Estimating Head Motion in Structural MRI Using a Deep Neural Network Trained on Synthetic Artifacts

2025-05-29

AI Summary

MRI head motion artifacts severely confound automated cortical thickness measurements, yet existing motion quantification methods rely on ground-truth motion labels, specialized hardware, or noisy real-world data. Method: We propose the first unsupervised 3D CNN framework for motion quantification, trained exclusively on controllably synthesized motion artifacts—requiring no ground-truth labels, prospective motion correction, or vendor-/protocol-specific adaptation. Contribution/Results: Our method achieves cross-site generalizability via synthetic data and attains R² = 0.65 against expert visual ratings across 14 independent datasets. It robustly replicates the well-established negative association between motion and cortical thickness in 12 of 15 datasets. Furthermore, predicted motion scores exhibit a biologically plausible, statistically significant inverse correlation with age (p < 0.001). This plug-and-play framework provides a scalable, objective, and vendor-agnostic tool for multi-site neuroimaging quality control.

Technology Category

• Computer Vision: Motion & Tracking
• Machine Learning: Calibration & Uncertainty Quantification
• Intelligent Robots: Motion and Path Planning

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Web Mining and Content Analysis: Web data quality in the era of algorithmically-generated content

Abstract

Motion-related artifacts are inevitable in Magnetic Resonance Imaging (MRI) and can bias automated neuroanatomical metrics such as cortical thickness. Manual review cannot objectively quantify motion in anatomical scans, and existing automated approaches often require specialized hardware or rely on unbalanced noisy training data. Here, we train a 3D convolutional neural network to estimate motion severity using only synthetically corrupted volumes. We validate our method with one held-out site from our training cohort and with 14 fully independent datasets, including one with manual ratings, achieving a representative $R^2 = 0.65$ versus manual labels and significant thickness-motion correlations in 12/15 datasets. Furthermore, our predicted motion correlates with subject age in line with prior studies. Our approach generalizes across scanner brands and protocols, enabling objective, scalable motion assessment in structural MRI studies without prospective motion correction.

Problem

• Estimating head motion in MRI scans objectively
• Overcoming limitations of manual review and noisy data
• Validating motion estimation across diverse scanner protocols

Innovation

• Deep neural network estimates head motion
• Uses synthetic artifacts for training
• Generalizes across scanner brands