Department of Electronics Engineering, University of Engineering and Technology Taxila, Taxila, Pakistan.
Department of Electrical Engineering, University of Engineering and Technology Taxila, Taxila, Pakistan.
Biomed Tech (Berl). 2022 Apr 11;67(3):201-218. doi: 10.1515/bmt-2022-0011. Print 2022 Jun 27.
Chronic obstructive pulmonary disease (COPD) and pneumonia are lethal pulmonary illnesses with equivocal nature of abnormal pulmonic acoustics. Using lung sound signals, the classification of pulmonary abnormalities is a difficult task. A standalone system was conceived for screening COPD and Pneumonia patients through signal processing and machine learning methodologies. The proposed system will assist practitioners and pulmonologists in the accurate classification of disease. In this research work, ICBHI's and self-collected lung sound (LS) databases are used to investigate COPD and pneumonia patient. In this scheme, empirical mode decomposition (EMD), discrete wavelet transform (DWT), and analysis of variance (ANOVA) techniques are employed for segmentation, noise elimination, and feature selection, respectively. To overcome the inherent limitation of ICBHI's LS database, the adaptive synthetic (ADASYN) sampling technique is used to eradicate class imbalance. Lung sound features are used to train fine Gaussian support vector machine (FG-SVM) for classification of COPD, pneumonia, and heathy healthy subjects. This machine learning scheme is implemented on low cost and portable Raspberry pi 3 model B+ (Cortex-A53 (ARMv8) 64-bit SoC @ 1.4 GHz through hardware-supported language. Resultant hardware is capable of screening COPD and pneumonia patients accurately and assist health professionals.
慢性阻塞性肺疾病（COPD）和肺炎是致命的肺部疾病，其肺部声学异常性质不明确。使用肺部声音信号对肺部异常进行分类是一项艰巨的任务。通过信号处理和机器学习方法，设计了一个独立的系统来筛选 COPD 和肺炎患者。该系统将帮助医生和肺病学家对疾病进行准确分类。在这项研究工作中，使用了 ICBHI 的和自采集的肺部声音（LS）数据库来研究 COPD 和肺炎患者。在该方案中，分别使用经验模态分解（EMD）、离散小波变换（DWT）和方差分析（ANOVA）技术进行分割、噪声消除和特征选择。为了克服 ICBHI 的 LS 数据库的固有局限性，使用自适应合成（ADASYN）采样技术消除类别不平衡。使用肺部声音特征来训练精细高斯支持向量机（FG-SVM），以对 COPD、肺炎和健康受试者进行分类。该机器学习方案在低成本和便携式 Raspberry pi 3 模型 B+（Cortex-A53（ARMv8）64 位 SoC@1.4GHz）上通过硬件支持的语言实现。由此产生的硬件能够准确地筛选 COPD 和肺炎患者，并为医疗保健专业人员提供帮助。
