Study design and patients
The protocol was approved by the Marburg Ethics Committees, the institutional review board for human studies, before study initiation (file number study 41/19). The study was performed in accordance with ethical standards the Declaration of Helsinki (October 2013) and written informed consent was obtained from all subjects.
Twenty- five patients with interstitial lung disease (ILD) were included in the study between July end the end of August 2019. The data collection took place at the Schön Klinik in Berchtesgadener Land (Germany), at the Centre for Pneumological Rehabilitation. All patients underwent inpatient treatment for pneumological rehabilitation at the time of acoustic registration. Exclusion criteria were an acute infection, which was confirmed clinically and/or haematologically.
Long-term registration of crackles and coughs
The LEOSound® device (Löwenstein Medical GmbH, Bad Ems, Germany) was used to record and evaluate respiratory as well as ambient sounds. The portable recorder obtained data from four microphones, three body microphones for respiratory sounds and the fourth microphone detecting ambient sounds integrated within the device. Two bioacoustic sensors were fixed over the basal lung segments, an additional sensor placed paratracheally (see Fig. 1).
The recordings were performed overnight to reduce artefacts and to ensure high quality. The registrations took place in the patients’ hospital room over a period of about 8 hours (22.00h to 06.00h) considering the individual sleeping time.
LEOSound analyser
The overnight recorded sounds were transferred to the associated software of the LEOSound analyser. A validated automatic evaluation of wheezing, coughing and respiratory rate was performed. The description of the recorder, sensitivity and specificity of the data analysis were described in detail in previous publications (21–23). An audio-visual verification of crackles, coughs and wheezes in 30s epochs complemented the software-based analysis (24).
LSARate software
The LSARate software was used for further evaluation of normal and adventitious respiratory sounds. The programme was developed as a supplement to the LEOSound analyser in order to carry out in-depth sound evaluations that were not automatically analysed. The in-depth scoring used qualifying recording ranges defined by the LEOSound analyser. The evaluation was done audio-visually for each respiratory cycle of the evaluable epochs. The data were evaluated by one trained person to avoid inter-individual differences. Cough and wheezing events were scored, and crackles were divided into inspiratory and expiratory crackles.
Diagnostic procedures
Lung function and diffusing capacity of the lungs for carbon monoxide (DLCO)
A body plethysmography was performed during the hospital stay according to international recommendations of the ATS (American Thoracic Society) and the ERS (European Respiratory Society) (25, 26). Furthermore, the DLCO was determined in patients with sufficiently high inspiratory vital capacity.
GAP Score
The GAP (Gender, Age, Physiology) score was used to describe the severity of interstitial lung disease. The cumulative score was composed of four parameters: gender, age, and the percentages of forced vital capacity (FVC) and DLCO (27).
Questionnaires
King’s Brief Interstitial Lung Disease (K-BILD) questionnaire
K-BILD was used to assess the subjective severity of breathlessness (27). K-BILD is a questionnaire specifically designed and validated for patients with ILD to assess health-related quality of life. The higher the score, the better the perceived health-related quality of life (28).
Statistical analysis
The statistical analysis was carried out with the programme IBM SPSS Statistics Version 24 (IBM, Ehningen, Germany). The Kolmogorov-Smirnov test was used for normal distribution of small sample sizes. The test showed deviations from normal distribution for the majority of the variables. The Mann-Whitney U test was used for group comparisons (IPF vs. ILD; GAP 2 vs. GAP 3). The Spearman-Rho correlation coefficient was calculated for correlation analyses. Statistical significance was assumed at p < 0.05 (two-sided).