Intra-breath measurements - supporting information of “Pattern recognition in intra-breath oscillometry measurements"

This measurement dataset is part of the supporting information for our article entitled <i>“Pattern recognition in intra-breath oscillometry measurements”</i>, which can be found through the references provided in the publication.In addition to presenting the detailed methodology and results of our analysis, the article serves as a general guide for conducting machine learning (ML)-based classification using intra-breath oscillometry (IBOsc) data. It includes guidelines on signal preprocessing, feature extraction, artefact detection, aggregation strategies, and interpretable ML workflows.<b>About Intra-Breath Oscillometry (IBOsc):</b>The article introduces intra-breath oscillometry in more detail. Briefly, oscillometry is a non-invasive method for evaluating the mechanical properties of the respiratory system. The novel intra-breath variant (IBOsc) employs a <b>single-frequency test signal</b> (typically 10 Hz in adults) and computes respiratory impedance over short, overlapping time windows (approximately 0.1 seconds), capturing dynamic changes throughout the breathing cycle.<b>Dataset Structure:</b>The dataset includes <b>multiple measurement visits of 871 patients</b> (the article only included the firts visits).Each patient has a <b>dedicated folder</b>, which may contain <b>multiple measurement files</b>. The file naming structure adheres to the following convention: patietnID_measurementID_fileID.txt.inpx <br>Note: One visit may contain multiple repeated measures with different file IDs. All measurements were acquired using the <b>tremoflo® C-100</b> device (THORASYS Thoracic Medical Systems Inc., Montreal, QC, Canada) in its research mode with <b>10 Hz single-frequency excitation</b>.Subjects were seated, used a nose clip, and supported their cheeks with their hands, following the technical recommendations for routine oscillometry.Recordings lasted 20–30 seconds, capturing several full breathing cycles.The <b>pressure and flow signals were sampled at 256 Hz</b> and exported using the manufacturer’s software.<b><i>labels.csv</i></b> file contains metadata for the dataset. This file includes an index of the patients, the number of measurement files available for each patient, and their corresponding diagnostic label: <b>healthy</b>, <b>COPD</b>, or <b>ILD</b>.<b>Data Content:</b>Each column in the <i>.txt.inpx</i> data files is explained in the article:<b>Time</b> (second): Timestamps for each sample.<b>Vol</b> (liter): Volume calculated from the flow signal.<b>Pcyl </b>(cmH₂O): Measured pressure.<b>Flow</b> (L/s): Respiratory flow signal.The paper provides step-by-step instructions on how to process these signals to compute the <b>mechanical impedance of the respiratory system (Zrs)</b> and offers explanations for all signal components involved.<br>