CHARMS	Full Name	Brief Description
CHARMS	CNN-Transformer Hybrid with Attention Regularization for MRI Super-Resolution	The proposed lightweight model for MRI super-resolution, combining CNN and Transformer elements with attention regularization.
RRAF	Reverse Residual Attention Fusion	Backbone module for hierarchical local feature extraction, integrating residual learning and attention.
RLFE	Residual Local Feature Extraction	Units within RRAF blocks, consisting of convolutions, ReLU activations, and ESA for feature encoding.
ESA	Enhanced Spatial Attention	Spatial attention operator that highlights high-frequency regions using dilated convolutions.
PCA	Pixel–Channel Attention	Mechanism merging pixel- and channel-level attention for fine-grained feature recalibration.
MDDTA	Multi-Depthwise Dilated Transformer Attention	Transformer block for efficient long-range dependency modeling with linear complexity.
GDDFN	Gated Depthwise Dilated Feed-Forward Network	Feed-forward component in the Transformer module, enhancing nonlinearity via gated convolutions.
THFIR	High-Frequency Information Refinement	Refinement module post-upsampling to restore high-frequency details and suppress artifacts.