A u-shaped coding-decoding network-based uveitis lesion segmentation method

By employing multi-branch feature fusion, frequency domain cross-layer feature fusion, hybrid Mamba interaction bottleneck, and uncertainty-aware recursive gating module in Uveitis FFA image lesion segmentation, the problems of insufficient multi-scale lesion, boundary recognition, and long-range context utilization were solved, achieving higher accuracy and more stable lesion segmentation.

CN122336309APending Publication Date: 2026-07-03SUZHOU UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610797294.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-06-04
Publication Date
2026-07-03

AI Technical Summary

Technical Problem

Existing methods for segmenting lesions in uveitis FFA images struggle to account for lesions at multiple scales, accurately extract diffuse boundaries, fail to utilize long-range context effectively, and exhibit unstable decoding in high-uncertainty regions, leading to inaccurate segmentation.

Method used

A U-shaped encoding and decoding network-based approach is adopted, which improves feature expression efficiency through a multi-branch feature fusion module, a frequency domain cross-layer feature fusion module, a hybrid Mamba interaction bottleneck module, and an uncertainty-aware recursive gating module. This approach balances the accurate localization of small lesions with the complete coverage of large lesions, ensuring boundary recognition accuracy. Furthermore, skip connections and loss function optimization are used to improve the training stability of the model under imbalanced conditions.

Benefits of technology

It improves the accuracy and stability of uveitis lesion segmentation, enabling more accurate identification and localization of retinal vascular leakage and macular edema leakage, reducing misjudgments in high-uncertainty regions, and enhancing the model's optimization performance on imbalanced datasets.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122336309A_ABST
    Figure CN122336309A_ABST
Patent Text Reader

Abstract

This invention relates to the field of image segmentation technology and discloses a method for segmenting uveitis lesions based on a U-shaped encoder-decoder network. First, a multi-branch feature fusion module is constructed as a basic unit, stacked to form a multi-scale feature enhancement encoder. Second, a frequency domain cross-layer feature fusion module is designed in the skip connections, utilizing Haar wavelet decomposition to extract boundary information and combining spatial and channel attention to filter and fuse features. A hybrid Mamba interaction bottleneck module is set at the top of the encoder to align and aggregate multi-level deep features, obtain global context through visual Mamba, and perform mutual correction of global and local features using a joint attention map. In the decoding stage, an uncertainty-aware recursive gating module is introduced, using the Shannon entropy output of the prediction head as the uncertainty map to dynamically modulate the initial gating generated by the gating network, achieving adaptive control of the intensity of shallow detail introduction.
Need to check novelty before this filing date? Find Prior Art