- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper proposes a method, GaLA-SVR, to reconstruct MRI scans into 3D volume measurements using 2D slices. The method corrects and synthesizes different 2D MRI slices from different angles by using a point-spread function to create more accurate and informed 3D reconstructions. This method is evaluated on the accuracy of different MRI reconstructions and how well the motion estimation is predicted.
Critical strengths
- This paper identified and explained the problem of 3D MRI reconstruction clearly, and the theory behind the method is well-defined with sound logic. It explains the limitations of current existing methods and proposes a logical solution.
- In the Experiment section, the results (Figure 3) and table (Table 1) match each other and it is clear where there are improvements on all benchmarks. The results are presented cleanly and make it easy to identify how the paper’s method is able to be close to the ground truth.
- In the Method section, the figure (Figure 2) displays the method relatively clearly, visually corresponding to each part of the method that is explained.
Critical weaknesses
- The datasets used for reconstruction of the 3D MRI volumes are under simulation, and not real-world patient motion data. This concerns if the results, no matter how well they performed, can perform equally as well in real patient scans.
- There is risk that the anatomy is not accurate to real-patient data and is instead derived from learned patterns from each stage. Even though there is noise reduction being done, it is possible to hallucinate structures that are not there and not accurate to human anatomy.
- Due to the limited amount of data, there might be too much of a generalization about the brain structure, as every patient’s brain will vary. If there is a case where a patient does not have the typical brain anatomy, it could result in more hallucinations about the entire structure.
- The comparison methods listed in the main results (Table 1 and Table 2) were mainly published more than two years ago; it is not possible to evaluate the performance compared with the SOTA methods in recent years.
Questions for authors
How well can the simulated MRI volume datasets be trusted for fair evaluation when part of the evaluation is for real motion accuracy from a patient’s movement?
6 comments
To reviewer NQ7t [Part II]
W2 & W3: Hallucination concern
For the concern that the anatomy may not be accurate on real-patient data, our response to W1 has shown that GaLA-SVR generalizes to real motion-corrupted neonatal scans, with the reconstructed anatomy consistent with the official dHCP [Ref1] reconstruction and no structure appearing that is absent from the reference (Table R1 and Figure 8 in the Appendix of the revised manuscript).
To further demonstrate that the reconstruction of GaLA-SVR is driven by the input slices rather than by learned patterns or anatomical priors alone, we compare it on the 76 dHCP test subjects with two prior-only predictions, which do not use the input slices at all. The first is the mean of the 675 training volumes, which is the same for every test subject. The second is, for each test subject, the training volume that is closest to its ground-truth volume, i.e., the best result that could be obtained by retrieving a brain from the training set. As shown in Table R2, they reach only 24.46 and 23.25 dB, whereas GaLA-SVR reaches 29.47 dB and is better on all 76 subjects. The performance therefore cannot be explained by learned patterns or anatomical priors alone. In addition, we conducted a perturbation experiment on the FeTA test set to verify that the reconstruction depends on the slice motion. We feed the same input slices to the network, but replace the slice motions estimated by the network with artificially perturbed ones before the PSF-aware aggregation. The perturbed motions are obtained by adding an independent random error to the ground-truth motion of each slice (a translation of mm plus a rotation of rad about the slice center). As shown in Table R3, the PSNR decreases monotonically from 30.24 dB with the ground-truth motions to 29.32, 26.89, and 22.34 dB at , 4, and 8 mm, which would not happen if the reconstruction were generated from learned patterns alone.
Table R2. Comparison with predictors that do not use the input slices on the 76 dHCP test subjects. Values are mean ± standard deviation.
| Prediction | PSNR ↑ | SSIM ↑ | NRMSE ↓ |
|---|---|---|---|
| Mean of the training volumes | 24.46 ± 1.02 | 0.810 ± 0.011 | 0.345 ± 0.035 |
| Most similar training volume | 23.25 ± 0.75 | 0.798 ± 0.008 | 0.396 ± 0.031 |
| GaLA-SVR | 29.47 ± 0.87 | 0.883 ± 0.007 | 0.194 ± 0.022 |
Table R3. Reconstruction of GaLA-SVR on the FeTA test set when the slice transformations passed to the aggregation are replaced. In-brain disp. is the mean displacement of the pixels inside the brain. Values are mean ± standard deviation.
