Automated Multi-sequence Cardiac MRI Segmentation Using Supervised Domain Adaptation

Andreas Maier; Nishant Ravikumar; Sulaiman Vesal

REVIEW 3 major objections 7 minor 17 references

Automated Multi-sequence Cardiac MRI Segmentation Using Supervised Domain Adaptation

T0 review · 3 major / 7 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Fine-tuning on just four late-gadolinium-enhanced cardiac MR scans lets a segmentation network pre-trained on T2 and bSSFP images reach 84.5% average Dice on 40 test subjects.

desk verdict A competent, incremental fine-tuning paper whose 40-subject test-set result is probably real, but whose headline 'significantly outperformed' claim rests on a five-subject validation comparison with no error bars and an inconsistency on training sample count. read the letter →

arxiv 1908.07726 v1 pith:XFSOCFEX submitted 2019-08-21 eess.IV cs.CV

Sulaiman Vesal , Nishant Ravikumar , Andreas Maier This is my paper · ORCID

classification eess.IVcs.CV

keywords cardiacMRIsegmentationsuperviseddomainadaptationtransferlearninglategadoliniumenhancementleftventriclefullyconvolutionalnetworkmulti-sequenceDiceloss

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to show that a cardiac MRI segmentation network can be adapted to a new imaging sequence with very few annotated examples. It first trains an encoder-decoder CNN on T2-weighted and balanced steady-state free precession images, then fine-tunes the same network on only four or five late gadolinium-enhanced (LGE) MR images. On a 40-subject LGE test set, the adapted network reaches an average Dice score of about 0.85, and on the validation set it clearly beats the same network trained from scratch on the few LGE samples (about 81% versus 67% average Dice). The point is that pixel-level annotations in one MR sequence can be leveraged to segment another sequence, reducing the annotation burden for LGE images.

What carries the argument

The load-bearing mechanism is two-stage transfer learning, a supervised domain adaptation scheme in which a small number of labeled target-domain (LGE) samples guide adaptation of a model trained on source-domain (T2+bSSFP) images. First, an encoder-decoder fully convolutional network with residual connections, skip connections, and a dilated-convolution bottleneck is trained on T2+bSSFP images using a multi-class soft Dice loss. Second, the same architecture is initialized with those learned weights and fine-tuned, with identical hyperparameters, on LGE-MR images. Data augmentation during fine-tuning further lifts validation Dice from 76.6% to 80.7%.

What would settle it

Run the same fine-tuning procedure on an external LGE-MR dataset acquired with a different scanner or field strength, using the same T2+bSSFP pre-trained weights and four annotated subjects for adaptation; if average Dice falls to the source-only level of about 31% rather than remaining near 85%, the adaptation does not generalize to new LGE domains.

Watch

Extended reading notes

Core claim

The central claim is that supervised domain adaptation by weight transfer across cardiac MR sequences makes LGE-MR segmentation feasible with very scarce target labels. A fully convolutional encoder-decoder trained on T2+bSSFP images with pixel-wise labels is initialized with those weights and then fine-tuned on just a handful of LGE-MR subjects. On a 40-subject LGE test set, the fine-tuned model obtains an average Dice of 0.844 ± 0.063 (myocardium 0.788, LV 0.912, RV 0.832). In contrast, a model trained from scratch on the same few LGE samples achieves 66.9% average Dice on the validation set, while the un-adapted source model essentially fails on LGE images with 31.3% average Dice. The paper attributes the improvement to better weight initialization and domain-invariant features learned from the source sequences.

Load-bearing premise

The claim rests on the assumption that anatomical features learned from T2 and bSSFP images transfer to LGE images through fine-tuning with only four or five annotated LGE subjects, despite large differences in contrast, resolution, slice thickness, and acquisition protocol.

Editorial extensions

If this is right

A practical path to LGE-MR segmentation with a handful of labeled subjects is to pre-train on annotated T2/bSSFP images and then fine-tune, reaching 0.844 average Dice on 40 held-out subjects.
Without fine-tuning, source-sequence knowledge does not transfer by itself: the un-adapted model scores only 0.313 average Dice on LGE images.
Training from scratch on the same few LGE labels is much weaker, with 0.669 validation Dice, so the method's value lies in weight initialization from multi-sequence data rather than in the architecture alone.
Data augmentation during fine-tuning adds roughly four Dice points on the validation set, from 0.766 to 0.807, indicating that aggressive augmentation helps in the small-target-label setting.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Beyond the paper, the same fine-tuning recipe could be applied to other scarce cardiac sequences, such as T1 mapping or edema-weighted images, whenever an abundant annotated sequence exists; the paper does not test this.
A T2-only or bSSFP-only pre-training ablation, which the paper does not report, would reveal whether combining both source sequences is necessary or whether any single annotated sequence suffices.
Because the network is 2D and slice thickness differs across sequences (5 mm for LGE versus 8–20 mm for source sequences), the transfer may rely on in-plane texture rather than volumetric anatomy; a slice-consistency or 3D evaluation would test this.
The fine-tuning set contains only five subjects with 5-fold cross-validation, so the paper does not measure how much performance varies with the choice of adaptation subjects; repeated sampling from a larger label pool would quantify that variance.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: this is a competent but incremental application of vanilla fine-tuning to LGE cardiac MRI segmentation, and the headline comparison is weaker than the prose suggests. The genuinely useful piece is Table 2, the 40-subject test-set performance of the domain-adapted model (avg Dice 0.844 ± 0.063). The novel part is not the method—supervised domain adaptation by fine-tuning is familiar and the authors cite the relevant literature—but the specific multi-sequence setting and the observation that T2+bSSFP pretraining gives a strong initialization for LGE with few labels. That is a legitimate contribution for the MS-CMRSeg challenge track and likely of practical interest to people working on CMR segmentation.

The paper does several things well. The architecture is clearly described, and the preprocessing and augmentation details are sufficient to approximate the pipeline. Reporting test-set Dice with standard deviations is good practice, and the comparison with a source-only model (31.3% Dice) demonstrates a real domain shift. The claim that fine-tuning is better than from-scratch training with few samples is plausible.

Where it gets soft: the abstract says 'trained with just four LGE-MR training samples,' while Section 2.4 says five LGE subjects with 5-fold cross-validation and Section 3 repeats five. That inconsistency matters because the validation comparison in Table 1 is the only evidence for the 'significantly outperformed' claim. The from-scratch baseline is a single average over five subjects with no per-structure variance; given that the model is selected on validation, fold choice could materially change the 13.8-point gap. The test-set numbers in Table 2 are more robust, but without a from-scratch baseline on the same 40 subjects or a comparison with an existing multi-sequence method (e.g., the MvMM baseline cited as [6]), the advantage over alternatives is not established. 'Significantly outperformed' is not supported by any statistical test. These are addressable, and I don't see a load-bearing flaw in the idea itself—fine-tuning from a related source domain is a reasonable way to handle scarce LGE labels.

Who this is for: researchers working on cardiac MRI segmentation, especially in low-label settings. It deserves referee time, but a serious referee should ask for a fixed-protocol from-scratch baseline with error bars and a comparison against prior multi-sequence methods. I would not desk-reject it; I would send it to review with expectations for revision.

Referee Report

3 major / 7 minor

Summary. The paper proposes a supervised domain adaptation approach for late gadolinium-enhanced (LGE) cardiac MRI segmentation. An encoder-decoder network is first trained on T2-weighted and bSSFP sequences with pixel-level annotations, then fine-tuned on a small number of LGE-MR subjects. The authors evaluate on the MS-CMRSeg 2019 challenge dataset and report validation Dice scores of 80.7% for the adapted network with augmentation versus 66.9% for a from-scratch baseline and 31.3% for a model trained only on T2/bSSFP. On the 40-subject test set they report an average Dice of 84.4%. The central claim is that feature transfer from T2/bSSFP to LGE with only a few target samples significantly outperforms training from scratch on the same LGE data.

