International Joint Conference on Neural Networks

Continuous monitoring of vital signs, such as respiration and heartbeat, plays a crucial role in early detection and even prediction of conditions that may affect the wellbeing of the patient. Sensing vital signs can be categorized into: contact-based techniques and contactless based techniques. Conventional clinical methods of detecting these vital signs require the use of contact sensors, which may not be practical for long duration monitoring and less convenient for repeatable measurements. On the other hand, wireless vital signs detection using radars has the distinct advantage of not requiring the attachment of electrodes to the subject’s body and hence not constraining the movement of the person and eliminating the possibility of skin irritation. In addition, it removes the need for wires and limitation of access to patients, especially for children and the elderly. This paper presents a thorough review on the traditional methods of monitoring cardio-pulmonary rates as well as the potential of replacing these systems with radar-based techniques. The paper also highlights the challenges that radar-based vital signs monitoring methods need to overcome to gain acceptance in the healthcare field. A proof-of-concept of a radar-based vital sign detection system is presented together with promising measurement results.

https://www.mdpi.com/1424-8220/20/5/1454/pdf

Human Vital Signs Detection Methods and Potential Using Radars: A Review.

The alarmingly high mortality rate and increasing global prevalence of cardiovascular diseases (CVDs) signify the crucial need for early detection schemes. Phonocardiogram (PCG) signals have been historically applied in this domain owing to its simplicity and cost-effectiveness. In this article, we propose CardioXNet, a novel lightweight end-to-end CRNN architecture for automatic detection of five classes of cardiac auscultation namely normal, aortic stenosis, mitral stenosis, mitral regurgitation and mitral valve prolapse using raw PCG signal. The process has been automated by the involvement of two learning phases namely, representation learning and sequence residual learning. Three parallel CNN pathways have been implemented in the representation learning phase to learn the coarse and fine-grained features from the PCG and to explore the salient features from variable receptive fields involving 2D-CNN based squeeze-expansion. Thus, in the representation learning phase, the network extracts efficient time-invariant features and converges with great rapidity. In the sequential residual learning phase, because of the bidirectional-LSTMs and the skip connection, the network can proficiently extract temporal features without performing any feature extraction on the signal. The obtained results demonstrate that the proposed end-to-end architecture yields outstanding performance in all the evaluation metrics compared to the previous state-of-the-art methods with up to 99.60% accuracy, 99.56% precision, 99.52% recall and 99.68% F1- score on an average while being computationally comparable. This model outperforms any previous works using the same database by a considerable margin. Moreover, the proposed model was tested on PhysioNet/CinC 2016 challenge dataset achieving an accuracy of 86.57%. Finally the model was evaluated on a merged dataset of Github PCG dataset and PhysioNet dataset achieving excellent accuracy of 88.09%. The high accuracy metrics on both primary and secondary dataset combined with a significantly low number of parameters and end-to-end prediction approach makes the proposed network especially suitable for point of care CVD screening in low resource setups using memory constraint mobile devices.

CardioXNet: A Novel Lightweight Deep Learning Framework for Cardiovascular Disease Classification Using Heart Sound Recordings

Cardiovascular disease is one of the leading factors for death cause of human beings. In the past decade, heart sound classification has been increasingly studied for its feasibility to develop a non-invasive approach to monitor a subject’s health status. Particularly, relevant studies have benefited from the fast development of wearable devices and machine learning techniques. Nevertheless, finding and designing efficient acoustic properties from heart sounds is an expensive and time-consuming task. It is known that transfer learning methods can help extract higher representations automatically from the heart sounds without any human domain knowledge. However, most existing studies are based on models pre-trained on images, which may not fully represent the characteristics inherited from audio. To this end, we propose a novel transfer learning model pre-trained on large scale audio data for a heart sound classification task. In this study, the PhysioNet CinC Challenge Dataset is used for evaluation. Experimental results demonstrate that, our proposed pre-trained audio models can outperform other popular models pre-trained by images by achieving the highest unweighted average recall at 89.7 %.

