A generalisable head MRI defacing pipeline: Evaluation on 2,566 meningioma scans

2025-05-19

AI Summary

Existing MRI de-identification methods often fail to simultaneously ensure complete facial feature removal and preserve anatomical integrity of brain structures in high-resolution images, hindering multi-center collaborative research. To address this, we propose a highly robust, anatomy-preserving cranial MRI de-identification framework. Our method introduces a registration-guided brain mask fusion strategy, integrating atlas-driven nonlinear registration, adaptive brain tissue segmentation, morphological refinement, and embedded multi-scale quality control. Evaluated on 2,566 heterogeneous clinical meningioma MRI scans, the framework achieves a 99.92% visual pass rate, a Dice coefficient of 0.9975 ± 0.0023 for brain mask accuracy, and successful de-identification in 2,564 out of 2,566 cases. The approach demonstrates strong generalizability and plug-and-play usability across diverse scanners and protocols. Source code is publicly available.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multimodal Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Security and Privacy: Data transparency and provenance

Abstract

Reliable MRI defacing techniques to safeguard patient privacy while preserving brain anatomy are critical for research collaboration. Existing methods often struggle with incomplete defacing or degradation of brain tissue regions. We present a robust, generalisable defacing pipeline for high-resolution MRI that integrates atlas-based registration with brain masking. Our method was evaluated on 2,566 heterogeneous clinical scans for meningioma and achieved a 99.92 per cent success rate (2,564/2,566) upon visual inspection. Excellent anatomical preservation is demonstrated with a Dice similarity coefficient of 0.9975 plus or minus 0.0023 between brain masks automatically extracted from the original and defaced volumes. Source code is available at https://github.com/cai4cai/defacing_pipeline.

Problem

• Ensuring reliable MRI defacing to protect patient privacy
• Addressing incomplete defacing or brain tissue degradation issues
• Developing a generalizable pipeline for high-resolution MRI defacing

Innovation

• Integrates atlas-based registration with brain masking
• Achieves 99.92% success rate on clinical scans
• Preserves anatomy with high Dice similarity coefficient