Multi-Contrast Super-Resolution MRI Through a Progressive Network

• MRI offers clear tissue structure and functional information without ionizing radiation, while pulse sequences generate complementary T1, T2, and PD contrasts.Source: §I · p. 1
• High-resolution MR imaging is difficult in clinical applications because MRI systems face trade-offs among cost-effectiveness and signal-to-noise ratio.Source: §I · p. 1
• MCSR combines multiple contrast images to provide more image priors for recovering details than single-contrast super-resolution.Source: §I · p. 1, A. Overall SR Process · p. 2
• The paper proposes a one-level non-progressive network for low up-sampling factors and a two-level progressive network for high up-sampling factors.Source: §I · p. 1
• The proposed networks use high-level feature-space integration, and their generator objective combines adversarial, MSE, perceptual, and texture-matching losses.Source: §I · pp. 1–2

Down-sampling and Zero-filling

• Low-resolution T2 images are generated by retaining central k-space data at 25%, 11.1%, or 6.25% for 2-, 3-, or 4-fold down-sampling, then zero-filling and inverse Fourier transforming.Source: B. Down-sampling and Zero-filling · p. 2

One-Level Non-progressive Network

• The one-level non-progressive model uses a WGAN-GP framework whose generator contains an encoder-decoder network and a reference feature extraction network.Source: C. One-Level Non-progressive Network · p. 2, Figure 1 · p. 3
• Reference images are converted into feature maps and fed into the encoder-decoder network to guide reconstruction.Source: C. One-Level Non-progressive Network · p. 2

Objective Function

• The generator objective combines adversarial, MSE, perceptual, and texture-matching losses.Source: D. Objective Function · p. 2
• Perceptual loss compares high-level VGG16 feature maps, while texture matching compares Gram matrices to preserve texture similarity and use correct style-transfer features.Source: D. Objective Function · pp. 3–4

E. Two-Level Progressive Network

• Progressive super-resolution divides a large up-sampling task into sequential steps, each increasing image resolution by a smaller factor.Source: E. Two-Level Progressive Network · p. 4
• The proposed two-level network combines two one-level networks to produce 4-fold MCSR, with each level performing 2-fold up-sampling.Source: E. Two-Level Progressive Network · p. 4
• The first level is supervised with 2-fold down-sampled images, while the second level is supervised with the original high-resolution images.Source: E. Two-Level Progressive Network · p. 4
• The progressive objective combines losses from both levels using inputs representing 4-fold LR images, 2-fold LR images, and original HR images.Source: E. Two-Level Progressive Network · p. 4

• SSIM compares luminance, contrast, and structural similarity between images.The reported SSIM scores are averaged over 7 × 7 patches.Source: F. Image Quality Evaluation Metrics · p. 4
• PSNR is a fidelity measure related to MSE and expressed using the maximum image value and MSE.Source: F. Image Quality Evaluation Metrics · p. 4
• IFC evaluates image quality through mutual information between a reference image and a distorted image, incorporating human visual perception.Unlike SSIM and PSNR, IFC is not best suited to evaluating structural similarity.Source: F. Image Quality Evaluation Metrics · p. 4
• The three metrics are used together because each has limitations and no metric accurately evaluates image quality in every circumstance.Source: F. Image Quality Evaluation Metrics · p. 4

• The IXI dataset contains registered T2- and PD-weighted MRI images from 578 patients.The study selected 7,000 T2–PD pairs for training and 1,955 pairs for testing, using 10-fold cross-validation.Source: A. Datasets · p. 4
• IXI HR images have size 256 × 256, and T2 LR inputs were created using 2-fold, 3-fold, and 4-fold down-sampling.Source: A. Datasets · p. 4
• The NAMIC dataset contains registered T1- and T2-weighted MRI images from 20 patients.The study used 1,620 pairs for training and 180 different pairs for testing.Source: A. Datasets · p. 5
• For NAMIC, T2 images were super-resolved with T1 images as reference, using 2-, 3-, and 4-fold LR inputs derived from 256 × 256 HR images.Source: A. Datasets · p. 5

