Making Opportunistic Vertebral Fracture Detection Work: What NICE EVA HTE34 Means for NHS Imaging Services and How deepcOS® Supports It

Across the NHS, vertebral fragility fractures (VFFs) appear on routine CT scans every day. Many are not reported because the scan was taken for another reason, the fracture is subtle, or the main clinical question lies elsewhere⁽¹⁾. These “hidden” fractures are often the earliest sign of osteoporosis and the moment when secondary fracture prevention could begin⁽²⁾.

NICE’s new Early Value Assessment (HTE34) brings national focus to this issue⁽¹⁾. For the first time, NHS trusts can use and fund a defined set of AI tools for opportunistic VFF detection during a structured three-year evidence generation period. Several of these technologies are available through deepcOS® AI Infrastructure.

This blog outlines what this means for imaging services, why infrastructure is essential to real-world adoption, and how deepcOS® can help NHS organisations meet both operational needs and NICE’s evidence expectations.

What NICE EVA HTE34 Really Means for NHS Organisations

HTE34 does three important things⁽³⁾:

1. It authorises funded use with conditions.

Five CT-based AI solutions (including Nanox's HealthOST and HealthVCF, as well as IB Lab's FLAMINGO, available on deepcOS®) can be used with core NHS funding during a three-year evidence-generation period, provided that:

• they are used within the indicated populations
• they have appropriate regulatory and DTAC approval
• the trust and the vendor are actively generating the evidence defined by NICE

2. It does not grant routine adoption.

The recommendations are conditional. NICE will reassess the technologies after three years and determine whether they should become routine NHS practice.

3. It shifts responsibility onto local services.

Trusts must:

• manage clinical and operational risks
• anticipate impact on radiology, MSK specialists and DEXA capacity
• ensure clear local protocols for downstream care
• consider equality and potential subgroup variation
• and negotiate contracts with awareness of long-term technology availability

This makes VFF AI not just a technology decision, but an operational, governance, and system-wide planning decision.

Why This Matters for Imaging Services

For imaging leaders, the relevance is direct.

Many vertebral fragility fractures are visible on CT scans that radiology teams already report. When they are missed, variation increases and opportunities to diagnose and manage osteoporosis earlier are lost. Identifying a vertebral fragility fracture is often the moment that triggers assessment for osteoporosis and secondary fracture prevention⁽⁴⁾.

Opportunistic VFF detection strengthens reporting completeness and increases the likelihood that patients receive appropriate follow-up care using imaging already performed as part of routine pathways.

What the Evidence Shows

NICE reviewed retrospective studies showing that the recommended technologies can identify additional moderate and severe VFFs that were not mentioned in the original radiology reports. Across these studies, the tools demonstrated high sensitivity and specificity⁽³⁾.

While NHS-specific evidence remains limited, the committee concluded that these technologies are likely to improve detection rates when integrated into real NHS reporting workflows, especially given that most missed fractures are visible and reconstructable on CT.

Three of the recommended technologies, Nanox HealthOST & HealthVCF and IB Lab FLAMINGO,  are available through deepcOS® today.

How Opportunistic Detection Fits Into Reporting

Opportunistic vertebral fracture detection only makes a difference if the findings are surfaced inside the workflows radiologists already use.

On deepcOS®, the EVA-listed AI solution analyses the CT series and produces its result. deepcOS® then:

• brings the AI findings into the PACS viewer 
• can present notifications in the worklist
• and can deliver the structured output into deepcOS® AIR, the platform’s reporting solution, where the findings can be documented consistently across vendors.

The reporting practitioner:

• reviews the flagged area
• interprets the findings
• and retains full clinical responsibility.
  

NICE emphasises that AI is a decision aid, not an autonomous diagnostic step. Its purpose is to reduce the likelihood that clinically meaningful fractures are overlooked, particularly during busy lists or when the vertebrae are not the primary focus of the scan.

The Challenge: Workflows, Capacity and System Impact

NICE identifies several operational realities that trusts must plan for:

• More specialist reviews: Additional flagged fractures often need confirmation by MSK radiologists.
  
• More DEXA referrals: Capacity varies, and many services are already stretched.
  
• More coordination with FLS / bone health teams: Because a newly identified VFF often becomes the entry point for osteoporosis assessment.
  
