肺音数据集，从电子听诊器声音中检测肺部疾病

该数据集包括来自一百一十二名受试者（35名健康人和77名不健康人）的呼吸音。受试者年龄从21岁到90岁，平均±SD为50.5±19.4，其中43名女性和69名男性。 每个数据文件的名称以编码为字母B、D或E的过滤器类型开头。后面跟着字母P、从1开始的唯一序列号和下划线。之后，文件名包括诊断、声音类型、胸部测量位置、受试者年龄和受试者性别。 数据中包括三种类型的过滤器。字母B与贝尔模式过滤一起使用，贝尔模式过滤放大[20-1000]Hz频率范围内的声音，但强调[20-2000]Hz范围内的低频声音。字母D与振膜模式过滤一起使用，该模式放大频率范围[20-20 0 0]Hz的声音，但强调频率范围[10 0-50 0]Hz内的声音。字母E用于扩展模式过滤，它放大频率范围为[20-1000]Hz的声音，但强调频率范围为[50-500]Hz。 胸部区域被编码为分别来自集合{A，P}，{L，R}和{L，M，U}的三个有序字母。这些字母具有以下含义；｛前：A，后：P｝，｛左：L，右：R｝，{下：L，上：U，中：M｝ 声音类型和受试者数量： 正常35，裂纹23，乳清41，裂纹8 支气管1，乳清和饼干2，支气管和饼干2 疾病诊断包括正常（N）、哮喘、COPD、BRON、心力衰竭、肺纤维化或胸腔积液。性别用字母F表示女性，用字母M表示男性。例如，名为“BP60_heart failure，Crep，P L L，83，F”的文件是从一名83岁心力衰竭女性患者的胸部左后下部区域获得的贝尔过滤的起皱声。Bell滤波器更适合听心音，心音的频率比肺部的声音低。患者编号很重要，因为它与注释文件中的疾病诊断和肺部声音类型交叉引用。 该数据集包括文件“data annotation.xlsx”，其中包含匿名人口统计信息（即年龄和性别），以及关于人体胸部特定位置的信息，记录是从那里捕获的。该文件还包含用于注释数据的各种字母符号的含义。 从听诊器导入的原始“.zsa”文件也包含在该集合中。10个文件中的每个文件都是根据其中包含的患者编号范围命名的。例如，文件“P1-P8.zsa”包含患者1至8的记录。分组是这段时间内检查的受试者数量的结果，每个文件最多可以包含12段录音。

This dataset includes respiratory sounds from 112 subjects (35 healthy individuals and 77 unhealthy individuals). The subjects range in age from 21 to 90 years, with a mean ± standard deviation of 50.5 ± 19.4, including 43 females and 69 males.
The name of each data file starts with a filter type encoded as the letters B, D, or E, followed by the letter P, a unique serial number starting from 1, and an underscore. The filename then includes the diagnosis, sound type, chest measurement location, subject age, and subject gender.
Three types of filters are included in the dataset. The letter B corresponds to the Bell-mode filter, which amplifies sounds in the [20, 1000] Hz frequency range but emphasizes low-frequency sounds within the [20, 2000] Hz range. The letter D corresponds to the diaphragm-mode filter, which amplifies sounds in the [20, 2000] Hz frequency range but emphasizes sounds within the [100, 500] Hz range. The letter E corresponds to the extended-mode filter, which amplifies sounds in the [20, 1000] Hz frequency range but emphasizes sounds within the [50, 500] Hz range.
Chest regions are encoded as three ordered letters selected from the sets {A, P}, {L, R}, and {L, M, U}, respectively. The letters have the following meanings: {anterior: A, posterior: P}, {left: L, right: R}, {lower: L, upper: U, middle: M}.
Sound types and their corresponding subject counts: normal (35 cases), crackles (23 cases), wheezes (41 cases), crackles (8 cases), bronchial breath sounds (1 case), wheezes and crackles (2 cases), bronchial breath sounds and crackles (2 cases).
Disease diagnoses include normal (N), asthma, COPD, BRON, heart failure, pulmonary fibrosis, and pleural effusion. Genders are denoted by the letter F for female and M for male.
For example, the file named "BP60_heart failure,Crep,PLL,83,F" is a crackle sound captured from the left lower posterior chest region of an 83-year-old female patient with heart failure using a Bell-mode filter. Bell-mode filters are more suitable for auscultating heart sounds, which have lower frequencies than pulmonary sounds. Patient numbers are important as they are cross-referenced with disease diagnoses and respiratory sound types in the annotation files.
The dataset includes the file "data annotation.xlsx", which contains anonymous demographic information (i.e., age and gender), information about specific chest locations where recordings were captured, and the meanings of various alphabetic symbols used for data annotation.
Original ".zsa" files imported from a stethoscope are also included in the dataset. Each of the 10 files is named based on the range of patient numbers it contains. For example, the file "P1-P8.zsa" contains recordings from patients 1 to 8. Grouping is based on the number of subjects examined during this period, and each file can hold up to 12 recording segments.

搜集汇总

背景与挑战

背景概述

该数据集包含112名受试者的呼吸音记录，用于从电子听诊器声音中检测肺部疾病，涵盖健康和不健康人群，年龄范围广泛（21-90岁），性别分布均衡。数据文件采用结构化命名，包含过滤器类型、疾病诊断、声音类型、胸部位置等关键信息，支持多种肺部疾病（如哮喘、COPD）和声音类型（如裂纹、乳清）的分析。数据集还提供注释文件和原始录音，便于研究和交叉引用，适用于肺部疾病检测和声音分类任务。