Mixed-method evaluation will test AI stress echocardiography in coronary artery disease care
Artificial intelligence is moving into cardiac imaging faster than many health systems can judge its real-world value. The EASE study sets out to examine EchoGo Pro, an AI-driven stress echocardiography analysis system developed by Ultromics Ltd, in a setting that matters clinically: the detection of cardiac ischaemia and possible coronary artery disease.
This is not a single-endpoint validation exercise. It is a mixed-method protocol designed to look beyond diagnostic accuracy and ask whether clinicians trust the output, where the system agrees or conflicts with cardiologists, what barriers block implementation, and whether the technology offers acceptable value for money in the NHS and beyond.
Diagnostic accuracy will be compared with manual interpretation
At the center of the study sits work package 1, where reports generated by EchoGo Pro will be compared with interpretation by three manual raters. The design places the AI system in direct contact with conventional expert reading, which is the only way to understand whether algorithmic support adds clarity or simply reproduces existing practice in a new format.
Stress echocardiography already depends on careful pattern recognition, and the clinical stakes are high. Missed inducible ischaemia can delay treatment for patients with obstructive CAD, while false-positive reads can trigger unnecessary downstream testing, anxiety, and costs. An AI system that can reliably flag abnormal studies could reduce variability in interpretation, especially in settings where reading expertise is unevenly distributed.
One sentence captures the core clinical challenge: the output has to be accurate enough to influence decisions, not just interesting enough to demonstrate technical promise.
The protocol does not present results yet. Its value lies in the structure of the comparison itself, which will allow investigators to judge whether the AI system matches, exceeds, or diverges from human readers when stress echo findings are translated into clinical judgments about cardiac ischaemia.
Clinician and patient acceptance will shape adoption
Work package 2 turns to the people who would actually live with the technology. Investigators plan interviews with clinicians, innovation and transformation staff, patients within the National Health Service, and staff within Ultromics. That combination allows the evaluation to move from laboratory-style performance into the messy realities of workflow, confidence, and institutional readiness.
Acceptance in cardiac imaging rarely depends on one feature alone. A system may be technically strong and still fail if clinicians do not trust the recommendations, if the interface slows workflow, or if the model behaves unpredictably in borderline cases. Patients bring another layer of meaning: whether they understand an AI-assisted report, whether they feel reassured by it, and whether they view machine support as an aid or an intrusion.
The study will also examine convergence and divergence between EchoGo Pro recommendations and cardiologists’ interpretations. Those disagreements may be the most clinically revealing part of the project. Divergence can expose edge cases in image quality, loading conditions, wall-motion assessment, or the thresholds clinicians use when deciding whether a study is truly abnormal. Agreement, by contrast, may show where AI can reinforce existing expertise and standardize decisions across sites.
Trust in AI will be measured across clinical and public audiences
Work package 4 widens the lens further with a quantitative cross-sectional survey of trust in AI tools used in cardiac care settings. The survey population includes clinicians, healthcare commissioners, and the general public. That design acknowledges a practical truth: adoption is not determined only by those reading the images, but also by those funding services and those receiving care.
Trust in AI is not abstract in this setting. If clinicians doubt the output, they may ignore it. If commissioners do not see value, deployment stalls. If the public sees AI as opaque or unsafe, acceptance can erode even when technical performance is sound. Survey data should help map where confidence is strong, where skepticism clusters, and which stakeholder groups need the most support before wider implementation is realistic.
Cost-effectiveness will be tested against NHS deployment pressures
Work package 5 focuses on cost, cost-effectiveness, and willingness to pay among cardiologists, healthcare commissioners, and the general public. This matters because stress echocardiography is not just a diagnostic test; it is a pathway gateway. Every decision influenced by image interpretation can ripple into further imaging, invasive angiography, medication changes, or discharge.
An AI system that speeds interpretation or reduces unnecessary variation could lower resource use. An AI system that generates extra uncertainty could do the opposite. The protocol is built to estimate deployment cost and to ask whether stakeholders believe the technology is worth paying for. In a publicly funded system such as the NHS, that question determines whether promising software becomes routine practice or remains a pilot.
Implementation questions are as important as algorithm performance
The EASE study reflects a broader shift in cardiac imaging research. Technical accuracy alone no longer settles the argument around AI adoption. Hospitals need to know whether the tool fits existing pathways, how clinicians react to disagreements with the algorithm, what training or oversight is required, and whether the system offers sufficient value to justify procurement and maintenance.
Because the study is mixed-method, it can capture both numbers and narrative. Quantitative comparisons will address diagnostic performance and economic value, while interviews and surveys should expose the human obstacles that often decide whether digital tools survive beyond early enthusiasm. In practice, these are the questions that separate a useful clinical support system from software that adds friction without changing care.
The protocol is also notable for its setting. Evaluation in the NHS gives the project a real-world frame, where resource constraints, variation in service delivery, and the need for scalable diagnostics are part of daily care. Findings from this work may therefore speak to other health systems that are considering AI support for stress echocardiography, but the immediate test is whether the technology can earn a place in routine cardiac assessment.
Ethics approval and dissemination are already in place
The study has received approval from the NHS Health Research Authority under IRAS No: 315284 and from the London South Bank University Ethics Panel, ETH2223-0164. Planned dissemination includes journal publications, conference presentations, seminars, and social media, along with research summaries and policy briefs tailored for NHS audiences.
Because this is a protocol, the real clinical question remains open: whether EchoGo Pro can prove accurate, acceptable, implementable, and economically defensible when it is tested against human interpretation in routine stress echocardiography. The answer will depend on results from all five work packages, not on diagnostic performance alone.