Accuracy of Deep Learning Echocardiographic View Classification in Patients with Congenital or Structural Heart Disease: Importance of Specific Datasets

Abstract

Introduction: Automated echocardiography image interpretation has the potential to transform clinical practice. However, neural networks developed in general cohorts may underperform in the setting of altered cardiac anatomy. Methods: Consecutive echocardiographic studies of patients with congenital or structural heart disease (C/SHD) were used to validate an existing convolutional neural network trained on 14,035 echocardiograms for automated view classification. In addition, a new convolutional neural network for view classification was trained and tested specifically in patients with C/SHD. Results: Overall, 9793 imaging files from 262 patients with C/SHD (mean age 49 years, 60% male) and 62 normal controls (mean age 45 years, 50.0% male) were included. Congenital diagnoses included among others, tetralogy of Fallot (30), Ebstein anomaly (18) and transposition of the great arteries (TGA, 48). Assessing correct view classification based on 284,250 individual frames revealed that the non-congenital model had an overall accuracy of 48.3% for correct view classification in patients with C/SHD compared to 66.7% in patients without cardiac disease. Our newly trained convolutional network for echocardiographic view detection based on over 139,910 frames and tested on 35,614 frames from C/SHD patients achieved an accuracy of 76.1% in detecting the correct echocardiographic view. Conclusions: The current study is the first to validate view classification by neural networks in C/SHD patients. While generic models have acceptable accuracy in general cardiology patients, the quality of image classification is only modest in patients with C/SHD. In contrast, our model trained in C/SHD achieved a considerably increased accuracy in this particular cohort.

Assisted probe guidance in cardiac ultrasound: A review

“…Similarly to other cardiac imaging modalities, echocardiography studies need several viewpoints of the heart structures ( 25 ). Although an unlimited number of various views are theoretically feasible, 27 views have been recognized as the views to be obtained during a full TTE evaluation ( 26 ).…”

“…Although an unlimited number of various views are theoretically feasible, 27 views have been recognized as the views to be obtained during a full TTE evaluation ( 26 ). Additionally, sonographers purposefully focus on substructures within an image, providing a variety of different perspectives by rotating and adjusting the US probe’s zoom level ( 25 ). Despite the multitude of different cardiac views, clinicians frequently use six standard views in a routine cardiac examination to assess the structure and function of the heart: the A4C, A2C, apical three chamber (A3C), parasternal short-axis mitral valve, parasternal short-axis papillary muscle, and parasternal short-axis apex ( 27 ) ( Figure 3 ).…”

“…Since CNNs developed in general cohorts may underperform in the setting of altered cardiac anatomy, Wegner et al ( 25 ) aimed to classify 17 cardiac views in datasets depicting structural and congenital heart disease (C/SHD). For this purpose, the authors proposed a non-congenital CVC model trained on 14,035 echocardiograms, and a CNN trained and tested on 139,910 and 35,614 frames from patients with C/SHD, respectively.…”

Artificial Intelligence in Congenital Heart Disease

“… Diller et al 10 2019 132 patients with a systemic RV and 67 normal controls (73,425 TGA; 33,394 ccTGA; and 24,354 normal apical 4-chamber frames) Echocardiograms CNN—classification and segmentation 159 40 Accuracy: 0.98 Model requires external validation. Wegner et al 36 2022 9,793 echocardiogram images from 262 patients with CHD (ToF, Ebstein, TGA) and 62 controls used to build a new model. Prior model was trained on 14,035 echocardiograms from patients without CHD for automated view classification.…”

A computer vision pipeline for fully automated echocardiogram interpretation

“…The first step in interpreting echocardiographic data (step 2 in Figure 1 Echo data is sometimes incorrectly classified due to subtle differences in image properties, ambiguous to the human eye. When assessing important patient characteristics, such as regional wall motion abnormalities, crucial information from up to 7 views (apical two -four chamber, parasternal short-axis at the mitral valve, papillary muscles and apex and parasternal longaxis) should be combined to make an informed judgement of the LV [53]. Hence, accurate echocardiographic view classification is essential as some anatomical abnormalities are only detected by distinct views.…”