Defining and Benchmarking a Data-Centric Design Space for Brain Graph Construction

2025-08-17

AI Summary

Conventional fMRI-based brain graph construction relies on rigid, fixed pipelines that neglect data-driven design choices. Method: This paper systematically defines and evaluates a data-centric design space for brain graph construction, spanning three stages: time-series signal processing, topological extraction, and graph feature characterization. We introduce key data-centric strategies—including high-amplitude BOLD filtering, adaptive sparsification, and multi-view node/edge feature representation—and integrate enhanced correlation metrics, lagged dynamic modeling, connectivity harmonization, and multi-view feature fusion. Contribution/Results: Experiments on the HCP1200 and ABIDE datasets demonstrate substantial improvements in psychiatric disorder classification accuracy. Our results empirically validate that data construction critically determines downstream task performance, establishing a new paradigm for reproducible and interpretable brain network modeling.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Humans and AI: Brain-Sensing and Analysis

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Machine learning and data science for the Web
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

The construction of brain graphs from functional Magnetic Resonance Imaging (fMRI) data plays a crucial role in enabling graph machine learning for neuroimaging. However, current practices often rely on rigid pipelines that overlook critical data-centric choices in how brain graphs are constructed. In this work, we adopt a Data-Centric AI perspective and systematically define and benchmark a data-centric design space for brain graph construction, constrasting with primarily model-centric prior work. We organize this design space into three stages: temporal signal processing, topology extraction, and graph featurization. Our contributions lie less in novel components and more in evaluating how combinations of existing and modified techniques influence downstream performance. Specifically, we study high-amplitude BOLD signal filtering, sparsification and unification strategies for connectivity, alternative correlation metrics, and multi-view node and edge features, such as incorporating lagged dynamics. Experiments on the HCP1200 and ABIDE datasets show that thoughtful data-centric configurations consistently improve classification accuracy over standard pipelines. These findings highlight the critical role of upstream data decisions and underscore the importance of systematically exploring the data-centric design space for graph-based neuroimaging. Our code is available at https://github.com/GeQinwen/DataCentricBrainGraphs.

Problem

• Systematically defining data-centric design for brain graph construction
• Evaluating impact of existing techniques on downstream performance
• Improving classification accuracy via thoughtful data configurations

Innovation

• Data-centric AI for brain graph construction
• Three-stage design space organization
• Improved classification via data-centric configurations