The under-recognition of hypertrophy, measuring error and variability, and difficulty distinguishing causes of increased wall thickness, such as hypertrophy, cardiomyopathy, and cardiac amyloidosis, all restricted early diagnosis and characterization of increased left ventricular (LV) wall thickness. For a study, the researchers sought to find how accurate a deep learning methodology was at measuring ventricular hypertrophy and predicting the source of increasing LV wall thickness. From January 1, 2008, to December 31, 2020, the cohort research included physician-curated cohorts from the Stanford Amyloid Center and Cedars-Sinai Medical Center (CSMC) Advanced Heart Disease Clinic for cardiac amyloidosis, as well as the Stanford Center for Inherited Cardiovascular Disease and the CSMC Hypertrophic Cardiomyopathy Clinic for hypertrophic cardiomyopathy. The deep learning method was developed and tested on retroactively gathering independent echocardiography films from Stanford Healthcare, CSMC, and the Unity Imaging Collaborative. The key outcome was the accuracy of the deep learning system in evaluating left ventricular dimensions and identifying individuals with hypertrophic cardiomyopathy and cardiac amyloidosis who had increased LV wall thickness. With parasternal long-axis videos, 12,001 from Stanford Health Care (6,509 [54.25] female; mean [SD] age, 61.6 [17.4] years) and 1,309 from CSMC (808 [61.7] female; mean [SD] age, 62.8 [17.2] years) and 8,084 from Stanford Health Care (4,201 [54.05] female; mean [SD] age, 69.1 [16.8] years) and 2,351 from CSMS (6,509 [54.2%] female; mean [SD] age, 69.6 [14.7] years) with apical 4-chamber videos. 23,745 patients in total were included in the research. The deep learning algorithm accurately evaluated intraventricular wall thickness (MAE, 1.2 mm; 95% confidence interval, 1.1-1.3 mm), LV diameter (MAE, 2.4 mm; 95% CI, 2.2-2.6 mm), and posterior layer thickness (MAE, 1.4 mm; 95% CI, 1.2-1.5 mm) and classified cardiac amyloidosis (area under the curve [AUC], 0.83) and hypertrophic cardiomyopathy (AUC, 0.98) separately from other causes of LV hypertrophy. The deep learning technique accurately measured ventricular parameters in various data sets from different domestic and international health care systems (domestic: R2, 0.96; international: R2, 0.90). The MAE for intraventricular septum thickness was 1.7 mm (95% CI, 1.6-1.8 mm), LV internal dimension was 3.8 mm (95% CI, 3.5-4.0 mm), and LV posterior wall thickness was 1.8 mm (95% CI, 1.7-2.0 mm) in the domestic data set. The MAE for intraventricular septum thickness was 1.7 mm (95% CI, 1.5-2.0 mm), LV internal dimension was 2.9 mm (95% CI, 2.4-3.3 mm), and LV posterior wall thickness was 2.3 mm (95% CI, 1.9-2.7 mm) in the international data set. In the domestic external validation site, the deep learning algorithm correctly diagnosed cardiac amyloidosis (AUC, 0.79) and hypertrophic cardiomyopathy (AUC, 0.89). The deep learning algorithm correctly recognized modest changes in LV wall geometric metrics and hypertrophy caused in the cohort research. Unlike human specialists, the deep learning method was automated, allowing for repeatable, exact measurements and perhaps laying the groundwork for precision heart hypertrophy diagnosis.

 

Link:jamanetwork.com/journals/jamacardiology/fullarticle/2789370