Significance. If the reported results are robust, the method is practically relevant because it addresses the high cost of annotating LGE-MR images: the proposed transfer-learning recipe is simple, uses public challenge data, and could reduce the number of required target-domain annotations. The paper also clearly demonstrates the domain-shift problem by showing that a T2/bSSFP-only model performs poorly on LGE images. However, the significance is limited by the lack of statistical validation of the headline comparison, an unresolved inconsistency in the number of LGE training samples, and the absence of any comparison with prior multi-sequence segmentation methods on the same test set.

major comments (3)

[Abstract and Section 2.4] The abstract states that the domain-adapted network was trained with just four LGE-MR training samples, while Section 2.4 says the model was re-trained with five LGE subjects using 5-fold cross-validation. This is a load-bearing discrepancy because the paper's contribution is specifically that very few target-domain samples are sufficient. The authors must state exactly how many LGE subjects were used, how the folds were constructed, and which fold or aggregation produced the numbers in Table 1.
[Table 1] The validation comparison in Table 1 reports single average Dice values over five subjects (0.669 for the from-scratch baseline and 0.807 for the proposed method) with no per-subject results, no standard deviation, and no significance test. The abstract's claim that the proposed method 'significantly outperformed' the baseline is therefore not supported by the evidence presented. The authors should report per-fold results, confidence intervals, and a paired statistical test (for example Wilcoxon signed-rank or a bootstrap interval) on the five validation subjects.
[Section 3, Table 2] The 40-subject test evaluation in Table 2 reports results only for the domain-adapted method. Without a from-scratch baseline or a published multi-sequence method (e.g., MvMM, reference [6]) evaluated on the same test set, the central claim that domain adaptation outperforms no adaptation cannot be assessed on the test data. The 84.4% Dice value is useful, but it does not by itself establish the advantage of the adaptation mechanism. The authors should provide a test-set comparison, or clearly state the challenge ranking and report all compared methods under the same protocol.

minor comments (7)

[Keywords] The keyword 'Myocardial Infraction' should be 'Myocardial Infarction'.
[Abstract] The phrase 'the test set comprises of 40 LGE-MR subjects' is grammatically incorrect; 'comprises' should be used without 'of'.
[Section 2.1] The definition 'domain Dtb = {S, P(X)}' is incomplete: the notation is introduced informally and the label space Y is mentioned before the domain is properly defined. Please tighten the notation for clarity.
[Section 2.3] The text says 'Since T1 and bSSFP images have very few slices', but the paper only uses T2, bSSFP, and LGE sequences; this should read 'T2 and bSSFP'.
[Section 3 and Table 1] The text says 'the average Dice improved to 80.9%' while Table 1 reports 0.807 for the same row; use a consistent rounding convention.
[Figure 3 caption] The caption contains the typo 'withe color', which should be 'white color'.
[Equation (1)] The soft Dice loss in Eq. (1) omits the conventional factor of 2 in the numerator, so a perfectly overlapping prediction gives a loss of 0.5 rather than 0. Please clarify whether this is intentional or a typographical error.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: empirical transfer-learning evaluation against held-out annotations.

full rationale

The central result is an empirical Dice measurement on 40 LGE subjects using challenge ground truth (Table 2), not a derivation from an input. The comparison between fine-tuned and from-scratch networks is an experimental protocol: Section 2.4 describes 5-fold cross-validation on LGE subjects and Section 3 reports the resulting metrics. No equation in the paper defines LGE test performance in terms of the backbone training loss or the T2/bSSFP labels, and no fitted parameter is subsequently renamed as a prediction. The only self-citation is reference [13] for the architectural motif (dilated convolutions), which does not by itself force the reported Dice values. Concerns about validation-set model selection, a per-subject from-scratch baseline without standard deviation, and absent significance testing are statistical-reporting issues, not circularity. The evaluation is self-contained against external annotations, so the paper is not circular.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The central result depends on standard deep learning assumptions and a small number of hand-chosen hyperparameters. The most load-bearing premise is feature transferability across MRI sequences, which is empirically tested rather than proven. No fitted constants or new entities are used.

