Searched over 200M research papers for "deep cough"
10 papers analyzed
These studies suggest that deep learning models, particularly convolutional neural networks, outperform traditional methods in identifying and classifying cough sounds with high sensitivity and specificity.
20 papers analyzed
Deep neural networks (DNNs) have shown significant promise in identifying cough sounds. Two primary approaches have been explored: convolutional neural networks (CNNs) and recurrent neural networks (RNNs). CNNs treat cough detection as a visual recognition problem, while RNNs address it as a sequence-to-sequence labeling problem. Experimental results indicate that both architectures outperform traditional methods, with CNNs achieving a higher specificity of 92.7% and RNNs attaining a higher sensitivity of 87.7%.
Another study introduced a pretrained DNN for cough classification, which involves a two-step process of pretraining and fine-tuning, followed by a Hidden Markov Model (HMM) decoder. This method demonstrated superior performance over traditional Gaussian Mixture Model-Hidden Markov Model (GMM-HMM) frameworks, achieving significant error reduction in sensitivity and specificity across both patient-dependent and patient-independent settings.
A wearable cough detection system employing a deep convolutional neural network was evaluated on healthy volunteers. This system achieved a classification sensitivity of 95.1% and a specificity of 99.5%, outperforming other reported cough detection systems.
Deep inspiration-provoked cough (DIPC) is associated with cough reflex arc hypersensitivity. In a study involving chronic cough patients and healthy subjects, DIPC was significantly more prevalent among those with chronic cough. The number of DIPC events correlated with the concentration of citric acid required to provoke coughs, indicating a hypersensitive cough reflex arc in these patients.
The DeepCough system, which uses deep learning algorithms to analyze cough sounds, has been developed as a primary screening tool for COVID-19. This system, based on clinically validated samples, achieved high accuracy, sensitivity, and specificity in detecting COVID-19. The results suggest that such a tool could significantly aid in the rapid detection and management of COVID-19.
Cold therapy has been shown to effectively manage pain associated with deep breathing and coughing post-cardiac surgery. Studies demonstrated that applying a cold gel pack to the chest incision area significantly reduced pain during these exercises, suggesting that cold therapy can be a valuable tool in postoperative care .
Deep learning models, particularly those utilizing transformer-based algorithms like BERT, have been effective in identifying chronic cough patients from electronic health records (EHRs). These models achieved high sensitivity and specificity by combining structured data (medication and diagnosis) with unstructured data (clinical notes), outperforming traditional rule-based algorithms.
Recent advancements in deep learning and neural networks have significantly improved the detection and classification of cough sounds, offering promising tools for diagnosing respiratory conditions and managing postoperative pain. These technologies not only enhance the accuracy and efficiency of cough detection but also provide valuable insights into the underlying mechanisms of cough reflex hypersensitivity and chronic cough identification.
Most relevant research papers on this topic