Brief

Researchers have developed GMN4AD, a novel Graph Matching Network designed for diagnosing Alzheimer's Disease using structural Magnetic Resonance Imaging (sMRI). This network addresses challenges posed by data heterogeneity across different sites by modeling interactions between brain graphs and employing a test-time domain adaptation strategy with contrastive learning to mitigate domain shifts during inference. Experiments on public datasets show GMN4AD outperforms existing state-of-the-art methods, offering a more robust and generalizable solution for early AD diagnosis. AI

Researchers have developed NeuroAgent, an LLM-driven framework designed to automate complex preprocessing and analysis for multimodal neuroimaging data. This system utilizes a hierarchical multi-agent architecture to generate, execute, and validate code for various imaging types like sMRI, fMRI, dMRI, and PET. Evaluations on a large dataset demonstrated NeuroAgent's capability to significantly reduce manual effort and enable end-to-end automated pipelines, achieving high accuracy in intent parsing and preprocessing correctness, with the strongest backend reaching 84.8% correctness. AI