Dual contextual learning for semi-supervised medical image classification - Takeaways - MDSpire

Dual contextual learning for semi-supervised medical image classification

  • By

  • Jiaying Liu

  • Chengyang Li

  • Sangsha Fang

  • Fangfang Deng

  • Qinghu He

  • Miao Cao

  • May 20, 2026

  • 0 min

Share

  • 1

    The Hierarchical Semantic Calibration (HSC) framework enhances pseudo-labeling reliability in semi-supervised medical image classification.

  • 2

    HSC utilizes local semantic neighborhood alignment to reduce labeling errors by enforcing consistency among similar pathological samples.

  • 3

    Global cluster-prototype calibration in HSC ensures consistent class-level representations across different augmented views.

  • 4

    Neighborhood-prototype consistency regularization in HSC adapts the alignment between local neighborhoods and global prototypes.

  • 5

    HSC achieves significant accuracy improvements, outperforming state-of-the-art methods with minimal labeled data in medical imaging.

Original Source(s)

Dual contextual learning for semi-supervised medical image classification

  • by Jiaying Liu, Chengyang Li, Sangsha Fang, Fangfang Deng, Qinghu He, Miao Cao

  • May 20, 2026

Related Content

A multidisciplinary RNA-guided approach to complement genomic analysis of unsolved patients with an inborn error of immunity

Dynamic U-shaped convolutional network for mouse cardiac image segmentation and quantification

Case Report: Synovial sarcoma with diffuse myxoid stroma and complete absence of epithelial differentiation in the extremity