European Society of Radiology has recommended an imaging equipment upgrade and replacement plan that looks at least five years ahead and is updated annually, noting the clinical and operational risks of aging equipment.[9] Life-cycle costing studies in radiodiagnosis similarly demonstrate that acquisition choices should be evaluated using the total stream of operating and technology costs rather than initial purchase price alone.[10]

Figure 3. Proposed enterprise capital prioritization matrix. The plotted assets are illustrative; the framework is intended to force explicit comparison of strategic value and lifecycle risk. The Asset Scorer tab plots your own asset onto these same quadrants from the Table 5 dimensions.

The capital decision should answer six questions

1. Clinical need. Does the asset materially improve diagnostic or therapeutic capability, quality, safety, or access?
2. Capacity need. Is the true constraint equipment, staffing, room availability, preparation, anesthesia, radiologist capacity, scheduling, or downstream flow?
3. Lifecycle risk. What is the probability and operational consequence of failure, vendor end-of-support, cybersecurity exposure, unavailable parts, or software obsolescence?
4. Enterprise fit. Does the asset conform to platform, protocol, coil and accessory, service, data, and training standards across the fleet?
5. Economic value. What is the total cost of ownership, incremental contribution, avoided cost, staffing implication, construction burden, and opportunity cost across the useful life?
6. Strategic value. Does the investment support priority service lines, geographic access, academic and research missions, differentiation, or network growth?

Table 5. Proposed imaging asset health index

Note the scale direction. Five of the six dimensions score 5 as worse, but strategic fit scores 5 as higher strategic consequence, which is a reason to invest rather than a defect. The Asset Scorer tab treats strategic fit as the horizontal axis of Figure 3 and the remaining five as the vertical risk axis, which is what the matrix requires.

Monitoring the capital portfolio

A capital dashboard should be continuously updated, not reconstructed during budget season. At a minimum it should include asset age, vendor support horizon, service cost, unplanned downtime, exam volume, utilization by hour, access delay, cancellations attributable to equipment, image quality concerns, cybersecurity status, staffing requirements, operating margin contribution, and known construction or infrastructure dependencies. Capital governance should review the portfolio at least quarterly, while the formal multi-year plan is refreshed annually.

From PACS management to digital production architecture

Traditional PACS management focused on image storage, retrieval, display, and uptime. Those functions remain essential, but enterprise radiology now requires a broader architecture: order and encounter workflows, modality worklists, identity, DICOM, HL7 and FHIR integration, VNA, enterprise viewers, image exchange, universal worklists, structured reporting, clinical communication, follow-up management, analytics, AI orchestration, cybersecurity, and cloud infrastructure. The informatics function should therefore be governed as part of the clinical production system rather than treated as a technical support service.[2,3,20]

Figure 4. Proposed enterprise imaging digital stack. The architecture connects acquisition to clinical outcomes through interoperable workflow and a governed AI and analytics layer. Governance, cybersecurity, data quality, identity management, and change control span every layer.

AI is an operating-model decision, not a software purchase

The maturation of radiology AI has exposed the limits of procurement-led implementation. A technically accurate algorithm can fail to create enterprise value if it is poorly integrated, applied to the wrong bottleneck, priced inappropriately, ignored by users, or allowed to drift after deployment. Shah and colleagues propose a practical lifecycle built around whether an AI tool works, helps, and stays effective: validation, deployment, value assessment, and post-deployment surveillance.[12] Venugopal and colleagues extend this logic by describing monitoring for temporal stability, predictive divergence, uncertainty, tiered alerting, root-cause analysis, and human oversight after deployment.[13]

Figure 5 rebuilt: the economic evidence funnel

Values reproduced from Molwitz et al., Radiology: Artificial Intelligence, 2026, as printed in Figure 5. The published figure prints the inclusion share as one percent; recomputed from its own counts, 21 of 1,879 is 1.12 percent. The 88 full texts evaluated appear in the figure only and are not stated in the body text.

View the published figure

Figure 5 as published. The 2026 systematic review by Molwitz et al. found that only 21 of 1,879 screened publications met criteria for quantified economic outcomes, underscoring the gap between technical AI literature and robust value evidence.[14]

1,879Records screened

The published search yield before eligibility screening.

21Economic-outcome studies

Studies meeting criteria for quantified economic outcomes.

23.9%Of full texts evaluated

Author-derived: 21 of the 88 full texts reached the economic-outcome threshold. The published figure reports the share of the 1,879 screened instead.

