Using Deep learning to Predict Cardiovascular Magnetic Resonance Findings from Echocardiography Videos.
Authors
Affiliations (6)
Affiliations (6)
- Department of Cardiology, Smidt Heart Institute, Cedars-Sinai Medical Center, Los Angeles, CA, USA.
- Department of Cardiology, Smidt Heart Institute, Cedars-Sinai Medical Center, Los Angeles, CA, USA; Biomedical Imaging Research Institute, Cedars-Sinai Medical Center, Los Angeles, CA, USA.
- Biomedical Imaging Research Institute, Cedars-Sinai Medical Center, Los Angeles, CA, USA.
- Department of Imaging and Medicine, Cedars-Sinai Medical Center, Los Angeles, CA, USA.
- Department of Cardiology, Smidt Heart Institute, Cedars-Sinai Medical Center, Los Angeles, CA, USA; Division of Artificial Intelligence in Medicine, Cedars-Sinai Medical Center, Los Angeles, CA, USA. Electronic address: [email protected].
- Department of Cardiology, Smidt Heart Institute, Cedars-Sinai Medical Center, Los Angeles, CA, USA. Electronic address: [email protected].
Abstract
Echocardiography is the most common modality for assessing cardiac structure and function. While cardiac magnetic resonance (CMR) imaging is less accessible, CMR can provide unique tissue characterization including late gadolinium enhancement (LGE), T1 and T2 mapping, and extracellular volume (ECV) which are associated with tissue fibrosis, infiltration, and inflammation. Deep learning has been shown to uncover findings not recognized by clinicians, however it is unknown whether CMR-based tissue characteristics can be derived from echocardiography videos using deep learning. To assess the performance of a deep learning model applied to echocardiography to detect CMR-specific parameters including LGE presence, and abnormal T1, T2 or ECV. In a retrospective single-center study, adult patients with CMRs and echocardiography studies within 30 days were included. A video-based convolutional neural network was trained on echocardiography videos to predict CMR-derived labels including LGE presence, and abnormal T1, T2 or ECV across echocardiography views. The model was also trained to predict presence/absence of wall motion abnormality (WMA) as a positive control for model function. The model performance was evaluated in a held-out test dataset not used for training. The study population included 1,453 adult patients (mean age 56±18 years, 42% female) with 2,556 paired echocardiography studies occurring at a median of 2 days after CMR (interquartile range 2 days prior to 6 days after). The model had high predictive capability for presence of WMA (AUC 0.873 [95%CI 0.816-0.922]) which was used for positive control. However, the model was unable to reliably detect the presence of LGE (AUC 0.699 [0.613-0.780]), abnormal native T1 (AUC 0.614 [0.500-0.715]), T2 0.553 [0.420-0.692], or ECV 0.564 [0.455-0.691]). Deep learning applied to echocardiography accurately identified CMR-based WMA, but was unable to predict tissue characteristics, suggesting that signal for these tissue characteristics may not be present within ultrasound videos, and that the use of CMR for tissue characterization remains essential within cardiology.