关键词: 磁共振指纹成像, 自监督学习, 参数量化, 成像物理模型, 流形结构

Abstract:

Magnetic resonance fingerprint (MRF) is an efficient multi-parameter quantitative imaging technology. However, traditional methods relying on signal dictionaries for parameter quantization are plagued by significant discretization errors and low matching efficiency. To overcome the limitations of existing supervised learning approaches that depend on pseudo-labels and lack physical interpretability, this study proposes a self-supervised parameter quantization method that integrates imaging physical models and manifold structure modeling. This method establishes reliable unlabeled constraints through Bloch equation-driven self-supervised physical consistency learning. By incorporating manifold structure-driven knowledge distillation, it transfers features of long frames to short frame models, realizing joint optimization of physical constraints and structural priors, thereby improving both accuracy and efficiency under unlabeled conditions. Experiments have verified this method’s superior accuracy and robustness, providing a novel approach for efficient and reliable MRF parameter estimation.

Key words: magnetic resonance fingerprint imaging, self-supervised learning, parameter quantization, imaging physical model, manifold structure