Audio for Audio is Better? An Investigation on Transfer Learning Models for Heart Sound Classification

By classifying the heart sound signals, it can provide very favorable clinical information to the diagnosis of cardiovascular diseases. According to the characteristics of heart sound signals which are complex and difficult to classify and recognize, a new method of feature extraction and classification about heart sound signal is proposed by a combination of wavelet scattering transform and twin support vector machine in this paper. The method is as follows: The heart sound signal data set is firstly divided into two parts, one as a training set and the other as a testing set. Then the wavelet scattering transform is applied to the heart sound signals in the training set and the testing set. The scattering transform is a new time-frequency analysis method. It overcomes the shortcomings of the traditional wavelet transform which has the time-shift changes. It has the advantages of translation invariance and elastic deformation stability. Thus obtain the scattering feature matrix of the heart sound signal. Due to the large dimension of scattering feature matrix, this paper uses multidimensional scaling (MDS) method to reduce the dimension. This method is compared with the classical dimension reduction method-principal component analysis (PCA). Finally, the dimensionality-reduced feature matrix is input into the twin support vector machine (TWSVM) for training. After training the classifier to get the optimal parameters, the dimensionality-reduced scattering feature matrix of the testing signal is input into the classifier for testing. Experimental results show that the classification accuracy of the proposed method can reach 98% or more, and the running time is greatly reduced compared with support vector machine (SVM).

/pdf/heart-sound-signal-classification-algorithm-a-combination-of-4jfwhcypgl.pdf

Heart Sound Signal Classification Algorithm: A Combination of Wavelet Scattering Transform and Twin Support Vector Machine

Phonocardiogram (PCG) signal represents recording of sounds and murmurs resulting from heart auscultation. Analysis of these PCG signals is critical in diagnosis of different heart diseases. Over the years, a variety of methods have been proposed for automatic analysis of PCG signals in time, frequency, and time-frequency domains. This paper presents a comprehensive survey of different methods proposed for automatic analysis of PCG signals with the objective to evaluate the current state-of-the-art and to determine the potential domains of effective analysis. An important aspect of our contribution is that the review is carried out as a function of domains of analysis rather than simply discussing different methods. Our method further splits analysis into pre-processing, localization, and classification, and details are presented in terms of techniques and classifiers used during these phases. Finally, results are summarized for normal heart beat, noisy heart beat, and different pathologies using metrices like accuracy and detection rate. In addition to time and frequency domain, time-frequency based methods including wavelet, empirical mode decomposition (EMD) and time-frequency representation (TFR) are selected for detailed analysis. The review concludes that the time-frequency representations like EMD and wavelets represent areas of exploration in future along with perspective of using different time-frequency techniques together.

/pdf/localization-and-classification-of-heart-beats-in-1k0sm5i8cu.pdf

Localization and classification of heart beats in phonocardiography signals —a comprehensive review

Non-invasive photoplethysmography (PPG) technology was developed to track heart rate during motion. Automated analysis of PPG has made it useful in both clinical and non-clinical applications. However, PPG-based heart rate tracking is a challenging problem due to motion artifacts (MAs) which are main contributors towards signal degradation as they mask the location of heart rate peak in the spectra. A practical analysis system must have good performance in MA removal as well as in tracking. In this article, we have presented state-of-art techniques in both areas of the automated analysis, i.e., MA removal and heart rate tracking, and have concluded that adaptive filtering and multi-resolution decomposition techniques are better for MA removal and machine learning-based approaches are future perspective of heart rate tracking. Hence, future systems will be composed of machine learning-based trackers fed with either empirically decomposed signal or from output of adaptive filter.

/pdf/heart-rate-tracking-in-photoplethysmography-signals-affected-534u7ij9sr.pdf

Shahid Ismail

Papers

Localization and classification of heart beats in phonocardiography signals —a comprehensive review

Heart rate tracking in photoplethysmography signals affected by motion artifacts: a review

Heart rate estimation in PPG signals using Convolutional-Recurrent Regressor