Closing the Bedside Diagnostic Gap: What a WHO Case Study Reveals About AI-Assisted Point-of-Care Cardiac Ultrasound
AISAP's point-of-care cardiac ultrasound platform, deployed at Sheba Medical Center, has been documented as a case study in a WHO Regional Office for Europe report on artificial intelligence implementation in health care. This article is written from the perspective of someone who was involved in executing that implementation, with the aim of offering a description of a specific clinical problem, the evidence generated by addressing it, and, perhaps more usefully, the implementation process itself - a part of the story that is rarely documented in enough detail to be genuinely useful to other institutions considering similar deployments.
A Recognized Gap in Health Workforce Capacity
Cardiac assessment has long depended on a scarce resource: clinicians who can both perform and interpret cardiac ultrasound at an expert level. In emergency departments, internal medicine wards and outpatient clinics, this dependency creates a structural bottleneck. When a patient presents with symptoms suggestive of heart failure or valvular disease, the diagnostic pathway often runs through a formal echocardiography service - a process that can take hours or days, not minutes. For non-cardiologists working at the front line of care, the absence of immediate, expert-level diagnostic capability at the bedside has been one of the most persistent gaps in routine clinical practice.
This gap is not a local anomaly. It reflects a broader global challenge in health workforce capacity. The World Health Organization has projected a shortfall of 11 million health workers by 2030 - a figure that helps explain why artificial intelligence in health care is increasingly framed not as a replacement for clinical expertise, but as a way to extend it.
Why This Case Study Carries Weight
The case study describing our work at Sheba was selected for inclusion in Bridging Theory and Practice - Implementation Insights on Artificial Intelligence in Health Care, produced by Working Group 1 (WG1) of the WHO Strategic Partners' Initiative for Data and Digital Health (SPI-DDH), and published by the WHO Regional Office for Europe in 2026.
SPI-DDH itself was established following a resolution of the WHO Regional Committee for Europe, tasking the WHO Regional Office for Europe with convening a strategic partnership council to encourage collaborative innovation in digital health across the European Region. WG1's membership reflects that mandate: ministries of health from across Europe and Central Asia, the OECD, the International Committee of the Red Cross, the International Organization for Migration, HIMSS, the Digital Medicine Society, EIT Health and a range of academic and health-system representatives from Finland, Ireland, Portugal, Sweden, Türkiye, the United Kingdom and beyond. The report itself was co-funded by the European Union.
Of an initial pool of many candidate submissions gathered through a structured questionnaire process, WG1 selected eleven case studies from seven countries for inclusion, using criteria that included the maturity of the project, the availability of genuine real-world implementation experience, and contributors' willingness to engage in detailed follow-up review. That selection process, and the institutional weight behind the report, is part of why this case study is a meaningful external validation of the work.
The Clinical Problem We Set Out to Address
At Sheba Medical Center, the challenge was well defined. Clinicians across emergency medicine, internal medicine and outpatient settings needed a way to assess cardiac structure and function without waiting for a cardiologist or a scheduled echocardiogram. The limited availability of experts capable of performing and interpreting cardiac ultrasound was identified as the critical constraint - not a lack of imaging equipment, but a lack of interpretive capacity at scale. This is a familiar pattern in diagnostic imaging AI more broadly: the hardware exists; the bottleneck is expert throughput.
The Intervention: AI-Assisted Point-of-Care Ultrasound (POCUS)
Sheba Medical Center partnered with AISAP to deploy a cloud-based AI platform built around 12 FDA-cleared modules, designed to support non-cardiologists in real-time interpretation of cardiac ultrasound and automated quantitative assessment of structural heart disease. In practice, this means a clinician without formal echocardiography training can acquire a cardiac ultrasound at the bedside and receive AI-supported, quantitative interpretive guidance in the same clinical encounter - collapsing a diagnostic pathway that traditionally spanned different departments and time frames into a single point-of-care workflow.
The Evidence: What Changed in Clinical Practice
The WHO case study cites a prospective study of 660 patients that quantifies the clinical impact of this approach:
- 55% of patients had a direct change in clinical decision-making based on AI-supported ultrasound findings.
- Over 30% of cases led to treatment escalation or de-escalation, early discharge, or a transition in the level of care.
- Scans were completed in under five minutes.
- Inter-observer agreement between AI-supported bedside assessment and formal echocardiography reports was high.
