GCUNet: a graph neural network-based contextual learning network for tertiary lymphoid structure semantic...

Area of Science:

• Computational pathology
• Image analysis
• Immunology

Background:

• Tertiary lymphoid structures (TLS) are crucial in immune responses and cancer.
• Accurate TLS segmentation in whole slide images (WSIs) is challenging due to image scale and the need for contextual information.
• Current patch-based segmentation methods limit performance by excluding external contextual data.

Purpose of the Study:

• To develop an advanced method for TLS semantic segmentation in WSIs.
• To improve the accuracy of identifying TLS boundaries and maturity by incorporating long-range and fine-grained contextual information.
• To introduce GCUNet, a novel graph neural network-based contextual learning network.

Main Methods:

• GCUNet progressively aggregates contextual information outside the target patch.
• A detail and context fusion block (DCFusion) integrates target details with aggregated context.
• Four TLS semantic segmentation datasets (TCGA-COAD, TCGA-LUSC, TCGA-BLCA, PUMCH-PAAD) were created, with three publicly released.

Main Results:

• GCUNet demonstrated superior performance in TLS semantic segmentation compared to state-of-the-art methods.
• An average improvement of at least 7.41% in mean F1-score (mF1) was observed over the best baseline.
• The proposed method effectively leverages contextual information for enhanced segmentation accuracy.

Conclusions:

• GCUNet offers a significant advancement in TLS semantic segmentation for WSIs.
• The method facilitates more accurate assessment of TLS and immune microenvironments in computational pathology.
• Publicly releasing datasets will promote further research in TLS semantic segmentation.