| Slice transformations | ED (mm) | In-brain disp. (mm) | PSNR ↑ | SSIM ↑ |
|---|---|---|---|---|
| GT | 0.00 | 0.00 | 30.24 ± 2.10 | 0.888 ± 0.007 |
| GT + 2 mm | 3.54 ± 0.05 | 2.15 ± 0.03 | 29.32 ± 2.19 | 0.884 ± 0.010 |
| GT + 4 mm | 6.95 ± 0.20 | 4.32 ± 0.06 | 26.89 ± 2.47 | 0.862 ± 0.020 |
| GT + 8 mm | 14.17 ± 0.32 | 8.58 ± 0.18 | 22.34 ± 2.27 | 0.814 ± 0.030 |
| GaLA-SVR (estimated) | 6.10 ± 2.78 | 3.07 ± 1.29 | 29.11 ± 2.08 | 0.881 ± 0.011 |
To examine how GaLA-SVR behaves when a subject does not have the typical brain anatomy, we tested it on structures that never appear in the training data. We inserted synthetic structures into the ground-truth volumes of dHCP test subjects (a hypointense sphere of radius 3 mm or 8 mm), and re-simulated the stacks with the original motion, slice thickness, and noise. Figure 4 in the Appendix of the revised manuscript shows that GaLA-SVR reconstructs the inserted structures at the correct location and with a similar extent, and that the error outside the structure remains at the same level as in the surrounding tissue. This indicates that the reconstruction follows the anatomy of the individual subject rather than a generalized brain structure, and that the presence of an atypical structure does not lead to hallucinations in the rest of the brain.
[Ref1] Hughes, Emer J., et al. "A dedicated neonatal brain imaging system." Magnetic Resonance in Medicine 2017
To reviewer NQ7t [Part III]
W4: Baseline evaluation concern
We would like to clarify that the original comparison already included the state-of-the-art learning-based SVR methods, among which FC-SVR (2024) and SVRMamba (2025) were published within the last two years. Since GaLA-SVR reconstructs a volume in an end-to-end fashion within a second, we therefore considered learning-based baselines with similar computational requirements to be a fairer comparison, with only one optimization-based method, SVRTK, as reference. Following the suggestion, we have added four recent methods to all comparisons: three optimization-based methods, NeSVoR (2023), GaussianSVR (2026), and GSVR (2026), and cSVR [Ref1] (2026), a recently published method that combines a feed-forward motion estimator with an optimization-based reconstruction. Note that the first three methods require extra initialization of slice transformations provided by an external pretrained SVoRT method. Their results are summarized in Table R4 and Table R5.
On CamCAN and dHCP, where sufficient training subjects are available, GaLA-SVR remains the best on all three metrics. Its PSNR gains over the strongest added method on each dataset, 0.91 dB over cSVR on CamCAN and 0.45 dB over NeSVoR on dHCP, are statistically significant (P<0.01), while being orders of magnitude faster than per-subject optimization. On FeTA, where only 65 training subjects are available, all learning-based methods are limited by the data size, and NeSVoR and GaussianSVR, which benefit from per-subject optimization with an already good initialization, achieve the highest PSNR (31.85 and 31.31 dB). The reconstruction of cSVR is optimization-based, so it is likely less affected by the limited training data. We also added the results and discussion in Section 4.2 in the revised manuscript.
Table R4. Quantitative comparison of reconstruction performance on FeTA, CamCAN, and dHCP for the added methods and GaLA-SVR. Values are mean ± standard deviation. Best results are in bold and second-best in italics.
| Method | FeTA PSNR ↑ | FeTA SSIM ↑ | FeTA NRMSE ↓ | CamCAN PSNR ↑ | CamCAN SSIM ↑ | CamCAN NRMSE ↓ | dHCP PSNR ↑ | dHCP SSIM ↑ | dHCP NRMSE ↓ |
|---|---|---|---|---|---|---|---|---|---|
| Optimization-based | |||||||||
| NeSVoR | 31.85 ± 4.15 | 0.890 ± 0.034 | 0.167 ± 0.111 | 25.94 ± 1.36 | 0.864 ± 0.009 | 0.183 ± 0.034 | 29.02 ± 0.94 | 0.878 ± 0.008 | 0.204 ± 0.028 |
| GaussianSVR | 31.31 ± 2.83 | 0.889 ± 0.009 | 0.166 ± 0.053 | 25.02 ± 2.20 | 0.842 ± 0.041 | 0.224 ± 0.065 | 27.55 ± 1.56 | 0.859 ± 0.015 | 0.245 ± 0.045 |
| GSVR | 25.65 ± 2.26 | 0.716 ± 0.075 | 0.317 ± 0.100 | 23.40 ± 2.76 | 0.782 ± 0.059 | 0.275 ± 0.091 | 27.85 ± 0.73 | 0.787 ± 0.081 | 0.226 ± 0.031 |
| Learning-based | |||||||||
| cSVR | 29.12 ± 2.43 | 0.879 ± 0.013 | 0.206 ± 0.040 | 26.27 ± 0.86 | 0.861 ± 0.007 | 0.182 ± 0.023 | 28.21 ± 0.75 | 0.872 ± 0.007 | 0.225 ± 0.023 |
| GaLA-SVR | 29.11 ± 2.08 | 0.881 ± 0.011 | 0.198 ± 0.021 | 27.18 ± 1.21 | 0.873 ± 0.015 | 0.171 ± 0.023 | 29.47 ± 0.87 | 0.883 ± 0.007 | 0.194 ± 0.022 |
Table R5. Quantitative comparison of transformation errors and inference time on FeTA, CamCAN, and dHCP for the added methods and GaLA-SVR. ED and GD are mean ± standard deviation; inference time is the mean runtime per subject. Best transformation results are in bold and second-best in italics.
