Intelligent Detection of Adventitious Sounds Critical in Diagnosing Cardiovascular and Cardiopulmonary Diseases

A Multi-Channel Stethograph System (STG system) was designed and developed as an electronic auscultation system for recording heart, lung, and trachea sounds non-invasively through an acoustic sensor array. The STG system consists of 16 acoustic sensors, a signal conditioning board, and a data logger (data acquisition, wireless transmission, sound visualization). The STG system captures breath with any adventitious sound event in 16 locations simultaneously to maximize the information of the specific sound event (for example, detection of the origin of mother adventitious sound and extracting its features), when compared with a single-channel stethoscope. This system can be an efficient tool to aid doctors or physicians in analyzing the adventitious sound from respiration diseases. However, it still requires the need for an experienced doctor or physician in the diagnosis and validation of adventitious sound. This paper presents a computerized method with an intelligent algorithm for detecting various adventitious sounds that are the key characteristics of cardiopulmonary diseases (CD) and assists the doctor/physician in the continuous diagnosis of lungs, which potentially can be beneficial during the COVID-19 progression. The proposed algorithm was able to detect breath patterns using trachea sound; location of the mother adventitious event using lung sounds (14 channels); determine the type of adventitious sound by correlating lung sound with trachea sound. The algorithm consists of breath pattern detection, candidate audio selection, breath pattern extraction, and adventitious sound detection. Digital signal processing techniques such as filtering, windowing, enveloping, discrete Fourier transform, and thresholds were used for identifying and classifying the inhalation and exhalation patterns in the lung sound in an independent (automatic) and intelligent way. The auscultation diagnosis algorithm can identify and distinguish discontinuous adventitious sounds which include wheeze, rhonchi, wheeze & rhonchi, and squawk, with an accuracy of 96.9%, 95.3%, 90%, and 100%. The algorithm was able to fully utilize the advantage of the multichannel system to simultaneously detect breath patterns, types of adventitious sound, and the location of the mother adventitious event that other algorithms cannot achieve. It has the potential to aid doctors/physicians in the early detection and monitoring of any lung disorders by providing objective evidence on adventitious sounds.