Heart failure & cardiomyopathies
Fully Automatic AI-Based Quantification of Left Ventricular (LV) Mass in Echocardiography: A Multimodality Validation
Authors' conclusion
AI echocardiography provides superior accuracy, diagnostic performance, and reproducibility for LV mass assessment compared with expert evaluation
Study context
- Population
- Patients undergoing echocardiography and computed tomography (CT)bedside test-retest cohort n = 40
- Participants
- 218 patients
- Comparison
- Artificial intelligence (AI)-based vs expert manual echocardiography LV mass measurement, CT as reference
- Outcome
- Accuracy of echocardiographic LV mass vs CT (ICC, MAPE)
Performance
As reported in the abstract
Agreement with CT for LV mass
| Test or model | Intraclass correlation coefficient (ICC), hybrid method | Mean absolute percentage error (MAPE), hybrid method | ICC, linear method | MAPE, linear method |
|---|---|---|---|---|
| AI | 0.76 | 0.16 | 0.49 | 0.44 |
AI
Intraclass correlation coefficient (ICC), hybrid method0.76
Mean absolute percentage error (MAPE), hybrid method0.16
ICC, linear method0.49
MAPE, linear method0.44
- AI vs Expert: reduction in measurement error, mainly due to more consistent interventricular septal diameter (IVSD) and posterior wall diameter (PWD) measurements
- ∼20%
- RWT+ cases reclassified by AI vs Expert
- 34%
Further findings
| Outcome | AI | Expert |
|---|---|---|
| Diagnostic accuracy for increased LV mass (AUC) | ||
| AUC, linear | 0.78 | 0.71 |
| AUC, hybrid | 0.84 | 0.77 |
| Test-retest reliability (coefficient of variation), bedside cohort | ||
| CV, both mass methods | 7.8% and 8.1% | — not stated in the abstract |
— not stated in the abstract
Test-retest reliability (coefficient of variation), bedside cohort
- Test-retest reliability highest with AI for both mass methods; reproducibility of manual assessment varied with operator skills
Original publication
Fully Automatic AI-Based Quantification of LV Mass in Echocardiography: A Multimodality Validation. JACC Cardiovasc Imaging. Published 29 October 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →