Benchmarking debiasing methods for artificial intelligence in neuroimaging research

01/02/2026 to 31/01/2028

This project aims to address the critical issue of bias and fairness in artificial intelligence (AI) applications in neuroimaging, with the goal of promoting fairness and equity in healthcare. The study will systematically benchmark debiasing methods and evaluate them against fairness metrics, which can then be used to develop fairer AI models for clinical use. Furthermore, an open-source toolbox is being developed that will enable researchers and stakeholders to evaluate fairness metrics and apply debiasing methods to various datasets and predictive tasks, particularly in the field of clinical neuroimaging. By reducing algorithmic bias, this project aims to improve the inclusivity and reliability of AI models in clinical neuroimaging.