Scaling to Multimodal and Multichannel Heart Sound Classification: Fine-Tuning Wav2Vec 2.0 with Synthetic and Augmented Biosignals
Abstract
Cardiovascular diseases (CVDs) are the leading cause of death worldwide, accounting for approximately 17.9 million deaths each year. Early detection is critical, creating a demand for accurate and inexpensive pre-screening methods. Deep learning has recently been applied to classify abnormal heart sounds indicative of CVDs using synchronised phonocardiogram (PCG) and electrocardiogram (ECG) signals, as well as multichannel PCG (mPCG). However, state-of-the-art architectures remain underutilised due to the limited availability of synchronised and multichannel datasets. Augmented datasets and pre-trained models provide a pathway to overcome these limitations, enabling transformer-based architectures to be trained effectively. This work combines traditional signal processing with denoising diffusion models, WaveGrad and DiffWave, to create an augmented dataset to fine-tune a Wav2Vec 2.0-based classifier on multimodal and multichannel heart sound datasets. The approach achieves state-of-the-art performance. On the Computing in Cardiology (CinC) 2016 dataset of single channel PCG, accuracy, unweighted average recall (UAR), sensitivity, specificity and Matthew's correlation coefficient (MCC) reach 92.48\%, 93.05\%, 93.63\%, 92.48\%, 94.93\% and 0.8283, respectively. Using the synchronised PCG and ECG signals of the training-a dataset from CinC, 93.14\%, 92.21\%, 94.35\%, 90.10\%, 95.12\% and 0.8380 are achieved for accuracy, UAR, sensitivity, specificity and MCC, respectively. Using a wearable vest dataset consisting of mPCG data, the model achieves 77.13\% accuracy, 74.25\% UAR, 86.47\% sensitivity, 62.04\% specificity, and 0.5082 MCC. These results demonstrate the effectiveness of transformer-based models for CVD detection when supported by augmented datasets, highlighting their potential to advance multimodal and multichannel heart sound classification.
The following analysis is generated by AI and may contain errors. It is separate from the authors’ abstract and is not peer review.
Research summary
This paper effectively combines advanced machine learning techniques with practical applications in cardiovascular health, showcasing a novel approach to heart sound classification that could significantly impact clinical practices. The integration of synthetic data generation methods addresses critical data limitations, making it a valuable contribution to the field of medical machine learning.
Methodology
The paper presents a comprehensive methodology that integrates traditional signal processing with advanced deep learning techniques, specifically leveraging Wav2Vec 2.0 and diffusion models (WaveGrad and DiffWave) for synthetic and augmented biosignal generation. The approach is innovative in that it addresses the scarcity of high-quality, synchronized datasets for heart sound classification by generating synthetic data, which is a significant contribution to the field. The methodology is well-structured, with clear steps for data augmentation, model training, and evaluation, although some details regarding hyperparameter tuning and model architecture could be elaborated further for clarity.
Results & evaluation
The experimental section is robust, utilizing multiple datasets (CinC 2016 and a wearable vest dataset) to validate the proposed models. The results demonstrate state-of-the-art performance on the CinC dataset and near-state-of-the-art results on the multichannel vest dataset. The metrics used for evaluation (accuracy, UAR, sensitivity, specificity, and MCC) are appropriate for the classification task, and the use of cross-validation enhances the reliability of the findings. However, the paper could benefit from a more detailed comparison with existing methods beyond just accuracy metrics to provide a clearer context of its contributions.
Reproducibility
The paper provides a reasonable level of detail regarding the implementation, including the hardware used and the data preprocessing steps. However, it lacks a direct link to a code repository or demo, which would greatly enhance reproducibility. Additionally, while the hyperparameter optimization process is mentioned, specific values and configurations used in the experiments are not fully disclosed, which could hinder replication efforts by other researchers.
Limitations
One limitation of the study is the reliance on synthetic data, which may not fully capture the complexities of real-world heart sounds. The performance on the multichannel vest dataset is lower than on the CinC dataset, indicating challenges in generalizing the model to noisier, real-world data. Furthermore, the paper does not address potential biases in the datasets used, which could affect the model's applicability across diverse populations.
Broader impact
The implications of this research are significant, particularly in the context of early detection of cardiovascular diseases, which is a leading cause of mortality worldwide. The ability to classify heart sounds accurately and inexpensively could enhance pre-screening methods and facilitate better patient outcomes. The integration of multimodal data (PCG and ECG) also opens avenues for more comprehensive diagnostic tools in cardiology. This paper effectively combines advanced machine learning techniques with practical applications in cardiovascular health, showcasing a novel approach to heart sound classification that could significantly impact clinical practices. The integration of synthetic data generation methods addresses critical data limitations, making it a valuable contribution to the field of medical machine learning.
Author of this paper? Send a correction.