• Age targeting matters: Most recommended solutions apply to patients aged 50+, reducing unnecessary system-wide activation.⁽⁴⁾
  
• Local variation: Imaging services differ in case mix, reporting models and downstream pathways.
  

This makes platform-based deployment essential: consistent routing, governance, monitoring and version control across scanners and sites.

Why Infrastructure Is the Missing Piece

AI adoption without infrastructure becomes fragile, costly and inconsistent. deepcOS® solves exactly the problems NICE warns about.

Targeted routing aligned with NICE indications

deepcOS® lets you route VFF AI only on:

• relevant CT series
• patients over the age thresholds
• or defined subgroups depending on capacity and risk.
  

This prevents unnecessary workload inflation and supports an evidence-based rollout.

Supporting downstream pathways

Within deepcOS®, trusts can configure how VFF-related AI findings move through reporting and operational workflows. For example, deepcOS® can:

• make AI findings visible in PACS, worklists, and the AIR® reporting solution, ensuring consistent documentation across vendors
• structure the AI output so that DEXA referral workflows or bone health teams can easily identify patients with a newly reported VFF
• standardise how results flow across scanners and sites, ensuring multi-site NHS trusts avoid variation in how VFF findings appear in reports
• support governance and audit by keeping model versioning, inference results, and acceptance/rejection decisions traceable in one place.

Vendor-neutral flexibility

With NICE’s 3-year evidence window, flexibility matters. deepcOS® allows you to:

• deploy multiple EVA-listed tools through a single integration
• swap or add solutions without rebuilding interfaces
• run comparative, side-by-side evaluations.
  

This protects your service from vendor lock-in while satisfying the need to generate evidence.

Turning EVA Requirements into a Real-World Evidence Plan with deepcOS®

NICE requests specific evidence gaps to be filled. deepcOS® supports all of them natively:

Diagnostic accuracy vs NHS standard care

Because all AI outputs, metadata, and study identifiers are stored centrally, deepcOS® enables trusts to run both retrospective and prospective evaluations without setting up separate pipelines. This allows:

• Retrospective studies on local CT data, by routing historical cohorts through the EVA-listed solutions and comparing AI findings with existing radiology reports.
  
• Identification of additional VFFs per 1,000 scans, by counting AI-detected fractures that were not mentioned in the initial report but confirmed during clinical audit.

Failure rates and causes

deepcOS® records:

• every successfully analysed case
• every failed inference
• vendor-specific error codes (where available)
• correlations with scanner type, image quality and acquisition protocol

Impact on radiology workload & downstream care

deepcOS® can provide aggregated views that help trusts understand system impact over time. For example:

• Flagged cases per modality and site: every inference is logged with modality and study metadata, so trusts can see how many CT scans are analysed and how many are flagged as potential VFFs by scanner, site or age band.
  
• Specialist review volumes: where trusts choose to record whether an AI finding was confirmed or overruled (for example, via simple accept/reject markers), deepcOS® can show how many cases required escalation or MSK review.
  
• DEXA referral patterns (where integrated): if DEXA orders or codes can be linked from RIS or another connected system to the original imaging episode, deepcOS® can support analysis of DEXA referral rates before and after AI introduction, by site or population group.⁽⁵⁾

Not all trusts will choose to connect every downstream dataset. The key point is that once AI outputs and imaging metadata are standardised and logged centrally, it becomes much easier to link them to other data sources and answer the questions NICE is asking about workload and pathway impact.

Equality & subgroup considerations

deepcOS® can support:

• aggregated demographic analyses
• comparison of AI performance across age, risk profiles and populations
• systematic evaluation of whether the AI behaves consistently across diverse groups.

A Defined Path Toward More Consistent Fracture Detection

NICE has not yet mandated routine adoption. Instead, HTE34 defines a structured three-year window in which the NHS can evaluate:

• reporting quality
• operational impact
• downstream care pathways
• and whether opportunistic VFF detection should become standard practice.

For imaging services, HTE34 provides something that has long been missing in this space: a clearer definition of what good looks like^(1,3).

More consistent identification of vertebral fragility fractures is becoming part of high-quality reporting in the NHS. Making that practical, safe, scalable and evidence-driven is exactly what deepcOS® is built for.