A DATASET AT A GLANCE...

Acute respiratory ailments have become more common now. The chemically-polluted air from industries, the emergence of new diseases and the continuous treatment for simple respiratory illnesses during childhood can all be indicated as the causes. “Billions of people still breathe unhealthy air”, states the World Health Organization (WHO) report, as on April 2022. So, in the forthcoming years, it will not be a surprise to see a ruthless rise in lung-based diseases affecting children, adults or elder people without partiality. So, is there any possibility to detect and diagnose the lung diseases, prior to becoming fatal? Yes. Modern tech does support by providing automated approaches for analyzing and classifying the respiratory sounds into a specific lung disorder. But how? It is because the changes created by diseases on lung secretions or tissue linings give varying sounds like, crackles or wheezes, when air moves through it. Examining change of respiratory sounds at different time instances could effectively help in classifying the occurrence of asthma, Chronic Obstructive Pulmonary Disease (COPD) or any other lung-related ailments. Kaggle repository presents a publicly available respiratory sound database for developing automated lung disease classification approaches and it could be found in https://www.kaggle.com/datasets/vbookshelf/respiratory-sound-database. Holding the respiratory sound recordings of 126 patients, which were recorded using digital stethoscopes and other recording techniques, the dataset has 920 annotated recordings of changing length between - 10s and 90s. About 5.5 hours of recordings with 6898 respiratory cycles (1864 contain crackles, 886 contain wheezes and 506 contain both crackles and wheezes) has been included in the dataset. For maintaining a practical scenario, the dataset has both respiratory sounds without any crackles or wheezes and noisy ones. Researchers intending to serve in the medical field, especially the lung-related ailments, can make use of the respiratory sound database. However, to avail the dataset, a researcher is required to do two things: (1) Cite “Rocha BM, Filos D, Mendes L, Vogiatzis I, Perantoni E, Kaimakamis E, Natsiavas P, Oliveira A, Jácome C, Marques A and Paiva RP, “Α Respiratory Sound Database for the Development of Automated Classification”, In Precision Medicine Powered by pHealth and Connected Health, Springer, Singapore, 2018, pp. 51-55”. (2) Mention special acknowledgements to the research teams at the University of Coimbra as well as the University de Aveiro in Portugal and the Aristotle University of Thessaloniki in Greece.

Image Credit: www.vecteezy.com