Editorial: Machine learning algorithms for brain imaging: new frontiers in neurodiagnostics and treatment

The field of neuroimaging has undergone profound transformation in recent years, driven primarily by rapid 3 advances in machine learning (ML), and especially deep learning (DL), techniques. These computational 4 innovations have been amplified by ongoing improvements in brain imaging modalities, such as structural 5 MRI (sMRI), functional MRI (fMRI), diffusion tensor imaging (DTI), and positron emission tomography 6 (PET), which now deliver increasingly high-resolution, multimodal views of brain structure, function, 7 connectivity, and metabolism.The intersection of neuroimaging and machine learning marks an inflection point in neuroscience; 9 a confluence where vast, high-dimensional brain data meet computational power capable of turning 10 complexity into clinically actionable insights. This convergence has led to the rapidly expanding domain of 11 machine learning for neuroimaging, inviting a new generation of studies that leverage ML to transcend the 12 limitations of traditional neuroimaging analysis.Traditional statistical approaches and model-based techniques often struggle with the defining features 14 of modern neuroimaging datasets: extreme high dimensionality, substantial heterogeneity across 15 subjects/scanners/protocols, nonlinear patterns, and large-scale "big data" from consortia and multimodal 16 fusion. In contrast, ML/DL paradigms provide powerful approaches to address the aforementioned 17 challenges by virtue of their ability to learn complex, non-linear relationships in the underlying data through Several challenges remain, such as overfitting, interpretability, explainability, computational demands, 23 and generalization across diverse datasets. Nevertheless, this integration holds transformative potential for 24 more accurate, individualized, and clinically translatable neuroscience insights.This Research Topic was conceived to provide a platform for researchers involved in a variety of related 26 domains such as neuroscience, medical imaging, and artificial intelligence to come together, and report 27 their state-of-the-art findings in the application of ML to neuroimaging. Our objective was to address both 28 the methodological and translational challenges involved, whilst also critically examining their clinical 29 relevance, robustness, interpretability, explainability and ethical implications in this rapidly evolving field.The Research Topic comprises 13 peer-reviewed articles, including original research papers and 31 methodological contributions, that collectively reflect the diversity and dynamism of current machine 32 learning research in brain imaging. This collection aims not only to push the methodological boundaries, but 33 to showcase the contributions made to translate the ML-driven neurimaging into real-world neurodiagnostic 34 and therapeutic applications. The collection also includes studies emphasizing multimodal data integration, where complementary