| Method | FeTA ED (mm) ↓ | FeTA GD (rad) ↓ | FeTA Inf. (s) ↓ | CamCAN ED (mm) ↓ | CamCAN GD (rad) ↓ | CamCAN Inf. (s) ↓ | dHCP ED (mm) ↓ | dHCP GD (rad) ↓ | dHCP Inf. (s) ↓ |
|---|---|---|---|---|---|---|---|---|---|
| Optimization-based | |||||||||
| NeSVoR | 5.96 ± 4.44 | 0.098 ± 0.082 | 115.49 | 6.37 ± 5.91 | 0.077 ± 0.079 | 146.80 | 3.98 ± 1.77 | 0.068 ± 0.038 | 106.42 |
| GaussianSVR | 3.35 ± 2.19 | 0.061 ± 0.043 | 647.33 | 6.33 ± 5.53 | 0.079 ± 0.074 | 1443.54 | 3.85 ± 2.04 | 0.073 ± 0.037 | 805.60 |
| GSVR | 11.38 ± 7.15 | 0.205 ± 0.136 | 78.14 | 9.16 ± 6.31 | 0.118 ± 0.086 | 318.68 | 4.35 ± 1.77 | 0.079 ± 0.029 | 165.26 |
| Learning-based | |||||||||
| cSVR | 6.31 ± 2.92 | 0.113 ± 0.058 | 4.44 | 4.56 ± 1.13 | 0.054 ± 0.016 | 7.63 | 5.23 ± 4.01 | 0.084 ± 0.081 | 5.30 |
| GaLA-SVR | 6.10 ± 2.78 | 0.110 ± 0.054 | 0.11 | 3.22 ± 1.04 | 0.041 ± 0.014 | 0.27 | 3.79 ± 3.62 | 0.066 ± 0.079 | 0.17 |
[Ref1] Xu, Junshen, et al. "cSVR: Convolutional Slice-to-Volume Reconstruction." MICCAI PIPPI Workshop 2026
To reviewer NQ7t [Part I]
We appreciate the reviewer's thoughtful evaluation and helpful recommendations, which have strengthened our paper. The manuscript has been revised accordingly, and we respond to each comment in turn below.
W1 & Q1: Real motion-corrupted data concern
To assess generalization beyond simulated motion, we further evaluate all methods on real motion-corrupted neonatal scans from the dHCP dataset [Ref1]. We select 30 subjects from the originally acquired, motion-corrupted T2-weighted stacks of the dHCP neonatal cohort. Each subject consists of 2 to 4 image stacks with in-plane resolution of 0.8 mm, slice gap of 0.8 mm, and slice thickness of 1.6 mm. Each stack is skull-stripped with SynthStrip [Ref2] from FreeSurfer, resampled in-plane to 1.0 mm, center-cropped to 128×128, intensity-normalized, and reoriented to the slice convention of the simulated stacks. The same preprocessing is applied to all methods. Since no ground-truth slice motion is available, we report volume reconstruction metrics only, using the official dHCP reconstruction of the same session. All learning-based models are trained only on simulated stacks on the dHCP dataset and applied to the real motion-corrupted data in a zero-shot manner.
As shown in Table R1, GaLA-SVR achieves the best zero-shot performance among all learning-based methods, with a PSNR of 21.00 dB and an SSIM of 0.761, outperforming the strongest learning-based baseline cSVR (20.66 dB). It also surpasses all optimization-based methods, including the recent Gaussian-based methods GaussianSVR and GSVR, as well as SVRTK and NeSVoR. Note that these four methods need long per-subject optimization time. NeSVoR, GaussianSVR, and GSVR start from slice transformations provided by a pretrained SVoRT. In contrast, GaLA-SVR is end-to-end and reconstructs a subject directly from the input stacks in 0.56 s with a single forward pass, about 200× faster than NeSVoR (110.5 s). This indicates that the learned prior transfers well to real acquisitions. Qualitative comparison can be found in Figure 8 in the Appendix of the revised manuscript. We have also added these results and discussion in the revised manuscript in Section 4.2.
Table R1. Quantitative comparison of reconstruction performance on real scans on the dHCP dataset. PSNR, SSIM, and NRMSE are reported as mean ± standard deviation; inference time is the mean runtime per subject. Best results are in bold and second-best in italics.
[Ref1] Hughes, Emer J., et al. "A dedicated neonatal brain imaging system." Magnetic Resonance in Medicine 2017
[Ref2] Hoopes, Andrew, et al. "SynthStrip: Skull-stripping for any brain image." NeuroImage 2022