Trustworthy generative and explainable AI for neuroimaging and neural signal analysis

Artificial intelligence is driving transformative change in neuroscience by revolutionizing the analysis of neuroimaging modalities—such as MRI, fMRI, PET—and electrophysiological recordings, including EEG and MEG. The integration of deep learning approaches into these data types has led to enhanced detection and characterization of neurological and psychiatric conditions, finer mapping of brain networks, and deeper understanding of brain-behavior relationships. Nevertheless, significant challenges persist: limited annotated datasets, data heterogeneity, model generalization issues, and a persistent demand for transparency undermine the reliable translation of AI methods into neuroscientific research and clinical workflows.Recent progress in generative artificial intelligence, encompassing GANs, VAEs, diffusion models, and foundation models, is helping to address some of these hurdles. In neuroscience, generative models support realistic data synthesis, augment datasets lacking sufficient samples, translate between different imaging and signal modalities, denoise or reconstruct incomplete data, and model rare or atypical brain conditions. Despite these advances, the opacity of modern AI systems poses ongoing concerns regarding trustworthiness and clinical acceptance. The need for robust explainable AI (XAI) techniques is paramount—saliency and relevance maps, concept-based explanations, and uncertainty quantification must be tailored to the nuances of neuroimaging and neural signal analysis for effective knowledge translation and regulatory compliance.This Research Topic aims to galvanize progress in designing trustworthy AI systems that combine generative modeling with explainability for neuroimaging and neural signal research. The focus is on fostering innovative, interpretable models and frameworks capable of meeting the technical and ethical challenges that arise in neuroscience. Our objective is to delineate pathways for the responsible adoption of AI-guided tools that facilitate discovery, enhance reliability, and earn the confidence of neuroscientists, clinicians, and stakeholders.To gather further insights within neuroimaging and neural signal analysis, we welcome articles addressing, but not limited to, the following themes:o Generative modeling for neuroimaging and electrophysiological datao Novel explainable AI methods for brain imaging and neural signal interpretationo Integration of generative and explainable AI in neuroscience researcho Multimodal and hybrid AI approaches for investigating brain structure and functiono Ensuring robustness, fairness, and mitigating bias in neuroscience AI modelso Federated and privacy-preserving learning for sensitive brain and neural datao Uncertainty quantification and reliability in brain data-driven predictiono Clinical and translational validation of AI methods in neuroscience Topic editors: Amira ECHTIOUI, Fathi KALLEL, Wassim ZOUCH, MOHAMED GHORBEL, Yassine Ben Ayed Source: https://www.frontiersin.org/research-topics/77981/trustworthy-generative-and-explainable-ai-for-neuroimaging-and-neural-signal-analysis