Heart failure & cardiomyopathies
Detection of Left Ventricular Outflow Obstruction From Standard B-Mode Echocardiogram Videos Using Deep Learning
Authors' conclusion
In this study, we developed an artificial intelligence model to detect LVOT obstruction from standard apical 4-chamber videos, highlighting patients who may benefit from more detailed cardiac workup for obstructive HCM.
Study context
- Population
- Patients with left ventricular outflow tract (LVOT) obstruction and control patients matched by age, sex, and septal thickness, with non-Doppler apical 4-chamber B-mode echocardiographic videos2,396 patients with LVOT obstruction and 6,177 control patients · patients with LVOT obstruction 2,396 · control patients 6,177 · Cedars-Sinai Medical Center test set (videos) 3,848
- Participants
- 2,396 vs 6,177
- Comparison
- Deep learning model on non-Doppler apical 4-chamber B-mode echocardiographic videos against outflow obstruction identified later in the same study by spectral Doppler
- LVOT obstructionn = 2,396
- control patientsn = 6,177
Methodological detailsmatched by age, sex, and septal thickness, held-out test sets from three centres
- Matched by age, sex, and septal thickness
- Held-out test sets from three centres
Main finding
As reported in the abstract
Detection of LVOT obstruction, Cedars-Sinai Medical Center test set
AUC
0.858
95% CI 0.847–0.870
Further findings
Detection of LVOT obstruction, Kaiser Permanente Northern California cohort
- AUC
- 0.817 (95% CI 0.740-0.922)
Detection of LVOT obstruction, Stanford Healthcare
- AUC
- 0.836 (95% CI 0.807-0.827)
- CI as printed in the abstract
Performance across patient subgroups
- Performance was consistent across patient subgroups, including those with hyperdynamic left ventricular function, pre-existing valvular disease, and small left ventricular cavity size
Original publication
Detection of Left Ventricular Outflow Obstruction From Standard B-Mode Echocardiogram Videos Using Deep Learning. JACC Adv. doi:10.
Summarised from the published abstract; every number is checked against it