Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Mitigating Sex Bias in Audio Data-driven COPD and COVID-19 Breathing Pattern Detection Models

Rachel Pfeifer, Sudip Vhaduri, James Eric Dietz

TL;DR

The paper tackles sex bias in audio-based COPD and COVID-19 breathing-pattern detection by evaluating fairness-aware post-processing with a threshold optimizer under demographic parity and equalized odds constraints. It uses binary decision-tree classifiers trained on MFCC features from two open-source breathing datasets, balancing cohorts to enable robust bias analysis. The study reports substantial bias mitigation, including an 81.43% reduction in demographic parity difference and a 71.81% reduction in equalized odds difference, with statistical significance, highlighting the potential for fairness-aware clinical AI in respiratory disease diagnosis. These findings underscore the importance of incorporating demographic fairness in data collection and model deployment to ensure equitable diagnostic performance across sexes.

Abstract

In the healthcare industry, researchers have been developing machine learning models to automate diagnosing patients with respiratory illnesses based on their breathing patterns. However, these models do not consider the demographic biases, particularly sex bias, that often occur when models are trained with a skewed patient dataset. Hence, it is essential in such an important industry to reduce this bias so that models can make fair diagnoses. In this work, we examine the bias in models used to detect breathing patterns of two major respiratory diseases, i.e., chronic obstructive pulmonary disease (COPD) and COVID-19. Using decision tree models trained with audio recordings of breathing patterns obtained from two open-source datasets consisting of 29 COPD and 680 COVID-19-positive patients, we analyze the effect of sex bias on the models. With a threshold optimizer and two constraints (demographic parity and equalized odds) to mitigate the bias, we witness 81.43% (demographic parity difference) and 71.81% (equalized odds difference) improvements. These findings are statistically significant.

Mitigating Sex Bias in Audio Data-driven COPD and COVID-19 Breathing Pattern Detection Models

TL;DR

The paper tackles sex bias in audio-based COPD and COVID-19 breathing-pattern detection by evaluating fairness-aware post-processing with a threshold optimizer under demographic parity and equalized odds constraints. It uses binary decision-tree classifiers trained on MFCC features from two open-source breathing datasets, balancing cohorts to enable robust bias analysis. The study reports substantial bias mitigation, including an 81.43% reduction in demographic parity difference and a 71.81% reduction in equalized odds difference, with statistical significance, highlighting the potential for fairness-aware clinical AI in respiratory disease diagnosis. These findings underscore the importance of incorporating demographic fairness in data collection and model deployment to ensure equitable diagnostic performance across sexes.

Abstract

In the healthcare industry, researchers have been developing machine learning models to automate diagnosing patients with respiratory illnesses based on their breathing patterns. However, these models do not consider the demographic biases, particularly sex bias, that often occur when models are trained with a skewed patient dataset. Hence, it is essential in such an important industry to reduce this bias so that models can make fair diagnoses. In this work, we examine the bias in models used to detect breathing patterns of two major respiratory diseases, i.e., chronic obstructive pulmonary disease (COPD) and COVID-19. Using decision tree models trained with audio recordings of breathing patterns obtained from two open-source datasets consisting of 29 COPD and 680 COVID-19-positive patients, we analyze the effect of sex bias on the models. With a threshold optimizer and two constraints (demographic parity and equalized odds) to mitigate the bias, we witness 81.43% (demographic parity difference) and 71.81% (equalized odds difference) improvements. These findings are statistically significant.
Paper Structure (27 sections, 2 figures)

This paper contains 27 sections, 2 figures.

Table of Contents

  1. Introduction
  2. Motivation
  3. Related Work
  4. Contributions
  5. Methodology
  6. Preliminaries
  7. Fairness and Bias
  8. Fairness Metrics
  9. Sensitive Features and Constraints
  10. Percentage Improvement
  11. Datasets
  12. Pre-processing and Feature Engineering
  13. Data Modeling
  14. Hyper-parameter Optimization
  15. Bias Mitigation
  16. ...and 12 more sections

Figures (2)

  • Figure 1: Bar graph with error bars representing findings of the demographic parity analysis
  • Figure 2: Bar graph with error bars representing findings of the equalized odds analysis