Automatic staging model of heart failure based on deep learning

Li, Dengao; Li, Xuemei; Zhao, Jumin; Bai, Xiao-Bo

doi:10.1016/j.bspc.2019.03.009

Cited by 35 publications

(20 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In contrast, the possibility exists of applying transfer learning techniques ( 204 ), where it is possible to deal with restricted data. Augmentation techniques suited to ECG signals might include noise addition ( 205 ), wavelet-based shrinkage filtering ( 206 ), and signal windowing or segmentation ( 206 ) with or without overlap. After signal pre-processing, if the research is focused on HRV, this should be extracted from the captured ECG signal.…”

Section: Discussionmentioning

confidence: 99%

A Proposal for a Data-Driven Approach to the Influence of Music on Heart Dynamics

Idrobo-Ávila

Correa

Muñoz-Bolaños

et al. 2021

Front. Cardiovasc. Med.

View full text Add to dashboard Cite

Electrocardiographic signals (ECG) and heart rate viability measurements (HRV) provide information in a range of specialist fields, extending to musical perception. The ECG signal records heart electrical activity, while HRV reflects the state or condition of the autonomic nervous system. HRV has been studied as a marker of diverse psychological and physical diseases including coronary heart disease, myocardial infarction, and stroke. HRV has also been used to observe the effects of medicines, the impact of exercise and the analysis of emotional responses and evaluation of effects of various quantifiable elements of sound and music on the human body. Variations in blood pressure, levels of stress or anxiety, subjective sensations and even changes in emotions constitute multiple aspects that may well-react or respond to musical stimuli. Although both ECG and HRV continue to feature extensively in research in health and perception, methodologies vary substantially. This makes it difficult to compare studies, with researchers making recommendations to improve experiment planning and the analysis and reporting of data. The present work provides a methodological framework to examine the effect of sound on ECG and HRV with the aim of associating musical structures and noise to the signals by means of artificial intelligence (AI); it first presents a way to select experimental study subjects in light of the research aims and then offers possibilities for selecting and producing suitable sound stimuli; once sounds have been selected, a guide is proposed for optimal experimental design. Finally, a framework is introduced for analysis of data and signals, based on both conventional as well as data-driven AI tools. AI is able to study big data at a single stroke, can be applied to different types of data, and is capable of generalisation and so is considered the main tool in the analysis.

show abstract

Section: Discussionmentioning

confidence: 99%

A Proposal for a Data-Driven Approach to the Influence of Music on Heart Dynamics

Idrobo-Ávila

Correa

Muñoz-Bolaños

et al. 2021

Front. Cardiovasc. Med.

View full text Add to dashboard Cite

show abstract

“…Previous work is more inclined to find the difference between the normal and CHF groups. In Table 11, researchers have evaluated the different severity levels of CHF disease, and a large number of features were used for binary and four-class classifications [27,28]. The form of the dataset is Recording × Segmentation × Time.…”

Section: Model Bmentioning

confidence: 99%

“…In that paper, they used 180 features, including 126 dynamic measurements and 54 static measurements [27]. In 2019, Li et al [28] proposed a four-stage classification problem using an end to end deep model, which extracted 20 features by convolution at max-pooling layers, and an accuracy of 97.6% was achieved.…”

Section: Introductionmentioning

confidence: 99%

“…From the previous work above, we have found that there are two points which need to be further explored: 1) They have focused on improving the accuracy of models, but the number of features in their pattern recognition models mostly exceeds 10 [23][24][25]27,28]. However, the limited number of features improves certainly the interpretability of the classification; 2) Compared with the diagnosis of the severity levels of CHF, researchers were more inclined to diagnose CHF patients from normal people.…”

Section: Introductionmentioning

confidence: 99%

“…However, the limited number of features improves certainly the interpretability of the classification; 2) Compared with the diagnosis of the severity levels of CHF, researchers were more inclined to diagnose CHF patients from normal people. For CHF risk assessment, their research was the consolidation of multiple NYHA scales into one category for identification [5,[26][27][28]. Therefore, researchers rarely distinguished the four severity scales of CHF disease clearly, which is a four-class classification problem.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Diagnosing Various Severity Levels of Congestive Heart Failure Based on Long-Term HRV Signal

et al. 2019

View full text Add to dashboard Cite

Previous studies have attempted to find autonomic differences of the cardiac system between the congestive heart failure (CHF) disease and healthy groups using a variety of algorithms of pattern recognition. By comparing previous literature, we have found that there are two shortcomings: 1) Previous studies have focused on improving the accuracy of models, but the number of features used has mostly exceeded 10, leading to poor generalization performance; 2) Previous works rarely distinguish the severity levels of CHF disease. In order to make up for these two shortcomings, we proposed two models: model A was used for distinguishing CHF patients from the normal people; model B was used for diagnosing the four severity levels of CHF disease. Based on long-term heart rate variability (HRV) (40000 intervals–8h) signals, we extracted linear and non-linear features from the inter-beat-interval (IBI) series. After that, the sequence forward selection algorithm (SFS) reduced the feature dimension. Finally, models with the best performance were selected through the leave-one-subject-out validation. For a total of 113 samples of the dataset, we applied the support vector machine classifier and five HRV features for CHF discrimination and obtained an accuracy of 97.35%. For a total of 41 samples of the dataset, we applied k-nearest-neighbor (K = 1) classifier and four HRV features for diagnosing four severity levels of CHF disease and got an accuracy of 87.80%. The contribution in this work was to use the fewer features to optimize our models by the leave-one-subject-out validation. The relatively good generalization performance of our models indicated their value in clinical application.

show abstract