free parameters (4)

learning_rate = 0.0001
Adam learning rate chosen by hand and fixed for both training stages; standard value, not fitted to the target result.
batch_size = 16 (source), 4 (fine-tuning)
Batch sizes set by hardware limits and small target dataset; not tuned to maximize test score.
input_crop_size = 224x224
Cropped to remove background; a preprocessing choice that could affect segmentation accuracy.
dilation_rates = 1 to 8
Architecture choice for the bottleneck to enlarge receptive field; motivated by prior work [13].

assumptions (3)

domain assumption Features learned on T2 and bSSFP images are transferable to LGE images via fine-tuning.
The entire method relies on this premise; if source and target feature distributions are disjoint, fine-tuning on four or five subjects would not improve accuracy. Stated in Section 2.1, where Dtb and Dl are defined.
domain assumption The label space (LV, RV, Myo, background) is consistent across T2, bSSFP, and LGE sequences.
The network is trained and fine-tuned with the same class labels; if labels were semantically inconsistent across sequences, pre-training would be misleading. Implicit in Section 2.3 and the segmentation task.
standard math The multi-class soft Dice loss in Equation (1) is a valid proxy for segmentation quality.
Standard loss used in medical imaging; the equation is a conventional soft Dice formulation.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Automated Multi-sequence Cardiac MRI Segmentation Using Supervised Domain Adaptation." pith.science (2026). https://pith.science/paper/XFSOCFEX

@misc{pith2026190807726,
  author       = {Pith},
  title        = {Pith review of: Automated Multi-sequence Cardiac MRI Segmentation Using Supervised Domain Adaptation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XFSOCFEX}},
  note         = {Machine review of arXiv:1908.07726}
}

abstract

Left ventricle segmentation and morphological assessment are essential for improving diagnosis and our understanding of cardiomyopathy, which in turn is imperative for reducing risk of myocardial infarctions in patients. Convolutional neural network (CNN) based methods for cardiac magnetic resonance (CMR) image segmentation rely on supervision with pixel-level annotations, and may not generalize well to images from a different domain. These methods are typically sensitive to variations in imaging protocols and data acquisition. Since annotating multi-sequence CMR images is tedious and subject to inter- and intra-observer variations, developing methods that can automatically adapt from one domain to the target domain is of great interest. In this paper, we propose an approach for domain adaptation in multi-sequence CMR segmentation task using transfer learning that combines multi-source image information. We first train an encoder-decoder CNN on T2-weighted and balanced-Steady State Free Precession (bSSFP) MR images with pixel-level annotation and fine-tune the same network with a limited number of Late Gadolinium Enhanced-MR (LGE-MR) subjects, to adapt the domain features. The domain-adapted network was trained with just four LGE-MR training samples and obtained an average Dice score of $\sim$85.0\% on the test set comprises of 40 LGE-MR subjects. The proposed method significantly outperformed a network without adaptation trained from scratch on the same set of LGE-MR training data.

Figures

Figures reproduced from arXiv: 1908.07726 by the authors.

**Figure 1.** Illustration of different CMR sequences: a) bSSFP slice, b) T2-weighted slice and c) LGE slice. The red arrows point to left ventricle on different sequences. 2 Method 2.1 Domain Adaptation Deep learning methods are typically sensitive to domain shift and perform poorly on a new set of data with a different marginal probability distribution. However, annotating data for every new domain is a very expensive task, par… view at source ↗

**Figure 2.** Overview of our network architecture for feature transfer learning between different CMR sequences. The encoder-decoder first trained with T2+bSSFP images and in the second stage the network with learned weights retrained with LGE iamges. 2.2 Network Architecture Our network architecture is a fully convolutional network inspired from [13] which comprises four encoder and decoder blocks, separated by a bottleneck blo… view at source ↗

