Influence of Gaussian White Noise on Medical Students' Capacity to Accurately Identify Pulmonary Sounds

Abstract

Background: The effect of background noise on auscultation accuracy for different lung sound classes under standardised conditions, especially at lower to medium levels, remains largely unexplored. This article aims to evaluate the impact of three levels of Gaussian white noise (GWN) on the ability to identify three classes of lung sounds.

Methods and materials: A pre-post pilot study assessing the impact of GWN on a group of students' ability to identify lung sounds was conducted. The three intensities were applied to the three classes of lung sounds: no GWN, signal-to-noise ratio (SNR), SNR-40 (medium level) and SNR-20 (high). This resulted with three exams, each containing nine questions. Fifty-two participants underwent a 4-day training programme and were tested on their identification of lung sound classes under the three levels of GWN, but seven subjects were excluded for not completing all three assessments. Statistical analysis was performed on 45 subjects, using non-parametric tests to analyse the data. A P-value of 0.05 was considered statistically significant.

Results: The GWN did not impact the overall lung sound identification capacity of medical students, with consistent scores of 66.7% across the three noise levels for all three lung sound classes combined. However, when considering sound classes separately, GWN affected the identification of normal (NAS) and discontinuous (DAS), but not continuous (CAS) types. Exam scores for NAS varied significantly across the three noise levels, with respective scores of 66.7%, 100% and 66.7%. Scores for DAS also varied, revealing 66.7%, 33.3% and 66.7%.

Conclusion: This study introduces a standardised simulation-based approach to investigate the effect of GWN on the accuracy of auscultation amongst medical students. Findings indicate that whilst CAS sounds are robust to background noise, the identification of NAS and DAS sounds can be compromised. The medium noise levels (SNR-40) of noise pollution had the greatest effect on the DAS lung sounds.

Figure 1

Study flowchart. Schematic diagram of all 52 subjects’ participation steps, showing the studying stages from subject enrolment to exam completion in the set order in three different groups.

Figure 2

Visualisation of GWN levels added to lung sounds. Spectrogram (top row) and waveform (bottom row) analysis from one 15 seconds recording. Brighter backgrounds in the spectrogram indicate increasing Gaussian white noise (GWN) intensity from lowest (no GWN), medium signal-to-noise ratio (SNR) −40, to highest SNR-20 (left to right column).

Figure 3

Website virtual patient ausculation section. Screenshot showing the top part of the website’s practice section, where six auscultation points with 15 seconds lung recordings are related to anatomic sites of the human body (back of the thorax).

Figure 4

Website individual sounds practice section. Screenshot showing the lower part of the website’s practice section, where the subject would study each lung sound individual before moving on to the cases.

Figure 5

Website test section. Screenshot showing the test section of the website.

Figure 6

Medical students’ exam scores for three classes of lung sounds under different levels of GWN. Impact of three levels of Gaussian white noise (GWN) on the ability of students to recognise continuous (CAS), discontinuous (DAS) and normal (NAS) lung sound classes. The noise levels are expressed in signal-to-noise ratio (SNR) from lowest levels (no GWN), medium (SNR-40) and to highest levels (SNR-20).