• Training used 112K 64 × 64 image patches, batch size 32, 50 epochs, and learning rate 2×10^-5.The discriminator was trained four times before each generator update.Source: B. Experimental Details · p. 5
• Hyperparameters λ1, λ2, and λ3 were selected using the objective-function tuning results shown in Fig. 3.The figure uses logarithmic x-axes.Source: B. Experimental Details · p. 5, Figure 3 · p. 5
• Perceptual loss, MSE loss, and Wasserstein distance were recorded each epoch to monitor convergence and compare feature-space, image-space, and distributional similarity.Source: A. Convergence Behavior · p. 5
• After 40 epochs, all monitored curves were stable and close to zero, indicating stable convergence and strong similarity to HR ground truth.Source: A. Convergence Behavior · p. 5
• The four ablation studies compare SISR, image synthesis, low-level multi-contrast integration, and high-level feature-space integration.Fig. 4 distinguishes the encoder-decoder network from the reference feature extraction network.Source: Figure 4 · p. 5

• The four ablation models test reference-image removal, direct image synthesis, low-level fusion, and high-level feature fusion.All models were evaluated using the IXI dataset to identify the most effective network structure for MCSR.Source: B. Ablation Study · p. 6
• Directly converting PD-weighted HR images into T2-weighted HR images produces the lowest SSIM, PSNR, and IFC scores and the largest pixel-value differences.This ablation also fails to recover fissure shapes visible in the other SR results.Source: B. Ablation Study · p. 6
• Using PD-weighted HR images improves MCSR over the single-contrast baseline, yielding higher PSNR, SSIM, and IFC scores and clearer structures.The highlighted curve remains difficult to see with the single-contrast model but becomes clearer when PD-weighted HR images are used.Source: B. Ablation Study · p. 6
• Combining T2-weighted LR and PD-weighted HR images in high-level feature space outperforms low-level image-space combination on PSNR, SSIM, and IFC.The high-level feature-space combination is the proposed network configuration.Source: B. Ablation Study · p. 6

C. Non-progressive Model MCSR Results with 2-fold, 3-fold and 4-fold Up-sampling

• The one-level model produces clearer shapes, texture details, and strong structural similarity even when LR images are highly down-sampled.Results are reported for 2-, 3-, and 4-fold up-sampling on the IXI and NAMIC datasets.Source: C. Non-progressive Model MCSR Results… · pp. 6–7, Figure 6 · p. 7, Figure 7 · p. 8
• Higher down-sampling factors produce more blurred LR images with fewer visible details, making the corresponding SR results harder to recover accurately.The paper reports lower image quality as the down-sampling factor increases.Source: C. Non-progressive Model MCSR Results… · p. 7
• Smaller down-sampling factors yield higher PSNR, SSIM, and IFC scores and smaller pixel-value differences from the HR ground truth.This pattern indicates better image quality for SR results generated from less aggressively down-sampled inputs.Source: C. Non-progressive Model MCSR Results… · p. 7

D. Progressive Model Further Improves Image Quality with 4-fold Up-sampling

• The two-level progressive model is introduced to further improve the quality of 4-fold MCSR results from highly down-sampled inputs.Its outputs are compared with those of the one-level non-progressive model.Source: C. Non-progressive Model MCSR Results… · p. 7
• The progressive model produces sharper edges, clearer texture details, higher PSNR, SSIM, and IFC scores, and smaller ground-truth differences than the non-progressive model.Both constrained and unconstrained progressive results outperform the non-progressive results; the constrained version scores higher than the unconstrained version.Source: C. Non-progressive Model MCSR Results… · p. 7

• The progressive model produces the highest SSIM and PSNR scores among the compared state-of-the-art MCSR methods.The one-level non-progressive model achieves the highest SSIM scores among all compared methods.Source: E. Comparing with State-of-the-art… · p. 8
• MCSR methods outperform SISR methods, indicating that multiple contrasts provide more image priors for super-resolution than a single contrast.Source: §V · p. 9
• Multi-contrast information combined in a high-level feature space yields better MCSR results than low-level image-space combination.Separate contrast-specific feature extraction and T2 feature skip connections support flexible feature extraction and stronger T2 characteristics in restoration.Source: §V · pp. 9–10
• Image quality decreases as the down-sampling factor increases because more blurred low-resolution images contain fewer recoverable details.Source: §V · p. 10
• The progressive model achieves significantly better 4-fold MCSR results than the one-level non-progressive network and can approach the latter’s 3-fold results.The authors attribute this improvement to strict parameter constraints and level-wise 2-fold training.Source: §V · p. 10

• The paper proposes a one-level non-progressive network for low-factor up-sampling and a two-level progressive network for large-factor up-sampling.Source: §VI · p. 10
• High-quality MCSR results are generated when multi-contrast information is combined in a high-level feature space.Source: §VI · p. 10
• The progressive network could be extended to larger up-sampling factors if more different contrast images are available.Source: §VI · p. 10