**Figure 3.** Visual comparison of segmentation results produced by different methods. From left to right are the raw LGE-MR images, ground truth, W/o Adaptation output trained from scratch on LGE-MR images only, output from model trained on T2+bSSFP images, and proposed network. The cardiac structures of LV, RV, and Myo are indicated in purple, blue and withe color respectively. Each row corresponds to one subject [PITH_FULL_I… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

17 extracted references · 17 canonical work pages

[6]

In: MICCAI

Zhuang, X.: Multivariate mixture model for cardiac segmentation from multi- sequence mri. In: MICCAI. (2016) 581–588

work page 2016
[1]

Journal of the American College of Cardiology 55(1) (2009) 1 – 16

Kim, H.W., Farzaneh-Far, A., Kim, R.J.: Cardiovascular magnetic resonance in patients with myocardial infarction: Current and emerging applications. Journal of the American College of Cardiology 55(1) (2009) 1 – 16

work page 2009
[2]

Group, E.S.D., Timmis, A., Wilkins, E., Wright, L., Townsend, N., Gale, C., Grobbee, R., Vos, R., Blum, M., Maniadakis, N., Vardas, P., Flather, M., Bax, J., Pinto, F.: European Society of Cardiology: Cardiovascular Disease Statistics

work page
[3]

IEEE Transactions on Pattern Analysis and Machine Intelli- gence (2019) 1–1

Zhuang, X.: Multivariate mixture model for myocardial segmentation combining multi-source images. IEEE Transactions on Pattern Analysis and Machine Intelli- gence (2019) 1–1

work page 2019
[4]

European Heart Journal - Cardiovascular Imaging 17(1) (05 2015) 76–84

Hammer-Hansen, S., Bandettini, W.P., Hsu, L.Y., Leung, S.W., Shanbhag, S., Mancini, C., Greve, A.M., Kber, L., Thune, J.J., Kellman, P., Arai, A.E.: Mecha- nisms for overestimating acute myocardial infarct size with gadolinium-enhanced cardiovascular magnetic resonance imaging in humans: a quantitative and kinetic study. European Heart Journal - Cardiovas...

work page 2015
[5]

Com- puterized Medical Imaging and Graphics (59) (2017) 13–27

Kurzendorfer, T., Forman, C., Schmidt, M., Tillmanns, C., Maier, A., Brost, A.: Fully automatic segmentation of left ventricular anatomy in 3-DLGE-MRI. Com- puterized Medical Imaging and Graphics (59) (2017) 13–27

work page 2017
[7]

Computerized Medical Imaging and Graphics 71 (2019) 49 – 57

Liu, J., Xie, H., Zhang, S., Gu, L.: Multi-sequence myocardium segmentation with cross-constrained shape and neural network-based initialization. Computerized Medical Imaging and Graphics 71 (2019) 49 – 57

work page 2019
[8]

In: Machine Learning in Medical Imaging

Chen, C., Dou, Q., Chen, H., Heng, P.A.: Semantic-aware generative adversarial nets for unsupervised domain adaptation in chest x-ray segmentation. In: Machine Learning in Medical Imaging. (2018) 143–151

work page 2018

Show all 17 references

[9]

In: CVPR

Russo, P., Carlucci, F.M., Tommasi, T., Caputo, B.: From source to target and back: Symmetric bi-directional adaptive gan. In: CVPR. (June 2018)

2018
[10]

In: 2017 IEEE International Conference on Computer Vision (ICCV)

Motiian, S., Piccirilli, M., Adjeroh, D.A., Doretto, G.: Uniﬁed deep supervised domain adaptation and generalization. In: 2017 IEEE International Conference on Computer Vision (ICCV). (Oct 2017) 5716–5726

2017
[11]

IEEE Transactions on Knowl- edge and Data Engineering 22(10) (Oct 2010) 1345–1359

Pan, S.J., Yang, Q.: A survey on transfer learning. IEEE Transactions on Knowl- edge and Data Engineering 22(10) (Oct 2010) 1345–1359

2010
[12]

In: MICCAI 2017

Ghafoorian, M., Mehrtash, A., Kapur, T., Karssemeijer, N., Marchiori, E., Pesteie, M., Guttmann, C.R.G., de Leeuw, F.E., Tempany, C.M., van Ginneken, B., Fe- dorov, A., Abolmaesumi, P., Platel, B., Wells, W.M.: Transfer learning for domain adaptation in mri: Application in bra...

