Scholarlink

<p>The global expansion of infectious lung diseases, like the COVID-19 pandemic, brought attention to the growing role of medical professionals in the adoption of automated lung sound diagnosis. Lung sounds carry essential information pertaining to pulmonary pathology. This work utilizes the ICBHI 2017 dataset, a publicly available repository of lung sound recordings encompassing normal, crackle, wheeze, and crackle embedded wheeze. A 12th order Butterworth filter is employed to remove noise and moving artifacts present in raw lung sounds. Fourteen features are extracted in time domain, ten, each in frequency and time–frequency domain from the four class of pre-processed lung sound signals. This work attempts to identify significant features from exhaustive feature set to demarcate lung sounds, as normal, crackles, wheezes, and the critical class, crackle-embedded wheezes. Due to its versatility, the t-statistic is commonly used to identify prominent features of different groups. Hence, t-test is applied on 34 features extracted in all three domain and significant features are selected based on obtained <i>p</i>-values. Statistical analysis of significant features is carried out using boxplot aside normal distribution plot. The significant features selected will foster the automated categorization of adventitious lung sounds.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An attribute-based framework to categorize adventitious lung sounds fostering automated diagnostic procedure

  • 期刊论文

  • 发表日期：

    2026年6月2日

查看全文
  • G. Shanthakumari,
  • E. Priya

摘要

The global expansion of infectious lung diseases, like the COVID-19 pandemic, brought attention to the growing role of medical professionals in the adoption of automated lung sound diagnosis. Lung sounds carry essential information pertaining to pulmonary pathology. This work utilizes the ICBHI 2017 dataset, a publicly available repository of lung sound recordings encompassing normal, crackle, wheeze, and crackle embedded wheeze. A 12th order Butterworth filter is employed to remove noise and moving artifacts present in raw lung sounds. Fourteen features are extracted in time domain, ten, each in frequency and time–frequency domain from the four class of pre-processed lung sound signals. This work attempts to identify significant features from exhaustive feature set to demarcate lung sounds, as normal, crackles, wheezes, and the critical class, crackle-embedded wheezes. Due to its versatility, the t-statistic is commonly used to identify prominent features of different groups. Hence, t-test is applied on 34 features extracted in all three domain and significant features are selected based on obtained p-values. Statistical analysis of significant features is carried out using boxplot aside normal distribution plot. The significant features selected will foster the automated categorization of adventitious lung sounds.