FairHealth: Open-Source Python Library for Trustworthy Healthcare AI in Low-Resource Settings

FairHealth is a Python library that operates as an open-source tool aimed at fostering reliable machine learning applications in healthcare, particularly in low-resource and low-income countries (LMICs) such as Bangladesh. It tackles four significant issues: the need for integrated fairness auditing for biosignals and clinical tabular data, the lack of privacy-preserving federated learning tools that align with standard ML processes, the absence of explainability tools suitable for low-bandwidth clinical decision support, and the lack of a toolkit for healthcare datasets from the Global South. Developed from five peer-reviewed studies, FairHealth comprises six modules, including federated learning with homomorphic encryption (fairhealth.federated) and hybrid fuzzy-SHAP explainability (fairhealth.explain). This library aims to enhance equity and trust in AI-based healthcare in underprivileged areas.