Auscultation of lung sounds in high noise environments using active noise reduction techniques

Samir Bhanubhai Patel, Purdue University

Abstract

Auscultation of lung sounds in an ambulance or aircraft (helicopter or fixed wing) used for medical evacuation is unachievable because of high ambient noise levels. For example, aircraft noise levels of 90 to 100 dB(A) are common while lung sounds have been measured in the 22 to 30 dB(A) range in free space and 65 to 70 dB(A) within a stethoscope coupler. Also, the frequency ranges of both lung sounds and aircraft noise have a strong overlap, limiting the utility of band-pass filtering. This clinical problem motivates the employment of dedicated signal processing techniques to extract high-fidelity lung sounds from the background noise. This thesis work addresses the digital signal processing and in particular, the adaptive filtering issues associated with an active noise reduction stethoscope that is under development. Noise corrupted lung sounds were collected from a healthy human subject simultaneously with the (reference) noise using a two microphone arrangement in a simulated C-130 aircraft environment in an acoustic chamber at noise levels ranging from 80 to 100 dB(A). The two microphones were closely spaced yet acoustically isolated within a passively shielded stethoscope coupler. Three adaptive filtering algorithms were employed after data collection: least mean square (LMS), normalized LMS (NLMS), and equation-error based adaptive infinite impulse response (IIREE). With the computationally simplest LMS approach, up to 25 dB of noise reduction in 100-600 Hz frequency range was observed. Although the increase in computational complexity of the NLMS approach as compared to the LMS is minor, an additional 5 dB or more of noise reduction was found. No significant performance increase was attained using the IIREE algorithm, despite its significantly higher computational complexity as compared to the NLMS approach. These results indicate that an active noise reduction stethoscope based on a digital adaptive filter and a passively isolated coupler with two microphones can be used to extract audible lung sounds from those detected in the presence of background aircraft noise, with the NLMS algorithm the best choice of the three considered for C-130 aircraft conditions.

Degree

Ph.D.

Advisors

Wodicka, Purdue University.

Subject Area

Electrical engineering|Biomedical research

COinS