A Robust Automatic Algorithm for Statistical Analysis and Classification of Lung Auscultations

Respiratory diseases affect more than 200 million people across the world and are one of the most intrinsic contributors towards deaths of adults and infants alike. Lung disorders range from mild symptoms like common cold and influenza, to life threatening instances like Pneumonia, Asthma and Lung Cancer. Therefore, early diagnosis of a respiratory disorder can often help prevent a tragedy. Medical Diagnostic of a lung disorder generally requires an auscultation of lung sounds, brief chest x-ray and in some cases can even include bronchoscopy, chest imaging and thoracoscopy. Auscultation is often subject to various biased opinions by different physicians and the results can be catastrophic if the physician is untrained. This research paper proposes statistical analysis and classification of the various auscultations of lung sounds. Here, the breathing rate of a person is chosen as the core parameter to segment the total number of breaths into mild, soft and hard breaths. In addition to this, the peak value of the envelope of the normalized signal is successfully used to predict the odds of having a lung disorder, from among Crackle, Pneumonia, Wheeze and Asthma. The proposed system reduces the need of a trained pra ctitioner which in turn makes the lung disorder diagnosis cost effective and also pro vides unbiased predictions. The time complexity of the system is very low which makes it suitable for the real time diagnosis of various lung disorders. The lung sounds are taken from the R.A.L.E, Canada repository.