Evaluation of Respiratory Sounds Using Image-Based Approaches for Health Measurement Applications

Madison Cohen-McFarlane; Pengcheng Xi; Bruce Wallace; Karim Habashy; Saiful Huq; Rafik Goubran; Frank Knoefel

doi:10.1109/OJEMB.2022.3202435

Evaluation of Respiratory Sounds Using Image-Based Approaches for Health Measurement Applications

IEEE Open J Eng Med Biol. 2022 Sep 13:3:134-141. doi: 10.1109/OJEMB.2022.3202435. eCollection 2022.

Authors

Madison Cohen-McFarlane^{1

2}, Pengcheng Xi³, Bruce Wallace^{2

4

5}, Karim Habashy⁶, Saiful Huq⁷, Rafik Goubran^{2

5}, Frank Knoefel^{8

1

9}

Affiliations

¹ AGE-WELL NCECarleton University Ottawa ON K1S 5B6 Canada.
² AGE-WELL SAM3 National Innovation HubCarleton University Ottawa ON K1S 5B7 Canada.
³ Digital Technologies Research CentreNational Research Council Canada Ottawa ON K1A 0R6 Canada.
⁴ AGE-WELL NCECarleton University Ottawa ON K1S 5B7 Canada.
⁵ Bruyère Research Institute Ottawa ON K1N 5C8 Canada.
⁶ National Research Council Canada Ottawa ON K1A 0R6 Canada.
⁷ Department of Systems and Computer Engineering, Carleton University Ottawa ON K1S 5B6 Canada.
⁸ Bruyère Research Institute, Bruyère Continuing CareElisabeth Bruyère Hospital Ottawa ON K1N 5C8 Canada.
⁹ AGE-WELL SAM3 National Innovation Hub Ottawa ON K1S 5B7 Canada.

Abstract

Goal: The evaluation of respiratory events using audio sensing in an at-home setting can be indicative of worsening health conditions. This paper investigates the use of image-based transfer learning applied to five audio visualizations to evaluate three classification tasks (C1: wet vs. dry vs. whooping cough vs. restricted breathing; C2: wet vs. dry cough; C3: cough vs. restricted breathing). Methods: The five visualizations (linear spectrogram, logarithmic spectrogram, Mel-spectrogram, wavelet scalograms, and aggregate images) are applied to a pre-trained AlexNet image classifier for all tasks. Results: The aggregate image-based classifier achieved the highest overall performance across all tasks with C1, C2, and C3 having testing accuracies of 0.88, 0.88, and 0.91 respectively. However, the Mel-spectrogram method had the highest testing accuracy (0.94) for C2. Conclusions: The classification of respiratory events using aggregate image inputs to transfer learning approaches may help healthcare professionals by providing information that would otherwise be unavailable to them.

Keywords: Acoustic signal processing; audio visualization; convolutional neural network; cough classification; respiratory classification.

Grants and funding

This work was supported in part by the National Research Council of Canada through Aging in Place Challenge Program, and in part by the Natural Sciences and Engineering Research Council of Canada through Discovery Grant Program and was completed in collaboration with the AGE-WELL SAM3 National Innovation Hub. The work of Frank Knoefel was supported by the University of Ottawa Brain and Mind – Bruyère Research Institute Chair in Primary Health Care Dementia Research.