cs.CVOct 13, 2026

Human-Inspired Asymmetric Collaboration for Dual-Student Semi-Supervised Medical Image Segmentation

Authors: Huiyi Ren, Ziming Fang, Jiaming Cai, Rui Xu, Haojie Li

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Dual-student learning enriches predictive diversity in SSMIS by introducing a strongly augmented view alongside a weakly augmented view. Insufficient strong augmentation may homogenize the students, whereas excessive augmentation can corrupt anatomical evidence and destabilize collaboration. This challenge echoes feedback-regulated human instruction, where learners receive differentiated tasks and the task difficulty is continually adjusted according to their evolving responses. Inspired by these, we propose Human-Inspired Asymmetric Collaboration (HAC), a closed-loop dual-student framework that maintains weak--strong learning through adaptive difficulty calibration and geometry preservation. Specifically, Reinforcement-guided Difficulty Calibration (RDC) formulates strong-augmentation control as a window-level reinforcement learning process, using augmentation history, realized perturbations, and training responses to dynamically adjust the augmentation magnitude, operation number, and magnitude range. Meanwhile, the Gramian Geometry Preservation Constraint (G2PC) preserves the early-stage three-branch geometry through global volume and pairwise area constraints, preventing shared supervision from homogenizing the two students. Comprehensive evaluations across LA, Pancreas, and BraTS-2019 validate both the segmentation effectiveness and capability-aware difficulty calibration of HAC.

Figures & tables

Figure 1: Motivation, core design, and performance of HAC. (a) A fixed augmentation difficulty becomes mismatched with evolving student capabilities. (b) HAC dynamically calibrates augmentation difficulty through a feedback-regulated loop. (c) HAC consistently outperforms representative dual-student methods across three benchmarks in terms of Jaccard.
Figure 2: Overview of HAC. After warm-up, RDC performs window-level updates to dynamically calibrate strong-augmentation difficulty from the evolving training state and feedback, while G2PC is applied at each iteration to preserve the target branch geometry through global volume and pairwise area constraints. For compactness, aa , dd , and tt denote AIS, DPS, and the Teacher, respectively. During inference, predictions from the two complementary students are fused.
Table 1: Results on the LA, Pancreas, and BraTS-2019 datasets under different labeled-data ratios. D , J , and H denote Dice, Jaccard, and 95HD, respectively. Higher Dice and Jaccard are better, while lower 95HD is better. Best and second-best results are shown in bold and underlined , respectively.
Figure 3: Qualitative visualization of segmentation results.
Figure 4: Training evolution of the RDC policy parameters.
Figure 5: Training dynamics of RDC and G2PC on LA with 10% labeled data. (a) Performance under fixed and adaptive augmentation. (b) Brach individuality measured by cosine distance, with and without G2PC.
Figure 6: Feature responses and CKA similarity between DPS and AIS under different fixed and adaptive augmentations on LA with 10% labeled data.
Table 8
Figure 7: Ablation study of the state components and reward terms in RDC.
Figure 10
More in this topic

Semi-Supervised Medical Image Segmentation

Semi-supervised medical image segmentation uses labeled and unlabeled scans to train models that delineate organs, lesions, and tissues with fewer annotations.

See all 49 papers

Explore similar work

CardsList

  1. Unifying Active Learning and Semi-Supervised Learning for Medical Image Segmentation

    Jul 27, 2026Bahram Jafrasteh, Cheng Wan, Heejong Kim +2Active LearningSemi-Supervised Medical Image Segmentation

    Department of Radiology, Weill Cornell Medicine, New York, NY, USA · Department of Electrical and Computer Engineering, Cornell University, Ithaca, NY, USA

  2. H2AL: Hyperbolic Hierarchy-aware Aggregative Learning for Registration-based Few-shot Medical Image Segmentation

    Aug 7, 2026Jia Wang, Jiaming Cai, Zunying Hu +4Hierarchical Representation LearningMedical Image Registration

    Dalian University of Technology Dalian, China · Beijing Children’s Hospital, Capital Medical University Beijing, China · Chongqing University of Posts and Telecommunications Chongqing, China

  3. Difficulty-Aware Sample Allocation for Adaptive Data Augmentation in Semantic Segmentation

    Aug 26, 2026Olasimbo Ayodeji Arigbabu, Abimbola Ismail ArigbabuData AugmentationSemantic Segmentation

    Independent Researcher, Germany. · Department of Science and Technology Education, Olabisi Onabanjo University, Ago-Iwoye, Ogun State, Nigeria.