AI Task-Shifting Revolutionizes Echocardiographic LVEF Assessment at National Heart Centre Singapore

In a groundbreaking advancement for cardiovascular diagnostics, researchers at the National Heart Centre Singapore (NHCS) have demonstrated that artificial intelligence (AI) task-shifting can transform how left ventricular ejection fraction (LVEF)—a critical measure of heart pumping efficiency—is assessed through echocardiography. This innovation allows minimally trained novices to perform point-of-care (POC) ultrasound scans with AI guidance, achieving accuracy comparable to expert sonographers while slashing costs and wait times.

With heart failure (HF) prevalence in Singapore exceeding the global average at around 4.5% and rising amid an ageing population—where cardiovascular diseases account for over 30% of deaths—this AI-driven approach addresses surging demand for echocardiograms. Traditional assessments rely on scarce skilled sonographers, leading to backlogs, but AI empowers primary care and task-shifting to non-experts, revolutionizing access to timely HF screening and treatment.

Understanding LVEF and Echocardiography Basics

Left ventricular ejection fraction (LVEF) quantifies the percentage of blood pumped out of the heart's left ventricle with each contraction, typically ranging 50-70% in healthy adults. Values below 50% signal reduced LVEF (HFrEF), guiding therapies like ACE inhibitors or SGLT2 inhibitors. Echocardiography, or cardiac ultrasound, visualises heart structures and function via sound waves, but requires expertise for image acquisition and interpretation—processes prone to variability.

In Singapore, where HF affects thousands and echo demand outstrips supply due to allied health shortages, bottlenecks delay diagnosis. AI task-shifting delegates acquisition to novices using handheld devices, with algorithms automating analysis, bypassing expert bottlenecks.

Challenges in Conventional Echocardiography

Standard transthoracic echocardiography (TTE) demands trained sonographers for high-quality apical views essential for LVEF calculation via Simpson's biplane method. Singapore faces radiographer and sonographer shortages, mirroring global trends, exacerbating wait times amid HF's 148% rise in Southeast Asia over three decades.

• Sonographer scarcity: Critical roles like echo techs in high demand, with job listings surging.
• High costs: TTE sessions cost S$1,403 per patient.
• Workflow delays: Hours for analysis, limiting scalability in primary care.
• Expert variability: Inter-observer LVEF differences up to 10%.

These hurdles hinder early HF intervention, where prompt LVEF-guided therapy improves outcomes.

The PANES-HF Study: Proof-of-Concept at NHCS

The Point-of-care AI-enhanced Novice Echocardiography for Screening Heart Failure (PANES-HF) trial at NHCS pioneered this shift. A novice coordinator, trained just two weeks (observation plus hands-on), used EchoNous Kosmos handheld ultrasound with AI TRIO real-time guidance—fanning/rotating probes for optimal views—and Us2.ai for automated LVEF from seven standard clips.

In 100 symptomatic patients (mean age 61, 56% male), yield reached 96% interpretable scans (scan time ~13 min). For LVEF <50% (27 cases):

• AUC: 0.880
• Sensitivity: 84.6%, Specificity: 91.4%
• Correlation with expert TTE: r=0.713
• Median absolute deviation: 6.03%

Novice learning curve: 60% time drop in first 20 scans, plateauing thereafter, with consistent accuracy. Outperformed NT-proBNP for LVEF detection.PANES-HF full study

Economic Evaluation: Quantifying Cost Savings

Building on PANES-HF, a March 2026 ESC Heart Failure analysis modeled decision trees for LVEF <50% diagnosis. Novice AI-POC: S$1,185/patient vs. sonographer TTE: S$1,403—saving S$218/patient, or S$21,669 for 100 cases at one centre.

Probabilistic sensitivity: 99.9% likelihood of savings. Inputs: PANES-HF accuracy, Singapore HF prevalence, costs (devices, training, staff). Scalable to polyclinics, easing tertiary burdens.Economic study details

Photo by Andy Wang on Unsplash

How AI Task-Shifting Works Step-by-Step

1. Training: 2 weeks: observe sonographers, practice landmarks (ASE guidelines).
2. Acquisition: Handheld probe on chest; AI TRIO labels structures, grades quality real-time.
3. Analysis: Us2.ai processes DICOM clips, computes LVEF via deep learning (validated vs. experts).
4. Output: Instant report; high NPV rules out HFrEF, flags referrals.
5. Integration: Primary care screening → therapy start → confirmatory TTE if needed.

Benefits and Risks of AI in Echo Task-Shifting

• Benefits: Faster (13 min vs. hours), cheaper, accessible (POC portable), high NPV for screening, addresses shortages.
• Risks: 4% non-interpretable (learning mitigated), AI bias (validated multi-ethnic), over-reliance (human oversight).
• Patient Attitudes: Positive; UTAUT2 study shows acceptance for AI task-shifting.

For higher-ed jobs in AI health, Duke-NUS leads training.

Singapore's Healthcare Context and NHCS Leadership

Singapore's ageing (25% over 65 by 2030) drives HF surge; CVD kills 13,500/year. NHCS, Asia's top heart centre, pioneers AI via CVS.AI Lab, partnering A*STAR. Beyond LVEF, SENSE AI speeds CAD CT (10 min vs. hours). Task-shifting aligns MOH goals for efficient care.NHCS AI news

Sonographer jobs abound; AI augments, not replaces.Singapore higher ed opportunities

Academic Collaborations Driving Innovation

Duke-NUS Medical School and NUS Saw Swee Hock School anchor research; authors like Jasper Tromp (NUS/Duke-NUS) bridge academia-clinical. Us2.ai (Singapore-based) validates with Kosmos. Ties to global: EchoNous, ESC. Trains next-gen via academic CV tips.

Future Outlook: Scaling AI Task-Shifting Globally

Ongoing trials: AI-ECG LVEF (NCT07038018). Potential: polyclinics, telehealth, LMICs. Challenges: regulation, equity. NHCS eyes nationwide rollout, cutting HF burdens. For careers in AI cardiology, explore university jobs at Duke-NUS/NUS.

Optimistic: "Task-shifting... cost-saving alternative."

Photo by Galen Crout on Unsplash

Stakeholder Perspectives and Actionable Insights

Clinicians praise scalability; patients embrace (UTAUT2). Insights: Train GP staff, integrate EHRs, monitor outcomes. Links: Rate My Professor for AI health educators; higher-ed jobs in med tech.