Value is context-dependent. The same review concluded that the economic value of AI was more favorable in high-volume or resource-intensive tasks when performance was adequate and pricing aligned with scale. Pay-per-use structures and lower specificity could erode economic value. AI should therefore compete for investment using the same clinical, operational, and economic discipline applied to scanners and other major technologies.

Table 6. Enterprise AI and informatics lifecycle

The AI Lifecycle tab turns this table into a phase-gate tracker that will not advance a tool past a phase whose required evidence has not been recorded.

PACS and enterprise informatics measures that matter

The demand problem is durable

US imaging demand is projected to increase substantially over the coming decades. Christensen and colleagues projected 2055 utilization to be 16.9 percent to 26.9 percent above 2023 levels across major modalities in a baseline scenario, with population growth explaining most of the increase and aging contributing additional growth.[15] That projection does not automatically prove a universal radiologist shortage, but it does establish a durable capacity challenge. Modality mix, geography, subspecialty distribution, attrition, local workflow, and productivity determine whether supply and demand are aligned in a given enterprise.

Figure 6 rebuilt: projected 2055 utilization increase versus 2023, baseline scenario

Values reproduced from Christensen et al., Journal of the American College of Radiology, 2025, as printed in Figure 6. The printed modality values span 16.9 percent (MRI) to 26.9 percent (nuclear medicine), which reproduces the 16.9 to 26.9 percent range stated in the body text exactly.

View the published figure

Figure 6 as published. Projected US imaging utilization growth by modality through 2055 under the baseline scenario reported by Christensen et al.[15] The baseline scenario assumes current per-person utilization persists.

Supply projections are also sensitive to assumptions about training and attrition. Christensen and colleagues estimated that the number of radiologists could grow by 25.7 percent by 2055 if residency positions remained flat and by 40.3 percent if recent residency growth continued; higher attrition would materially reduce those totals.[16] This means the management question is not whether there will be enough radiologists in the abstract. It is whether the enterprise can design work to convert available physician and nonphysician expertise into reliable clinical capacity.

Workload is not distributed evenly

Large-scale registry evidence demonstrates why averages can conceal operational risk. In an analysis of 46.4 million examinations across 167 US facilities, the average number of examinations per radiologist per day changed little overall, yet the top quartile of radiologists experienced a 30.6 percent increase in examinations per day and a 19.7 percent increase in clinical days worked per quarter.[17] The workforce can therefore expand while workload concentration worsens for selected clinicians.

Figure 7 rebuilt: workforce and workload shifts, ACR registry cohort

Values reproduced from Zamani et al., Journal of the American College of Radiology, 2026, as printed in Figure 7. Two of the four printed values, the 23.6 percent increase in working radiologists and the 4.9 percent decline in days per radiologist per quarter, appear in the figure only and are not stated in the report’s body text. The Workload Concentration tab uses them.

View the published figure

Figure 7 as published. Workforce and workload shifts from the ACR registry cohort analyzed by Zamani et al.[17] The result illustrates why enterprise capacity monitoring must examine distributions, not only averages. Based on 46.4 million examinations across 167 facilities.

Turnover provides a second signal. National claims-based analysis found radiologist turnover rising from 5.3 percent in 2013 to 8.5 percent in 2022.[18] Retention literature emphasizes compensation, leadership, development, social connection, recognition, well-being, and job design rather than recruitment alone.[19] In this context, chronic overload should be treated as an operational control failure with workforce consequences, not merely as an individual resilience problem.

Table 7. Workforce architecture: levers beyond recruitment

Implementing a demand-capacity workforce model

The operating model should begin with demand rather than current staffing. Volume should be segmented by modality, site, hour, day, acuity, exam complexity, procedure time, and required professional skill. Capacity should then be modeled separately for scanners, technologists, nursing, radiologists, procedural staff, anesthesia, transport, registration, and rooms.

Monitoring and sustainability

Band definitions

Bands are the arithmetically obvious partition of the report’s own 1 to 5 scale. They are not validated cut points and no external benchmark is implied.

Pillar 5 states two findings from the same author team, published as companion papers in the same journal and year, both indexed to 2023 and both projecting to 2055. Demand rises 16.9 to 26.9 percent depending on modality.[15] Radiologist supply rises 25.7 percent if residency positions stay flat, or 40.3 percent if recent residency growth continues.[16] The report then makes the argument this model exists to test: the real question is not the headline comparison, it is whether the enterprise can convert available expertise into reliable clinical capacity.

