Beyond Heart Murmur Detection: Automatic Murmur Grading from Phonocardiogram

Objective: Murmurs are abnormal heart sounds, identified by experts through cardiac auscultation. The murmur grade, a quantitative measure of the murmur intensity, is strongly correlated with the patient's clinical condition. This work aims to estimate each patient's murmur grade (i.e., absent, soft, loud) from multiple auscultation location phonocardiograms (PCGs) of a large population of pediatric patients from a low-resource rural area. Methods: The Mel spectrogram representation of each PCG recording is given to an ensemble of 15 convolutional residual neural networks with channel-wise attention mechanisms to classify each PCG recording. The final murmur grade for each patient is derived based on the proposed decision rule and considering all estimated labels for available recordings. The proposed method is cross-validated on a dataset consisting of 3456 PCG recordings from 1007 patients using a stratified ten-fold cross-validation. Additionally, the method was tested on a hidden test set comprised of 1538 PCG recordings from 442 patients. Results: The overall cross-validation performances for patient-level murmur gradings are 86.3% and 81.6% in terms of the unweighted average of sensitivities and F1-scores, respectively. The sensitivities (and F1-scores) for absent, soft, and loud murmurs are 90.7% (93.6%), 75.8% (66.8%), and 92.3% (84.2%), respectively. On the test set, the algorithm achieves an unweighted average of sensitivities of 80.4% and an F1-score of 75.8%. Conclusions: This study provides a potential approach for algorithmic pre-screening in low-resource settings with relatively high expert screening costs. Significance: The proposed method represents a significant step beyond detection of murmurs, providing characterization of intensity which may provide a enhanced classification of clinical outcomes.

Left Ventricle Contouring of Apical Three-Chamber Views on 2D Echocardiography

We propose a new method to automatically contour the left ventricle on 2D echocardiographic images. Unlike most existing segmentation methods, which are based on predicting segmentation masks, we focus at predicting the endocardial contour and the key landmark points within this contour (basal points and apex). This provides a representation that is closer to how experts perform manual annotations and hence produce results that are physiologically more plausible. Our proposed method uses a two-headed network based on the U-Net architecture. One head predicts the 7 contour points, and the other head predicts a distance map to the contour. This approach was compared to the U-Net and to a point based approach, achieving performance gains of up to 30\% in terms of landmark localisation (<4.5mm) and distance to the ground truth contour (<3.5mm).

Contrastive Learning for View Classification of Echocardiograms

Analysis of cardiac ultrasound images is commonly performed in routine clinical practice for quantification of cardiac function. Its increasing automation frequently employs deep learning networks that are trained to predict disease or detect image features. However, such models are extremely data-hungry and training requires labelling of many thousands of images by experienced clinicians. Here we propose the use of contrastive learning to mitigate the labelling bottleneck. We train view classification models for imbalanced cardiac ultrasound datasets and show improved performance for views/classes for which minimal labelled data is available. Compared to a naive baseline model, we achieve an improvement in F1 score of up to 26% in those views while maintaining state-of-the-art performance for the views with sufficiently many labelled training observations.

The CirCor DigiScope Dataset: From Murmur Detection to Murmur Classification

Cardiac auscultation is one of the most cost-effective techniques used to detect and identify many heart conditions. Computer-assisted decision systems based on auscultation can support physicians in their decisions. Unfortunately, the application of such systems in clinical trials is still minimal since most of them only aim to detect the presence of extra or abnormal waves in the phonocardiogram signal. This is mainly due to the lack of large publicly available datasets, where a more detailed description of such abnormal waves (e.g., cardiac murmurs) exists. As a result, current machine learning algorithms are unable to classify such waves. To pave the way to more effective research on healthcare recommendation systems based on auscultation, our team has prepared the currently largest pediatric heart sound dataset. A total of 5282 recordings have been collected from the four main auscultation locations of 1568 patients, in the process 215780 heart sounds have been manually annotated. Furthermore, and for the first time, each cardiac murmur has been manually annotated by an expert annotator according to its timing, shape, pitch, grading and quality. In addition, the auscultation locations where the murmur is present were identified as well as the auscultation location where the murmur is detected more intensively.

Segmentation and Optimal Region Selection of Physiological Signals using Deep Neural Networks and Combinatorial Optimization

Physiological signals, such as the electrocardiogram and the phonocardiogram are very often corrupted by noisy sources. Usually, artificial intelligent algorithms analyze the signal regardless of its quality. On the other hand, physicians use a completely orthogonal strategy. They do not assess the entire recording, instead they search for a segment where the fundamental and abnormal waves are easily detected, and only then a prognostic is attempted. Inspired by this fact, a new algorithm that automatically selects an optimal segment for a post-processing stage, according to a criteria defined by the user is proposed. In the process, a Neural Network is used to compute the output state probability distribution for each sample. Using the aforementioned quantities, a graph is designed, whereas state transition constraints are physically imposed into the graph and a set of constraints are used to retrieve a subset of the recording that maximizes the likelihood function, proposed by the user. The developed framework is tested and validated in two applications. In both cases, the system performance is boosted significantly, e.g in heart sound segmentation, sensitivity increases 2.4% when compared to the standard approaches in the literature.