Heart failure & cardiomyopathiesDiagnostic or prognostic study
Invasive haemodynamic validation of an artificial intelligence echocardiographic tool for detecting heart failure with preserved ejection fraction
- external validation
- retrospective
- real-world referral cohort
Authors' conclusion
This AI model demonstrated high specificity and positive predictive value (PPV) for detecting HFpEF, reliably identifying patients with more advanced haemodynamic abnormalities. Its performance remained robust across variable intervals between transthoracic echocardiography (TTE) and right heart catheterization (RHC), and in patients with indeterminate clinical scores. Due to limited sensitivity, the tool is best utilized to enrich identification of patients with clearly abnormal haemodynamics rather than to exclude HFpEF, particularly in early or borderline cases. While broader use as a screening tool is promising, prospective validation studies are necessary to confirm its utility in general populations.
Study context
- Population
- Patients with unexplained dyspnoea who underwent transthoracic echocardiography (TTE) and right heart catheterization (RHC)including 28 with both rest and exercise haemodynamics
- Participants
- 47 patients
- Comparison
- Artificial intelligence (AI) tool for detecting heart failure with preserved ejection fraction (HFpEF) compared to invasive haemodynamic criteria
- Outcome
- Invasively defined HFpEF: PCWP ≥15 mmHg at rest or ≥25 mmHg with exercise
Performance
As reported in the abstract
| Test or model | sensitivity | specificity | PPV | NPV |
|---|---|---|---|---|
| AI tool | 30% | 88% | 86% | 33% |
- Patients with exercise RHC (n = 28) meeting haemodynamic HFpEF criteria
- 71%
Further findings
| Outcome | AI tool | Invasive haemodynamic criteria | Estimate | P value |
|---|---|---|---|---|
| PCWP/CO slope >2 mmHg/L/min (secondary analysis, n = 25) | ||||
| specificity | 100% | Not reported in the abstract | ||
| PPV | 100% | Not reported in the abstract | ||
| sensitivity | 27% | Not reported in the abstract | ||
| NPV | 16% | Not reported in the abstract | ||
| Haemodynamics in AI-positive vs AI-negative patients — AI-positive · significantly higher in AI-positive | resting PCWP 20 vs 15 mmHg | P = 0.029 | ||
| Haemodynamics in AI-positive vs AI-negative patients — AI-positive · significantly higher in AI-positive | mean PA pressure 29 vs 24 mmHg | P = 0.02 | ||
| Haemodynamics in AI-positive vs AI-negative patients — AI-positive · significantly higher in AI-positive | PVR 2.1 vs 1.3 WU | P = 0.002 |
Indeterminate H2FPEF scores (n = 25)
- AI correctly identified, of those with invasively confirmed HFpEF
- 80%
Performance by TTE-RHC interval
- Model performance was consistent across TTE-RHC intervals <365 and <90 days
Original publication
Summarised from the published abstract; every number is checked against it