An Adversarial Framework for Mitigating Gender Bias in Coronary Heart Disease Prediction

• In the rapidly evolving landscape of healthcare, Artificial Intelligence (AI) and Machine Learning (ML) are increasingly being employed to classify and diagnose patients. While these technologies offer immense potential, they also introduce significant challenges, particularly in the realm of bias and fairness. Our project focuses on a critical issue at the intersection of AI and healthcare: gender bias in the prediction of Coronary Heart Disease (CHD).
• Coronary Heart Disease, a type of Cardiovascular Disease (CVD), has historically been associated with significant gender disparities in diagnosis and treatment. Often perceived as a “man’s disease”, female patients of CHD have experienced higher rates of misdiagnoses and suboptimal care. Our project aims to address this issue by developing an innovative machine learning model that can accurately and efficiently predict CHD while mitigating gender-based biases.
• We implement a neural network (NN) model using an adversarial configuration to tackle this challenge. Our approach involves:
  • A primary neural network model for CHD classification
  • A secondary “discriminator” model designed to detect and penalize gender-based biases
  • By leveraging this adversarial setup, we encourage our primary model to focus on relevant diagnostic features while discouraging reliance on sensitive attributes like gender. This methodology aims to promote more equitable healthcare outcomes, particularly for underrepresented patient populations in CHD diagnosis and treatment.

Curious to learn more? Email me and I’ll forward you to the team! Or, check out their repo here