The 1st International Workshop on Neural Network Models for Brain Connectome Analysis (BrainNN2022) at IEEE BigData

The last decade has witnessed an impressive revolution in data science and machine learning characterized by deep learning methods. It is well known that deep learning techniques that are designed for 2D/3D grid-structured data like images or sequences are not directly applicable to non-Euclidean data (i.e., graphs and manifolds). This gap has sparked a surge in geometric deep learning research in various domains such as social networks, bioinformatics, and chemical engineering.

Recently, characterizing the connectome of the human brain by structural or functional connectivities has become one of the most pervasive paradigms in neuroimaging studies. Understanding the structural and functional mechanisms of the human brain has been an intriguing pursuit for researchers with various goals, including neural system simulation, mental disorder therapy, as well as general artificial intelligence. Brain connectome is inherently equipped with a graph structure defined on brain regions or voxels, entailing the use of geometric deep learning models, but questions remain to be explored on how to account for the error in brain parcellation; how to learn compact representations of brain connectomes with accuracy, efficiency, and robustness; how to handle heterogeneity underlying raw neuroimaging data; what kind of brain connectome structure is useful for downstream clinical analysis; how to improve the interpretability of neural network models on brain connectome datasets, etc.

In this first workshop of BrainNN, we encourage discussions and investigations of theories, methods, and applications of neural network models for brain connectome analysis. We seek contributions that include but are not limited to the following research topics:

  • Shallow or deep neural networks for neuroimage computing

  • Geometric deep learning and graph neural networks (GNNs) for brain connectome analysis

  • Multi-modal neuroimage fusion and integration with neural network models

  • Spatial-temporal neuroimage analysis with neural network models

  • Novel applications of neural network models in neuroimage acquisition, reconstruction, and analysis

  • New principles of deep learning in neuroimaging such as transfer learning, pre-training, knowledge distillation, meta-learning

  • Interpretable neural network models for mental disorder analysis and biomarker discovery

  • New datasets, benchmarks, and empirical studies of neural network models for brain connectome analysis and neuroimage computing

  • Ethics, privacy, fairness and robustness of neural network models for brain connectome analysis

In short, this BrainNN workshop aims at promoting novel concepts and methodologies of neural network models for brain connectome analysis and neuroscience to enable neurological discoveries and understandings. This workshop intends to bring together researchers from interdisciplinary fields in statistics, data science (DS), artificial intelligence (AI), neuroimaging, biomedical informatics, and public health by sharing visions of investigating new theories, methods, and applications at the intersection of neural networks and neuroscience.