2017
[13]

In: STACOM

Vesal, S., Ravikumar, N., Maier, A.: Dilated convolutions in neural networks for left atrial segmentation in 3d gadolinium enhanced-mri. In: STACOM. (2018) 319–328

2018
[14]

In: CVPR

He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: CVPR. (June 2016) 770–778

2016
[15]

In: ICLR

Yu, F., Koltun, V.: Multi-scale context aggregation by dilated convolutions. In: ICLR. (2016)

2016
[16]

In: 2016 Fourth International Conference on 3D Vision (3DV)

Milletari, F., Navab, N., Ahmadi, S.: V-net: Fully convolutional neural networks for volumetric medical image segmentation. In: 2016 Fourth International Conference on 3D Vision (3DV). (Oct 2016) 565–571

2016
[2017]

European Heart Journal 39(7) (11 2017) 508–579

2017

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [6]

In: MICCAI

Zhuang, X.: Multivariate mixture model for cardiac segmentation from multi- sequence mri. In: MICCAI. (2016) 581–588

work page 2016

[2] [1]

Journal of the American College of Cardiology 55(1) (2009) 1 – 16

Kim, H.W., Farzaneh-Far, A., Kim, R.J.: Cardiovascular magnetic resonance in patients with myocardial infarction: Current and emerging applications. Journal of the American College of Cardiology 55(1) (2009) 1 – 16

work page 2009

[3] [2]

Group, E.S.D., Timmis, A., Wilkins, E., Wright, L., Townsend, N., Gale, C., Grobbee, R., Vos, R., Blum, M., Maniadakis, N., Vardas, P., Flather, M., Bax, J., Pinto, F.: European Society of Cardiology: Cardiovascular Disease Statistics

work page

[4] [3]

IEEE Transactions on Pattern Analysis and Machine Intelli- gence (2019) 1–1

Zhuang, X.: Multivariate mixture model for myocardial segmentation combining multi-source images. IEEE Transactions on Pattern Analysis and Machine Intelli- gence (2019) 1–1

work page 2019

[5] [4]

European Heart Journal - Cardiovascular Imaging 17(1) (05 2015) 76–84

Hammer-Hansen, S., Bandettini, W.P., Hsu, L.Y., Leung, S.W., Shanbhag, S., Mancini, C., Greve, A.M., Kber, L., Thune, J.J., Kellman, P., Arai, A.E.: Mecha- nisms for overestimating acute myocardial infarct size with gadolinium-enhanced cardiovascular magnetic resonance imaging in humans: a quantitative and kinetic study. European Heart Journal - Cardiovas...

work page 2015

[6] [5]

Com- puterized Medical Imaging and Graphics (59) (2017) 13–27

Kurzendorfer, T., Forman, C., Schmidt, M., Tillmanns, C., Maier, A., Brost, A.: Fully automatic segmentation of left ventricular anatomy in 3-DLGE-MRI. Com- puterized Medical Imaging and Graphics (59) (2017) 13–27

work page 2017

[7] [7]

Computerized Medical Imaging and Graphics 71 (2019) 49 – 57

Liu, J., Xie, H., Zhang, S., Gu, L.: Multi-sequence myocardium segmentation with cross-constrained shape and neural network-based initialization. Computerized Medical Imaging and Graphics 71 (2019) 49 – 57

work page 2019

[8] [8]

In: Machine Learning in Medical Imaging

Chen, C., Dou, Q., Chen, H., Heng, P.A.: Semantic-aware generative adversarial nets for unsupervised domain adaptation in chest x-ray segmentation. In: Machine Learning in Medical Imaging. (2018) 143–151

work page 2018