Modality detail

Demand growth values are as printed in Figure 6. Demand index is 1 plus that growth. Margin and ratio are computed against the effective supply index above.

How to read this honestly

The report uses two of Figure 7’s four values in its text: the top quartile’s 30.6 percent increase in examinations per day and 19.7 percent increase in clinical days worked per quarter. The other two, a 23.6 percent increase in working radiologists and a 4.9 percent decline in days per radiologist per quarter, are printed in the figure but not discussed. These summary changes can be combined for sensitivity analysis. Their product is an illustrative index, not a directly observed total; top-quartile membership and joint distributions are not fixed by these summary statistics.

Illustrative index ratio

Table 5 supplies the dimensions and the anchors. Figure 3 supplies the two axes and the four quadrant names. This tab joins them: the five risk dimensions form the vertical axis, strategic fit forms the horizontal axis, and the plotted point falls into whichever quadrant Figure 3 defines. The quadrant boundaries sit at the midpoint of the printed scale, which is where Figure 3 places its own Moderate gridlines.

Live plot on the Figure 3 matrix

Quadrant names and axis labels reproduce Figure 3. The plotted point is yours. The six illustrative assets from the published figure are shown faintly for orientation; their positions are read from the published figure and are illustrative there as well.

View the published figure

Figure 3 as published. Proposed enterprise capital prioritization matrix. Scoring should integrate safety, downtime, capacity, cybersecurity, vendor support, total cost of ownership, and service-line strategy.

Dimension profile

Strategic fit is drawn in a separate colour because its scale runs in the opposite direction: a high score there argues for investment, while a high score on any other dimension describes a problem.

The report’s argument is that AI failure is usually an operating-model failure rather than an algorithm failure. A technically accurate tool can still fail if it is poorly integrated, applied to the wrong bottleneck, priced badly, ignored by users, or allowed to drift.[12,13] Table 6 answers that by attaching required evidence to every phase. This tab holds the sequence to it.

A high-quality radiology scorecard should resist two common errors. The first is metric proliferation, where hundreds of indicators create noise rather than control. The second is output fixation, where leaders monitor volume and report turnaround without measuring the processes that create them. The recommended scorecard uses a limited set of enterprise measures with clear definitions and assigns each measure an owner, a threshold, a review cadence, an escalation rule, and a balancing measure.

Table 8. Proposed enterprise radiology excellence scorecard

A practical monthly operating review

1. Start with enterprise outcomes. Safety, access, patient experience, quality, workforce stability, and financial performance.
2. Review exceptions and outliers before averages. The purpose is to identify where the operating system is not behaving as designed.
3. Separate common-cause variation from special-cause events before launching corrective actions.
4. Require every red metric to identify five things: the constraint, the countermeasure, the owner, the expected effect, the balancing measure, and the review date.
5. Close the loop. At the next review, determine whether the countermeasure changed the target process and whether any new risk was created.

Figure 8. Proposed five-level maturity model for radiology enterprise excellence. The staircase runs from site-defined management with reactive controls to predictive controls, rapid-cycle improvement, monitored AI, and continuous reinvestment.

Table 9. Maturity descriptors by pillar

Select a level to read across all five pillars at once. Table 9 prints descriptors for levels 1, 3 and 5; levels 2 and 4 are named in Figure 8 but not itemized by pillar, and are shown here as the transitions they describe rather than invented descriptors.

Figure 9. Proposed 24-month implementation roadmap: establish control before attempting scale. The exact sequence should be adjusted for local risk, readiness, and capital cycles.

First 90 days: establish control

1. Appoint and charter. Name the executive sponsor and enterprise imaging accountable executive; approve the charter and decision-rights matrix.
2. Build one inventory. Sites, modalities, assets, software, interfaces, workforce, protocols, and major vendors, in a single validated record with owners.
3. Define the KPI dictionary. Baseline the most material access, quality, workforce, capital, and digital measures against one approved definition each.
4. Name the top ten of each. Identify the top 10 enterprise variation problems and the top 10 capital and lifecycle risks using objective criteria.
5. Select no more than three pilots. Each should test multiple pillars, such as MRI access, CT protocol standardization, or enterprise AI triage.

Table 10. Executive readiness checklist

A concise readiness test for use before launching an enterprise imaging transformation. Answer each item, then check the evidence column: the report requires evidence, not assertion.

Verification ledger

Method notes

Structural audit of the source

Selected peer-reviewed evidence base

Numbered references correspond to citations in the report. Filter by topic to find the evidence behind a particular pillar.