By the second quarter of 2025, the platform had been used for more than 3,000 scans across six departments and the emergency room, with close to 150 trained clinicians actively using the system as part of routine workflows and electronic medical record integration. These are not pilot-stage anecdotes; they describe a system that moved from proof of concept into sustained clinical use - which is precisely the transition that most health-care AI deployments fail to complete.
Implementation: The Part of the Story That Usually Goes Untold
What distinguishes this case study - and much of the WHO report's broader argument - is the attention paid to how the technology was adopted, not just what it achieved. Having been directly involved in the deployment, I can confirm that several of these dynamics were exactly as demanding in practice as they read in the report.
1. A deliberately phased rollout. Deployment did not begin as a hospital-wide initiative. It started with two clinicians in a single internal medicine department, expanded to three departments and six users based on early feedback and demonstrated clinical impact, then scaled to six departments and fifteen users, before ten emergency department clinicians were onboarded. This incremental, evidence-gated expansion model is a recurring theme across the WHO report's case series: successful AI adoption in clinical settings tends to be nonlinear and adaptive rather than a single large-scale launch.
2. Structured but insufficient-on-its-own training. A standardized 2-3 hour training program combined instruction on the AISAP platform, focused cardiac ultrasound education, and hands-on scanning tailored to each clinician's prior experience. Importantly, training alone did not produce proficiency - clinicians needed repeated, real-world use to build confidence in applying the tool under clinical pressure. This distinction between initial competence and sustained proficiency is a point too often glossed over in AI deployment literature.
3. Iterative workflow engineering. The interface itself evolved based on real-world use, with technical adjustments made specifically to reduce the number of clicks and shorten the time between image acquisition and interpretation. In a point-of-care setting, seconds matter, and interface friction is not a cosmetic issue - it determines whether a tool is actually used under time pressure.
4. Cultural and organizational change management. Moving from a workflow built around delayed, centralized echocardiography interpretation to one built around real-time, AI-supported bedside diagnosis required active clinician engagement and departmental alignment - not merely technical integration. Adoption depended on support from department heads and institutional leadership to embed the platform within standard clinical policy, not just clinical enthusiasm at the individual level.
Why This Case Study Matters Beyond Cardiology
The Sheba Medical Center implementation is instructive for anyone evaluating AI-enabled diagnostic imaging, clinical decision support, or point-of-care ultrasound more broadly - whether the audience is a hospital innovation office, a medical device developer, or a health system policy-maker. The WHO report situates this case within a wider framework for evaluating AI capabilities in health care (prediction, classification, association and optimization), and within a growing set of structured evaluation tools - including the OPTICA checklist developed at Clalit Health Services, the Coalition for Health AI's model card framework, and NHS England's Digital Technology Assessment Criteria - all aimed at helping health organizations assess clinical appropriateness, data quality and deployment risk before and during adoption.
Across the WHO report's eleven case studies, several patterns recur that are consistent with what was observed at Sheba: technical sophistication alone does not determine success. What differentiates AI deployments that progress beyond pilot stage is strong clinical involvement, integration with existing workflows, alignment with organizational priorities, and the presence of internal champions able to bridge technical and clinical perspectives. The implementation at Sheba was supported institutionally through the hospital's ARC (Accelerate, Redesign, Collaborate) Center for Digital Innovation, which enabled ongoing collaboration between clinicians and AISAP's technical team through real-time feedback and phased deployment.
A Replicable Model, Not a Finished Story
The WHO case study is careful to frame this as one institution's implementation experience rather than a universal blueprint. AISAP's platform has since expanded across additional departments at Sheba and to other institutions, but the documented evidence in the report is specific to Sheba Medical Center's experience. That specificity is precisely what makes it useful: it offers a concrete, evidence-based account of how a point-of-care cardiac ultrasound AI tool moved from two clinicians in one department to a multi-department, emergency-room-integrated diagnostic workflow, with measurable effects on clinical decision-making.
For health systems considering AI-based diagnostic support - in cardiology or elsewhere - the lesson is less about the algorithm and more about the surrounding infrastructure: phased deployment, sustained training beyond initial onboarding, iterative interface refinement, and institutional alignment from department leadership. Technology can shorten the path from image to interpretation. Whether it changes clinical practice at scale depends on all of the above.
This article draws on the case study "AI Implementation for Point-of-Care Cardiac Ultrasound (Israel)," published in Bridging Theory and Practice - Implementation Insights on Artificial Intelligence in Health Care*, the report of Working Group 1 of the WHO Strategic Partners' Initiative for Data and Digital Health (SPI-DDH), WHO Regional Office for Europe, 2026.*
