Novel two-stage deep learning framework for automated pressure injury classification - Takeaways - MDSpire
Coming Soon: Introducing MDSpire News. Learn more
Conexiant’s news site is now MDSpire News. Learn more

Innovative Two-Phase Deep Learning System for Automated Classification of Pressure Injuries

  • By

  • Ting-Yu Lai

  • Yi-Jiun Chou

  • Chun-You Liu

  • Chien-Wei Chen

  • Ching-Ting Lin

  • Wei-Chun Wang

  • Yimin Hsu

  • Ming-Li Hsieh

  • Shih-Sheng Chang

  • March 27, 2026

Share

  • 1

    The study developed an AI framework for automatic pressure injury staging from raw clinical images without manual lesion localisation.

  • 2

    A two-stage deep learning approach was utilized, combining YOLOv9 for lesion detection and DenseNet161 for staging.

  • 3

    The object detection model achieved a mean average precision (mAP) of 0.796, while the staging model demonstrated an accuracy of 0.775.

  • 4

    The framework showed performance comparable to experienced nurses, improving diagnostic accuracy and reducing subjectivity in PI staging.

  • 5

    Challenges include handling intra-wound heterogeneity and variability in image quality, which may affect the system's performance.

Original Source(s)

Novel two-stage deep learning framework for automated pressure injury classification

  • by Ting-Yu Lai, Yi-Jiun Chou, Chun-You Liu, Chien-Wei Chen, Ching-Ting Lin, Wei-Chun Wang, Yimin Hsu, Ming-Li Hsieh, Shih-Sheng Chang

  • March 27, 2026

Related Content

Achieving Long-Term Success in Recurrent Rhinophyma

Case report describes recurrence-free outcomes at 18 months following treatment for triple-recurrent disease.

Medical Oddities: Prayer for Pain?

This week’s research collection showed that biology, behavior, and environment often intersect in unexpected ways.

Topical Recombinant Collagen Following Fractional Radio Frequency Microneedling for Nonsurgical Facelift: Case Report