<p>Objectively and accurately evaluating the quality of Magnetic Resonance Images (MRI) remains a challenging task. Current mainstream approaches for image quality assessment (IQA) are primarily developed for the general domain. However, these methods mainly focus on the overall quality perception of natural scenes and fail to capture fine-grained characteristics that are crucial for MRI. To address this limitation, we first establish a large-scale, expert-annotated MRI perceptual quality preference dataset. We then propose MPQA, a multi-dimensional perceptual quality assessment model for MRI, which is capable of predicting both overall and multi-dimensional quality scores. This model offers detailed insights into the factors influencing image quality perception. Extensive experiments demonstrate that our model outperforms existing baselines in various evaluation tasks, showcasing its strong accuracy, robustness and generalizability.</p>