Deep learning-based lung sound analysis for intelligent stethoscope

Auscultation is crucial for the diagnosis of respiratory system diseases.However,traditional stethoscopes have inherent limitations,such as inter-listener variability and subjectivity,and they cannot record respiratory sounds for offline/retrospective diagnosis or remote prescriptions in telemedicine.The emergence of digital stethoscopes has overcome these limitations by allowing physicians to store and share respiratory sounds for consultation and education.On this basis,machine learning,particularly deep learning,enables the fully-automatic analysis of lung sounds that may pave the way for intelligent stethoscopes.This review thus aims to provide a comprehensive overview of deep learning algorithms used for lung sound analysis to emphasize the significance of artificial intelligence(AI)in this field.We focus on each component of deep learning-based lung sound analysis systems,including the task categories,public datasets,denoising methods,and,most importantly,existing deep learning methods,i.e.,the state-of-the-art approaches to convert lung sounds into two-dimensional(2D)spectrograms and use convolutional neural networks for the end-to-end recognition of respiratory diseases or abnormal lung sounds.Additionally,this review highlights current challenges in this field,including the variety of devices,noise sensitivity,and poor interpretability of deep models.To address the poor reproducibility and variety of deep learning in this field,this review also provides a scalable and flexible open-source framework that aims to standardize the algorithmic workflow and provide a solid basis for replication and future extension:https://github.com/contactless-healthcare/Deep-Learning-for-Lung-Sound-Analysis.

Is sibship composition a risk factor for childhood asthma? Systematic review and meta‑analysis

Background Following the“hygiene hypothesis”,the role of sibship composition in asthma and wheezing has been extensively studied,but the fndings are inconsistent.For the frst time,this systematic review and meta-analysis synthesized evidences from studies investigating the association of sibship size and birth order with risk of asthma and wheezing.Methods Fifteen databases were searched to identify eligible studies.Study selection and data extraction were performed independently by pairs of reviewers.Meta-analysis with robust variance estimation(RVE)was used to produce pooled risk ratio(RR)efect estimates from comparable numerical data.Results From 17,466 identifed records,158 reports of 134 studies(>3 million subjects)were included.Any wheezing in the last≤1.5 years occurred more frequently in infants with≥1 sibling[pooled RR 1.10,95%confdence interval(CI)1.02–1.19]and≥1 older sibling(pooled RR 1.16,95%CI 1.04–1.29).The pooled efect sizes for asthma were overall statistically nonsignifcant,although having≥1 older sibling was marginally protective for subjects aged≥6 years(pooled RR 0.93,95%CI 0.88–0.99).The efect estimates weakened in studies published after 2000 compared with earlier studies.Conclusions Being second-born or later and having at least one sibling is associated with a slightly increased risk of temporary wheezing in infancy.In contrast,being second-born or later is associated with marginal protection against asthma.These associations appear to have weakened since the turn of the millennium,possibly due to lifestyle changes and socioeconomic development.

Detection of abnormalities in dyspneic patients using a new lung imaging modality

Background Although chest radiography is a useful examination tool,it has limitations.Because not all chest conditions can be detected on a radiograph,radiography cannot necessarily rule out all irregularities in the chest.Therefore,further imaging studies may be required to clarify the results of a chest radiograph,or to identify abnormalities that are not readily visible.The aim of this study was to compare traditional chest radiography with acoustic-based imaging （vibration response imaging） for the detection of lung abnormalities in patients with acute dyspnea.Methods The current investigation was a pilot study.Respiratory sounds throughout the respiratory cycle were captured using an acoustic-based imaging technique.Consecutive patients who presented to the emergency department with acute dyspnea and a normal chest radiograph on admission were enrolled and underwent imaging at the time of presentation.Dynamic and static images of vibration （breath sounds） and a dynamic image score were generated,and assessments were made using an evaluation form.Results In healthy volunteer controls （n=61）,the mean dynamic image score was 6.3±1.9.In dyspneic patients with normal chest radiographs （n=51） and abnormal chest radiographs （n=48）,the dynamic image scores were 4.7±2.7 and 5.1±2.5,respectively （P ＜0.05）.The final assessment of the vibration images indicated abnormal findings in 15％,86％ and 90％ of the participants in the above groups,respectively （P ＜0.05）.Conclusions In patients with acute dyspnea who present with normal chest radiographs,respiratory sound analyses often showed abnormal values.Hence,the ability of acoustic-based recordings to offer objective and noninvasive measurements of abnormal sound transmission may be useful in the clinical setting for patients presenting with acute dyspnea.