An open auscultation dataset for machine learning-based respiratory diagnosis studies.

Area of Science:

• Medical Informatics
• Biomedical Engineering
• Artificial Intelligence in Healthcare

Background:

• Machine learning holds promise for auscultation-based diagnosis, particularly in prescreening applications.
• A significant limitation is the scarcity of high-quality, annotated datasets for training these models.
• Developing robust machine learning diagnostic tools requires comprehensive and accessible data.

Purpose of the Study:

• To establish a novel, open-access auscultation dataset for respiratory diagnosis research.
• To facilitate the development and validation of machine learning algorithms for lung sound classification.
• To address the critical need for high-quality training data in machine learning-enabled auscultation.

Main Methods:

• Compilation of an auscultation dataset including samples and annotations from both patients and healthy individuals.
• Development and application of a machine learning approach for analyzing the dataset.
• Classification of lung sounds to identify different respiratory diseases.

Main Results:

• Establishment of a publicly available open auscultation dataset.
• Demonstration of a machine learning approach for lung sound classification using the new dataset.
• The dataset supports scientific investigation and practical application in respiratory diagnosis.

Conclusions:

• The newly established open auscultation dataset is valuable for machine learning-based respiratory diagnosis.
• This resource can accelerate research and development in automated lung sound analysis.
• The dataset has significant scientific importance and practical potential for improving diagnostic capabilities.