Caroline Currie
Caroline Currie has a total of 23 co-authors. Here are the top co-authors by number of shared publications:
Spencer Kieu
Eko Health Inc. Emeryville CA USA
2 shared publications- Adam E. Saltman
Eko Devices, Inc. Oakland CA USA
1 shared publication - Anand Subramanian
Sri Jayadeva Institute of Cardiovascular Sciences and Research
1 shared publication Cholenahalli Manjunath
Sri Jayadeva Institute of Cardiovascular Sciences and Research Bengaluru India
1 shared publicationCody Hitchcock
Eko Devices, Inc. Oakland CA USA
1 shared publicationDaniel Barbosa
Eko Devices, Inc. Oakland CA USA
1 shared publication
Latest Publications
- Open Access
- 4 February 2025
- Published by Wolters Kluwer Health in Journal of the American Heart Association
- Vol. 14 (3) , e036882
Development and Evaluation of a Deep Learning–Based Pulmonary Hypertension Screening Algorithm Using a Digital Stethoscope
Background: Despite the poor outcomes related to the presence of pulmonary hypertension, it often goes undiagnosed in part because of low suspicion and screening tools not being easily accessible such as echocardiography. A new readily available screening tool to identify elevated pulmonary artery systolic pressures is needed to help with the prognosis and timely treatment of underlying causes such as heart failure or pulmonary vascular remodeling. We developed a deep learning–based method that uses phonocardiograms (PCGs) for the detection of elevated pulmonary artery systolic pressure, an indicator of pulmonary hypertension. Methods: Approximately 6000 PCG recordings with the corresponding echocardiogram‐based estimated pulmonary artery systolic pressure values, as well as ≈169 000 PCG recordings without associated echocardiograms, were used for training a deep convolutional network to detect pulmonary artery systolic pressures ≥40 mm Hg in a semisupervised manner. Each 15‐second PCG, recorded using a digital stethoscope, was processed to generate 5‐second mel‐spectrograms. An additional labeled data set of 196 patients was used for testing. GradCAM++ was used to visualize high importance segments contributing to the network decision. Results: An average area under the receiver operator characteristic curve of 0.79 was obtained across 5 cross‐validation folds. The testing data set gave a sensitivity of 0.71 and a specificity of 0.73, with pulmonic and left subclavicular locations having higher sensitivities. GradCAM++ technique highlighted physiologically meaningful PCG segments in example pulmonary hypertension recordings. Conclusions: We demonstrated the feasibility of using digital stethoscopes in conjunction with deep learning algorithms as a low‐cost, noninvasive, and easily accessible screening tool for early detection of pulmonary hypertension.
- 4 February 2025
- Published by Wolters Kluwer Health in Journal of the American Heart Association
- Vol. 14 (3) , e036882
Open AccessDevelopment and Evaluation of a Deep Learning–Based Pulmonary Hypertension Screening Algorithm Using a Digital Stethoscope
Background: Despite the poor outcomes related to the presence of pulmonary hypertension, it often goes undiagnosed in part because of low suspicion and screening tools not being easily accessible such as echocardiography. A new readily available screening tool to identify elevated pulmonary artery systolic pressures is needed to help with the prognosis and timely treatment of underlying causes such as heart failure or pulmonary vascular remodeling. We developed a deep learning–based method that uses phonocardiograms (PCGs) for the detection of elevated pulmonary artery systolic pressure, an indicator of pulmonary hypertension. Methods: Approximately 6000 PCG recordings with the corresponding echocardiogram‐based estimated pulmonary artery systolic pressure values, as well as ≈169 000 PCG recordings without associated echocardiograms, were used for training a deep convolutional network to detect pulmonary artery systolic pressures ≥40 mm Hg in a semisupervised manner. Each 15‐second PCG, recorded using a digital stethoscope, was processed to generate 5‐second mel‐spectrograms. An additional labeled data set of 196 patients was used for testing. GradCAM++ was used to visualize high importance segments contributing to the network decision. Results: An average area under the receiver operator characteristic curve of 0.79 was obtained across 5 cross‐validation folds. The testing data set gave a sensitivity of 0.71 and a specificity of 0.73, with pulmonic and left subclavicular locations having higher sensitivities. GradCAM++ technique highlighted physiologically meaningful PCG segments in example pulmonary hypertension recordings. Conclusions: We demonstrated the feasibility of using digital stethoscopes in conjunction with deep learning algorithms as a low‐cost, noninvasive, and easily accessible screening tool for early detection of pulmonary hypertension.
