Valvular & structural heart disease
Detecting Bicuspid Aortic Valve From Echocardiographic Reports Using Natural Language Processing: A Veterans Affairs Study
Authors' conclusion
This NLP approach enables large-scale retrospective identification of BAV patients from clinical text, creating the largest BAV cohort to date and facilitating future cardiovascular research and clinical decision-making.
Study context
- Population
- Echocardiographic reports from the Veterans Affairs Corporate Data Warehouse555 training, 170 validation reports
- Outcome
- BAV detection (valve leaflet structure identification)
Methodological detailsrule-based natural language processing (NLP) (MedSpaCy), development and held-out validation
- Rule-based NLP (MedSpaCy)
- Development and held-out validation
Performance
As reported in the abstract
BAV detection
| Test or model | Precision | Sensitivity | F1-score |
|---|---|---|---|
| NLP system | 0.925 | 0.939 | 0.932 |
NLP system
Precision0.925
Sensitivity0.939
F1-score0.932
Further findings
Affirmed bicuspid aortic valve (BAV) when applied to 14,453,591 echocardiographic documents from 3,478,658 patients
- Patients identified with affirmed BAV
- 83,461 (2.40%)
Concordance with ICD-10 code Q23.81
- NLP concordance among patients identified by ICD-10 code Q23.81
- 86.1%
- Manual review confirmed NLP accuracy in discordant cases
Original publication
Detecting Bicuspid Aortic Valve From Echocardiographic Reports Using Natural Language Processing: A Veterans Affairs Study. JACC Adv. Published 16 December 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →