Reliability-Role Collaborative Learning for Semi-Supervised Cardiac Image Segmentation
Authors: Zhongrui Xie, Wenfeng Zhang, Wei Hu, Ruitao Xie, Wenjie Wang, Zhouyun Xie, Yi Zhang, Tao Jia
Organizations: College of Computer and Information Science, Research Center of Biomimetic Intelligence and Computational Biomedicine, Chongqing Normal University, Chongqing, China · Shenzhen University of Advanced Technology, Shenzhen, China · Wuwei Tumour Hospital of Gansu Province, Wuwei, China
Semi-supervised cardiac image segmentation aims to exploit abundant unlabeled images together with limited annotations, but the reliability of pseudo-label supervision is highly uneven across spatial regions. In a dual-network setting, interior regions are usually more stable, whereas boundary regions, low-confidence areas, and disagreements between networks can introduce noisy supervision. However, existing methods typically impose uniform supervision on all unlabeled pixels, overlooking reliability differences and limiting discriminative learning in challenging areas. To address this issue, we propose R2CL (Reliability-Role Collaborative Learning), a framework for semi-supervised cardiac image segmentation. R2CL models confidence relationships between two networks to assign roles to unlabeled regions: mutually uncertain regions are used for entropy regularization, confidence-discrepant regions for directional knowledge transfer, and mutually confident regions as reliable supervision anchors. Reliability-Role Mixed Cross Learning (RR-MCL) implements these roles through region-mixed training, directional cross-network supervision, and entropy regularization over mutually uncertain regions. Reliability-Transition Prototype Contrastive Learning (RT-PCL) constructs query features from foreground-adjacent reliability-transition regions and learns network-specific class prototypes from mutually confident regions, improving feature discrimination around ambiguous boundaries. High-confidence Context Refinement (HCR) exploits reliable interior pixels and trusted dense blocks within mutually confident regions to complement stable contextual supervision. R2CL achieves stable improvements in region-overlap and boundary-related metrics across multiple datasets, with more evident advantages on the foreground-sparse MVAA-CT dataset.
Figures & tables
Fig. 1: Overall framework of R2CL. (a) RR-MCL constructs reliability-role mixed cross learning by converting unlabeled regions into region-mixed training, directional cross-network supervision, and entropy regularization signals. (b) RT-PCL performs foreground-adjacent reliability-transition query mining and network-specific class-prototype contrast. (c) HCR exploits reliable interior pixels and trusted dense blocks in mutually confident regions to enhance structural consistency.
Method
Labeled
Unlabeled
DSC ↑
Jaccard ↑
95HD ↓
ASD ↓
U-Net † [ 1 ]
7(10%)
0
77.07
65.14
8.47
2.65
14(20%)
0
84.12
74.21
6.46
2.03
70(All)
0
91.47
84.68
1.40
0.34
CPS † [ 7 ]
7(10%)
63(90%)
86.74
77.43
6.64
1.80
MC-Net+ † [ 10 ]
86.31
76.71
8.23
2.20
BCP † [ 12 ]
88.23
79.71
4.54
1.41
TABLE I: Comparison on the ACDC test set. Best and second-best results are bold and underlined; † denotes reproduced results.
Method
Labeled
Unlabeled
DSC ↑
Jaccard ↑
95HD ↓
ASD ↓
3D U-Net † [ 20 ]
2
0
74.55 ± 3.83
60.13 ± 4.90
4.56 ± 2.10
1.18 ± 0.44
4
0
81.06 ± 0.93
68.34 ± 1.27
1.83 ± 0.43
0.59 ± 0.14
MC-Net+ † [ 10 ]
2
1058
73.28 ± 2.21
58.92 ± 3.36
3.23 ± 1.02
0.74 ± 0.25
CAML † [ 26 ]
72.33 ± 3.17
57.25 ± 4.02
3.79 ± 1.47
0.80 ± 0.42
BCP † [ 12 ]
75.92 ± 1.98
61.22 ± 2.56
2.58 ± 0.67
0.73 ± 0.16
MLRP † [ 22 ]
71.46 ± 3.52
56.94 ± 4.13
4.26 ± 1.67
1.15 ± 0.94
TABLE II: Five repeated labeled-case sampling results on the MVAA-CT validation set. Results are reported as mean ± standard deviation. Best and second-best mean values are bold and underlined; † denotes reproduced results.
Fig. 2: Segmentation visualizations on ACDC test cases and MVAA-CT validation cases. Panel (A) shows ACDC with 7 labeled cases (10%), and panel (B) shows MVAA-CT under the 4-labeled setting. Comparisons are intended within each dataset.
Dataset
Variant
DSC ↑
Jaccard ↑
95HD ↓
ASD ↓
ACDC
Baseline
86.54
77.32
5.96
1.79
+RR-MCL
88.39
79.78
3.23
0.68
+RR-MCL+RT-PCL
89.22
81.32
1.85
0.64
+RR-MCL+HCR
89.56
81.61
2.41
0.66
R2CL(Ours)
89.93
82.16
1.43
0.47
MVAA-CT
Baseline
71.93
56.45
4.98
1.28
TABLE III: Stepwise component ablation on ACDC and MVAA-CT. ACDC uses the 7-labeled setting, and MVAA-CT uses the 2-labeled setting. Component ablation is conducted on the same default split.
Fig. 3: RR-MCL training dynamics on MVAA-CT: region proportions, directional KL reduction, and raw/mixed-view uncertainty. Here, ΩCC , ΩUU , Ω1←2 , and Ω2←1 denote mutually confident, mutually uncertain, network-2-guided, and network-1-guided regions, respectively.
Fig. 4: Per-case DSC and ASD distributions of the full model and RT-PCL ablation variants on the ACDC test set.
Semi-supervised medical image segmentation uses labeled and unlabeled scans to train models that delineate organs, lesions, and tissues with fewer annotations.
Semi-supervised medical image segmentation aims to train reliable segmentation models using limited labeled data and abundant unlabeled data. However, existing methods typically rely on prediction confidence or model consistency to estimate pseudo-label reliability, but these criteria are insufficient to characterize both semantic reliability and geometric plausibility. To address this issue, we propose ProGeo-RC, a prototype-guided semantic reliability and geometric error calibration framework for semi-supervised medical image segmentation. ProGeo-RC consists of two key modules: Prototype-guided Semantic Reliability Estimation (PSRE) and Structured Semantic-Geometric Error Calibration (S-SGEC). PSRE uses probabilistic class prototypes to estimate the semantic reliability of unlabeled predictions and calibrate pixel-wise teacher pseudo labels, while S-SGEC introduces signed-distance-field-based geometric constraints to refine ambiguous boundaries and suppress spurious foreground responses outside target structures. Experiments on the ACDC and MVAA-CT datasets show that ProGeo-RC achieves better performance compared with representative semi-supervised segmentation methods. Under the 10% labeled setting on ACDC, ProGeo-RC improves DSC and Jaccard by 0.37% and 0.49% over the corresponding second-best results, while reducing 95HD and ASD by 0.61 and 0.21, respectively. These results indicate that joint semantic reliability estimation and geometric error calibration can provide more reliable pseudo supervision for semi-supervised medical image segmentation.
Yunfei Zhu, Wenfeng Zhang, Pan Zeng +5
College of Computer and Information Science, Research Center of Biomimetic Intelligence and Computational Biomedicine, Chongqing Normal University, Chongqing, China · School of Educational Sciences, Chongqing Normal University, Chongqing, China · School of Physics and Electrical Engineering, Chongqing Normal University, Chongqing, China
Accurate cardiac ultrasound segmentation is essential for reliable assessment of ventricular function in intelligent healthcare systems. However, echocardiographic images are challenging due to low contrast, speckle noise, irregular boundaries, and domain shifts across devices and patient populations. Existing methods, largely based on appearance-driven learning, often fail to preserve boundary precision and structural consistency under these conditions. To address these issues, we propose a Contour-Guided Query Refinement Network (CGQR-Net) for boundary-aware cardiac ultrasound segmentation. The framework integrates multi-resolution feature representations with contour-derived structural priors. An HRNet backbone preserves high-resolution spatial details while capturing multi-scale context. A coarse segmentation is first generated, from which anatomical contours are extracted and encoded into learnable query embeddings. These contour-guided queries interact with fused feature maps via cross-attention, enabling structure-aware refinement that improves boundary delineation and reduces noise artifacts. A dual-head supervision strategy jointly optimizes segmentation and boundary prediction to enforce structural consistency. The proposed method is evaluated on the CAMUS dataset and further validated on the CardiacNet dataset to assess cross-dataset generalization. Experimental results demonstrate improved segmentation accuracy, enhanced boundary precision, and robust performance across varying imaging conditions. These results highlight the effectiveness of integrating contour-level structural information with feature-level representations for reliable cardiac ultrasound segmentation.
Zahid Ullah, Sieun Choi, Jihie Kim
Department of Computer Science and Artificial Intelligence, Dongguk University, Seoul 04620, Republic of Korea
Accurate delineation of electrocardiogram (ECG), the segmentation of meaningful waveform features, is crucial for cardiovascular diagnostics. However, the scarcity of annotated data poses a significant challenge for training deep learning models. Conventional semi-supervised semantic segmentation (SemiSeg) methods primarily focus on consistency from unlabeled data, underutilizing the information exchange possible between labeled and unlabeled sets. To address this, we introduce CardioMix, a framework built on a bidirectional CutMix strategy guided by cardiac patterns for ECG segmentation. This approach enriches the labeled set with realistic variations from unlabeled data while simultaneously applying stronger supervisory signals to the unlabeled set, as the cardiac pattern-guided mixing ensures all augmented samples remain physiologically meaningful. Our framework is designed as a plug-and-play module, demonstrating high compatibility with various SemiSeg algorithms. Extensive experiments on SemiSegECG, a public multi-dataset benchmark for ECG delineation, demonstrate that CardioMix consistently outperforms existing CutMix-based fusion strategies across diverse datasets and labeled ratios as a plug-and-play module compatible with various SemiSeg algorithms.