A Three-Paper Series on Cardiac Cine-MRI Analysis: From Boundary-Aware Segmentation to Clinical Indices

DOI: https://doi.org/10.62517/jmhs.202605408

Author(s)

Yali Wan1,*, Deyi Xiong2

Affiliation(s)

1School of Information Technology and Engineering, Guangzhou College of Commerce, Guangzhou, China 2Chongqing Academy of Metrology and Quality Inspection, Chongqing, China *Corresponding Author

Abstract

We present a three-paper series on automated cardiac cine-MRI analysis using the ACDC benchmark. The series progresses from architectural innovation to semi-supervised learning to multi-task clinical prediction. Paper 1 introduces a lightweight hybrid CNN-Transformer with boundary-aware skip refinement and a compound boundary-aware loss, achieving mDice 0.859 ± 0.004 (3-fold CV) with RV/MYO/LV HD95 of 3.03 ± 0.71 / 1.94 ± 0.05 / 1.33 ± 0.14 pixels. Paper 2 extends to semi-supervised learning via temporal consistency on 28 unused cine frames per patient, achieving mDice 0.713 ± 0.011 (3-fold CV) while exploiting 15× as much data per patient. Paper 3 proposes joint segmentation plus clinical-index prediction (EDV, ESV, EF) with uncertainty-weighted multi-task loss, achieving mDice 0.823 ± 0.003 (3-fold CV) alongside EF MAE 11.66 ± 1.75% (direct) / 3.47 ± 1.14% (Simpson on model). All three papers now report segmentation metrics under a unified per-slice ED+ES aggregation rule shared across the series, so the cross-paper numbers are directly comparable; the widened Paper 2 gap relative to earlier patient-level ED-only reports reflects this unified protocol, under which ES frames are intrinsically harder than ED. This overview paper synthesises the three contributions, situates them in the literature, discusses the engineering trade-offs (parameter count, inference cost, training time, data efficiency), and outlines a unified research direction: cardiac cine-MRI analysis that is cheap to train, sparing with labels, and tied to clinical outputs.

Keywords

Cardiac Cine-MRI; ACDC, Deep Learning; Research Series; Semi-Supervised Learning; Multi-Task Learning

References

[1] Bernard O, Lalande A, Zotti C, et al. Deep Learning Techniques for Automatic MRI Cardiac Multi-structures Segmentation and Diagnosis: Is the Problem Solved?. IEEE Transactions on Medical Imaging, 2018, 37(11):2514-2525. DOI:10.1109/TMI.2018.2837502. [2] Kendall A, Gal Y, Cipolla R. Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2018:7482-7491. DOI:10.1109/CVPR.2018.00781. [3] Chen J, Lu Y, Yu Q, et al. TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation. arXiv:2102.04306, 2021. DOI:10.48550/arXiv.2102.04306. [4] Cao H, Wang Y, Chen J, et al. Swin-Unet: Unet-like Pure Transformer for Medical Image Segmentation. European Conference on Computer Vision Workshops (ECCVW), 2022, LNCS 13803:205-218. DOI: 10.1007/978-3-031-25066-8_9. [5] Ronneberger O, Fischer P, Brox T. U-Net: Convolutional Networks for Biomedical Image Segmentation. Medical Image Computing and Computer-Assisted Intervention (MICCAI), 2015, LNCS 9351:234-241. DOI: 10.1007/978-3-319-24574-4_28. [6] Oktay O, Schlemper J, Folgoc L L, et al. Attention U-Net: Learning Where to Look for the Pancreas. arXiv:1804.03999, 2018. DOI:10.48550/arXiv.1804.03999. [7] Isensee F, Jaeger P F, Kohl S A A, et al. nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nature Methods, 2021, 18:203-211. DOI: 10.1038/s41592-020-01008-z. [8] Milletari F, Navab N, Ahmadi S A. V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation. Fourth International Conference on 3D Vision (3DV), 2016:565-571. DOI:10.1109/3DV.2016.79. [9] Tarvainen A, Valpola H. Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results. Advances in Neural Information Processing Systems (NeurIPS), 2017, 30:1195-1204. [10] Sohn K, Berthelot D, Carlini N, et al. FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence. Advances in Neural Information Processing Systems (NeurIPS), 2020, 33:596-608. [11] Venugopalan J, Tong L, Hassanzadeh H R, Wang M D. Multimodal deep learning models for early detection of Alzheimer's disease stage. Scientific Reports, 2021, 11:3254. DOI: 10.1038/s41598-020-74399-w. [12] Suresh H, Hunt N, Johnson A, et al. Clinical Intervention Prediction and Understanding with Deep Neural Networks. Proceedings of the 2nd Machine Learning for Healthcare Conference (MLHC), 2017, PMLR 68:322-337. DOI:10.48550/arXiv.1705.08498. [13] Weninger L, Rippel O, Koppers S, Merhof D. Segmentation of Brain Tumors and Patient Survival Prediction: Methods for the BraTS 2018 Challenge. International Workshop on Brain Lesion (BrainLes), 2018, LNCS 11384:3-12. DOI: 10.1007/978-3-030-11726-9_1. [14] Heker M, Ben-Cohen A, Greenspan H. Hierarchical Fine-Tuning for Joint Liver Lesion Segmentation and Lesion Classification in CT. 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2019:895-898. DOI:10.1109/EMBC.2019.8857127. [15] Esteva A, Kuprel B, Novoa R A, Ko J, Swetter S M, Blau H M, Thrun S. Dermatologist-level classification of skin cancer with deep neural networks. Nature, 2017, 542:115-118. DOI: 10.1038/nature21056. [16] Tang Y, Yang D, Li W, et al. Self-Supervised Pre-Training of Swin Transformers for 3D Medical Image Analysis. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022:20730-20740. DOI:10.1109/CVPR52688.2022.02011.