S <math xmlns="http://www.w3.org/1998/Math/MathML"><mmultiscripts><mrow></mrow> <mrow></mrow> <mn>3</mn></mmultiscripts> </math> TU-Net: Structured convolution and superpixel transformer for lung nodule segmentation.
Authors
Affiliations (5)
Affiliations (5)
- The College of Electronic and Electrical Engineering, Shanghai University of Engineering Science, Shanghai, 201600, China.
- The College of Electronic and Electrical Engineering, Shanghai University of Engineering Science, Shanghai, 201600, China. [email protected].
- The Department of Medical Imaging, Shanghai Electric Power Hospital, Shanghai, 200000, China.
- The Department of Cloud Network, Shanghai IDEAL INFORMATION Industry Co. LTD, Shanghai, 200000, China.
- The Department of Molecular and Cellular Biology, Center for Brain Science, Harvard University, Cambridge, MA, USA.
Abstract
Accurate segmentation of lung adenocarcinoma nodules in computed tomography (CT) images is critical for clinical staging and diagnosis. However, irregular nodule shapes and ambiguous boundaries pose significant challenges for existing methods. This study introduces S<sup>3</sup>TU-Net, a hybrid CNN-Transformer architecture designed to enhance feature extraction, fusion, and global context modeling. The model integrates three key innovations: (1) structured convolution blocks (DWF-Conv/D<sup>2</sup>BR-Conv) for multi-scale feature extraction and overfitting mitigation; (2) S<sup>2</sup>-MLP Link, a spatial-shift-enhanced skip-connection module to improve multi-level feature fusion; and 3) residual-based superpixel vision transformer (RM-SViT) to capture long-range dependencies efficiently. Evaluated on the LIDC-IDRI dataset, S<sup>3</sup>TU-Net achieves a Dice score of 89.04%, precision of 90.73%, and IoU of 90.70%, outperforming recent methods by 4.52% in Dice. Validation on the EPDB dataset further confirms its generalizability (Dice, 86.40%). This work contributes to bridging the gap between local feature sensitivity and global context awareness by integrating structured convolutions and superpixel-based transformers, offering a robust tool for clinical decision support.