Evaluation of Risk of Bias in Neuroimaging-Based Artificial Intelligence Models for Psychiatric Diagnosis

Abstract

Background: The evaluation of risk of bias in neuroimaging-based artificial intelligence (AI) models for psychiatric diagnosis has garnered significant attention in recent scholarly discourse. This literature review synthesizes key findings from various studies that highlight the implications of bias in AI systems, particularly in the context of mental health diagnostics. Literature Review: Ethical considerations are further explored by (Goisauf & Cano Abadía, 2022), who highlight the societal implications of systemic biases in AI, particularly regarding gender and sex. This perspective is critical, as it addresses the potential for AI tools to perpetuate inequities in mental healthcare diagnostics. The comprehensive literature review by (Muley et al., 2023) categorizes the risks associated with AI in healthcare, including the risk of bias and perpetuation of inequities. Their call for collaboration among stakeholders to evaluate and regulate AI systems is vital for maximizing benefits while minimizing adverse impacts on patient care. The narrative review by (R. Gashkarimov et al., 2023) focuses on the application of machine learning techniques in diagnosing schizophrenia, acknowledging the challenges posed by data heterogeneity and the necessity for robust methodologies. This highlights the importance of addressing the variability in data sources to improve diagnostic accuracy. Further examination of bias in EHR-based AI models by (Chen et al., 2023) reveals the pervasive nature of implicit and selection biases, advocating for standardized frameworks to ensure fairness in AI applications. This insight is crucial for understanding how biases in healthcare data can affect AI outcomes. Conclusion: In conclusion, the literature reveals a nuanced understanding of the multifaceted challenges posed by bias in neuroimaging-based AI models for psychiatric diagnosis. There is a critical need for ongoing research, collaboration, and the development of standardized frameworks to ensure equitable and effective AI applications in mental healthcare. Addressing these challenges is essential for improving diagnostic accuracy and ensuring that AI technologies contribute positively to patient outcomes.