- Open Access
- 17 October 2023
- Published by Wolters Kluwer Health in Journal of the American Heart Association
- Vol. 12 (20) , e030377
Deep Learning Algorithms to Detect Murmurs Associated With Structural Heart Disease
Background: The success of cardiac auscultation varies widely among medical professionals, which can lead to missed treatments for structural heart disease. Applying machine learning to cardiac auscultation could address this problem, but despite recent interest, few algorithms have been brought to clinical practice. We evaluated a novel suite of Food and Drug Administration‐cleared algorithms trained via deep learning on >15 000 heart sound recordings. Methods and Results: We validated the algorithms on a data set of 2375 recordings from 615 unique subjects. This data set was collected in real clinical environments using commercially available digital stethoscopes, annotated by board‐certified cardiologists, and paired with echocardiograms as the gold standard. To model the algorithm in clinical practice, we compared its performance against 10 clinicians on a subset of the validation database. Our algorithm reliably detected structural murmurs with a sensitivity of 85.6% and specificity of 84.4%. When limiting the analysis to clearly audible murmurs in adults, performance improved to a sensitivity of 97.9% and specificity of 90.6%. The algorithm also reported timing within the cardiac cycle, differentiating between systolic and diastolic murmurs. Despite optimizing acoustics for the clinicians, the algorithm substantially outperformed the clinicians (average clinician accuracy, 77.9%; algorithm accuracy, 84.7%.) Conclusions: The algorithms accurately identified murmurs associated with structural heart disease. Our results illustrate a marked contrast between the consistency of the algorithm and the substantial interobserver variability of clinicians. Our results suggest that adopting machine learning algorithms into clinical practice could improve the detection of structural heart disease to facilitate patient care.
- 17 October 2023
- Published by Wolters Kluwer Health in Journal of the American Heart Association
- Vol. 12 (20) , e030377
Open AccessDeep Learning Algorithms to Detect Murmurs Associated With Structural Heart Disease
Background: The success of cardiac auscultation varies widely among medical professionals, which can lead to missed treatments for structural heart disease. Applying machine learning to cardiac auscultation could address this problem, but despite recent interest, few algorithms have been brought to clinical practice. We evaluated a novel suite of Food and Drug Administration‐cleared algorithms trained via deep learning on >15 000 heart sound recordings. Methods and Results: We validated the algorithms on a data set of 2375 recordings from 615 unique subjects. This data set was collected in real clinical environments using commercially available digital stethoscopes, annotated by board‐certified cardiologists, and paired with echocardiograms as the gold standard. To model the algorithm in clinical practice, we compared its performance against 10 clinicians on a subset of the validation database. Our algorithm reliably detected structural murmurs with a sensitivity of 85.6% and specificity of 84.4%. When limiting the analysis to clearly audible murmurs in adults, performance improved to a sensitivity of 97.9% and specificity of 90.6%. The algorithm also reported timing within the cardiac cycle, differentiating between systolic and diastolic murmurs. Despite optimizing acoustics for the clinicians, the algorithm substantially outperformed the clinicians (average clinician accuracy, 77.9%; algorithm accuracy, 84.7%.) Conclusions: The algorithms accurately identified murmurs associated with structural heart disease. Our results illustrate a marked contrast between the consistency of the algorithm and the substantial interobserver variability of clinicians. Our results suggest that adopting machine learning algorithms into clinical practice could improve the detection of structural heart disease to facilitate patient care.