Back to all papers

Attention-enhanced TransUNet for ultrasound liver parenchyma segmentation toward assessment of schistosomiasis-associated liver disease.

November 18, 2026pubmed logopapers

Authors

Gao L,Li Y,Sun P,Wang X,Ye F

Affiliations (4)

  • Hubei Key Laboratory of Cognitive and Affective Disorders, Institute of Biomedical Sciences, School of Medicine, Jianghan University, Wuhan, China.
  • School of Artificial Intelligence, Jianghan University, Wuhan, China.
  • Department of Ultrasonic Imaging, Wuhan Daishan Hospital, Wuhan, China.
  • State Key Laboratory of Magnetic Resonance Spectroscopy and Imaging, National Center for Magnetic Resonance in Wuhan, Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of Sciences, Wuhan, China.

Abstract

BackgroundSchistosomiasis-associated liver disease is the most common clinical form of chronic schistosomiasis. Chronic hepatic schistosomiasis is mainly manifested as liver fibrosis, cirrhosis and other liver dysfunction, and even induces liver cancer in the late stage. Ultrasound is widely used for the diagnosis of schistosomiasis-associated liver disease, grading of liver fibrosis, and assessment of of therapeutic effects.ObjectiveThe study aims to detect and segment the ultrasound liver parenchyma with high accuracy.MethodsWe propose an attention-enhanced TransUNet that improves skip-connection feature transfer by integrating axial attention with residual connections, while retaining the Vision Transformer module for global context modeling. Axial attention re-weights features along spatial axes to emphasize direction-sensitive structures such as liver boundaries, and residual refinement mitigates information degradation during multi-scale fusion. We compared the proposed method against UNet and the original TransUNet using Dice, IoU, and Sensitivity. Ablation studies further demonstrated that combining axial attention with residual connection outperformed axial attention alone.ResultsThe proposed model achieved a satisfactory performance in the delineation of the ultrasound liver parenchyma region, with a Dice coefficient of 0.965, an IoU of 0.931 and a Sensitivity value of 0.970.ConclusionThe results support the feasibility of the attention-enhanced TransUNet model as an upstream module for accurate ultrasound image analysis in schistosomiasis-associated liver disease such as fibrosis staging and cirrhosis.

Topics

Journal Article

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAISlice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.