Global Imaging AI Adoption 2025: KLAS Findings

Imaging AI adoption is accelerating worldwide, with about half of organisations live, Europe leading and Latin America and the Middle East and Africa around 30%. Adoption rises with imaging volume. Most use cases are pixel AI, including CT head for stroke, mammography and chest x-ray, plus MSK fractures and bone age. Platforms from PACS vendors are widely considered. Average active use is 2.2 cases, with faster acquisition, triage and impression generation emerging.

Key Points

• Europe leads imaging AI adoption across surveyed regions.
• About half of organisations have imaging AI live in routine use.
• Adoption increases with annual imaging volume of providers.
• Pixel AI dominates with stroke, mammography and chest x-ray use.
• AI platforms from PACS vendors are widely considered for scale.

KLAS Research’s Global Imaging AI 2025 report takes stock of how imaging artificial intelligence is taking root across health systems outside the United States. Drawing on feedback from 369 organisations in Asia-Oceania, Canada, Europe, Latin America and the Middle East-Africa region, it shows rapid uptake of AI and broad experimentation. Almost half of respondents already run at least one imaging AI tool not just in pilots or research, but in day-to-day practice. The mix spans pixel-level analysis, workflow and reporting support, and platform layers that knit multiple algorithms together.

Europe Leads Global Adoption

Adoption has surged, yet the pace varies by geography and size. Europe leads the field thanks to coordinated national initiatives, particularly in the United Kingdom, and generally higher healthcare IT budgets. Western and northern Europe show especially lively activity, even as representation from parts of southeastern and eastern Europe remains lighter in the sample.

By contrast, Latin America and the Middle East-Africa report lower overall adoption, at roughly three in ten organisations. Many providers in these regions license tools from Europe, Asia or the U.S., which in turn increases the need for local validation to ensure performance on regional populations. Asia-Oceania sits around the one-third mark for live use, and Canada is close to that. KLAS’ separate U.S.-focused 2024 report found that 52% of U.S. organisations had imaging AI in routine use. Adoption was higher among high-volume providers, reaching roughly two-thirds for sites performing ≥500k studies per year.

Scale matters. Smaller sites performing fewer than 100,000 annual studies are least likely to be live, often citing limited capacity to purchase, validate and implement. Adoption climbs as volume rises: midsize and especially very large organisations pass the halfway mark, with the largest providers adopting at roughly three-quarters.

What Organisations Are Actually Using

KLAS groups use into three broad families:

• Pixel AI: algorithms that read images and videos to assist diagnosis.
• Operational and reporting AI: tools that streamline acquisition, triage and documentation.
• AI platforms: orchestrators or marketplaces that bring many algorithms together under one roof.

Pixel AI dominates current practice. Stroke assessment on CT head, breast imaging in screening and diagnostics, and chest radiography are the big three. Beyond these, organisations report musculoskeletal fracture detection, bone age estimation, cardiac CT and MR, chest CT and brain MR for neurological conditions. Labels can blur from site to site: for instance, what one team calls lung imaging another may log as chest CT. However, the pattern is clear: clinically high-impact, high-volume pathways are getting attention first.

Operational and reporting AI is gaining ground. Faster image acquisition, particularly for MRI, and automated worklist triage are notable use cases. Early interest is also forming around generative tools that draft the radiology impression for clinicians to edit. Platform adoption still sits in the 10–20% range today, yet it is the most frequently mentioned future move as organisations seek cleaner governance and integration.

On average, respondents report 2.2 active use cases. More than seven in ten have one or two. A handful run far more, but the overall picture suggests steady, staged expansion rather than a wholesale switch-over.

Names That Come Up Most Often

Across the sample, respondents mention 108 different vendors. A handful recur frequently:

• RapidAI features strongly in stroke workflows and is validated across all five regions in the sample.
• Gleamer is widely used for musculoskeletal work, particularly fracture detection, and has broad European reach with mentions on other continents as well.
• annalise.ai is prominent for chest x-ray and head CT, with strong traction in the UK and presence across Southeast Asia and Australia.
• Lunit appears widely for chest and mammography, especially among larger providers.
• ScreenPoint Medical is used for mammography and benefits from integration with several Picture Archiving and Communication System (PACS)-based platforms.

Platform interest centres on established imaging IT players: Sectra Amplifier, AGFA HealthCare RUBEE AI and Philips AI Manager are most often cited, with GE HealthCare, PaxeraHealth, Fujifilm and MV also in the mix. Independent platforms, including Blackford Analysis, Aidoc, deepc and CARPL.ai, draw attention where organisations want a vendor-neutral layer or need to span multiple PACS environments. Blackford Analysis, which is owned by Bayer and primarily considered in English-speaking Australia and Canada, is included based on research carried out before Bayer announced plans to discontinue its AI platform business, a move perceived as a significant setback for third-party AI platforms.

Why Platforms Matter

A clear theme in the findings is consolidation around platform models. Hospitals want simpler validation, cleaner contracting and smoother integration, especially when they run several PACS systems or expect to mix algorithms from multiple suppliers. PACS vendors are extending their role by offering orchestration and marketplaces, letting radiology teams access a portfolio of tools through a familiar environment. Independent platforms meet similar needs where neutrality is preferred.

Traditional imaging IT remains central. In the respondent pool, Sectra and AGFA HealthCare are the most frequently mentioned PACS vendors, with Philips, GE HealthCare and Fujifilm also present. Intelerad, Dedalus, PaxeraHealth, MV, INFINITT, Medsynaptic, PACS-native offerings in specific markets, and several others round out a varied ecosystem.

Who Is Adopting — And How Deeply

The sample spans everything from sites handling fewer than 20,000 studies a year to those reading more than a million. Roughly one-third fall between 100,000 and 299,000 annual studies. The relationship between volume and adoption is clear: the bigger the throughput, the more likely imaging AI is live. Smaller sites tend to move carefully, given the resource demands of selection, validation and implementation.

Even where AI is live, breadth varies. Most organisations run one or two use cases; a few deploy far more. In practice, that means imaging AI is shifting from novelty to routine, but not yet saturating every pathway. The direction is steady expansion rather than abrupt transformation.

What Comes Next

Although many providers already have imaging AI in place, the market is still early in its maturity curve. A large share of organisations intend to add use cases or bring in additional vendors. Platforms look set for the sharpest growth, giving radiology and enterprise imaging teams a common chassis for evaluation, deployment, monitoring and updates.

Operational gains are likely to widen too. Faster acquisitions and smarter triage can free capacity and shorten turnaround times, while early experiments in drafting the impression point toward semi-automated documentation that keeps clinicians firmly in the loop.

What the Data Shows

KLAS’ Global Imaging AI 2025 report captures a field in motion. Around half of non-U.S. organisations in the sample have imaging AI live today, primarily targeting stroke, chest and breast imaging. Europe sets the pace, while Latin America and the Middle East-Africa are preparing to move faster. Large-volume providers are the earliest adopters, and platform strategies from PACS vendors and independents are becoming the default route to scale. The trend is unmistakable: from one-off tools to coordinated ecosystems, imaging AI is settling into everyday clinical work, preparing for its next phase of growth.

Conflict of Interest

None.