Pulmonary vascular disease & thromboembolism
Echocardiographic artificial intelligence for pulmonary hypertension classification
Authors' conclusion
This preliminary study suggests promising potential for our ML model in predicting echocardiographic PH.
Study context
- Population
- Patients who underwent right heart catheterisation, classified as non-PH, precapillary pulmonary hypertension (PH) or postcapillary PHderivation n=720
- Participants
- 885 patients
- Comparison
- Machine learning (ML) model vs guideline-based echocardiographic assessment for PH classification
- Outcome
- Macro-average classification accuracy for PH
Methodological detailsderivation cohort and prospective validation data set
- Derivation cohort and prospective validation data set
Performance
As reported in the abstract
Macro-average classification accuracy, derivation cohort
| Test or model | Value | AUC normal | AUC precapillary PH | AUC postcapillary PH |
|---|---|---|---|---|
| ML model | 59.4% | 0.789 | 0.766 | 0.742 |
| Guideline-based assessment | 51.6% |
Value
ML model59.4%
Guideline-based assessment51.6%
AUC normal
ML model0.789
Guideline-based assessment— not stated in the abstract
AUC precapillary PH
ML model0.766
Guideline-based assessment— not stated in the abstract
AUC postcapillary PH
ML model0.742
Guideline-based assessment— not stated in the abstract
- ML model vs Guideline-based assessment: P < 0.01 — derivation cohort
Further findings
| Outcome | P value |
|---|---|
| Macro-average PH classification accuracy, prospective validation data set59.4% vs 57.8%; comparable | P = 0.638 |
- Macro-average PH classification accuracy, prospective validation data set: validation n=165
Original publication
Echocardiographic artificial intelligence for pulmonary hypertension classification. Heart. Published 30 January 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →