EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

LGF‐Net: A multi‐scale feature fusion network for thyroid nodule ultrasound image classification

Yao Xiao, Yan Zhuang, Wenwu Ling, Shouyu Jiang, Ke Chen, Guoliang Liao, Yuhua Xie, Yao Hou, Lin Han, Zhan Hua, Yan Luo, Jiangli Lin iD

Journal of Applied Clinical Medical Physics · 2025-08 · 26 (8)

DOI: 10.1002/acm2.70149 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Applied Clinical Medical Physics, online erschienen am 25.07.2025. Verfasst von Yao Xiao und 11 weiteren Autor:innen. Beteiligte Einrichtung: College of Biomedical Engineering, Sichuan University, Chengdu, China. Hauptthemen laut MEDLINE: Thyroid Nodule, Neural Networks, Computer, Image Processing, Computer-Assisted, Image Interpretation, Computer-Assisted. Die Arbeit wurde bisher 7-mal zitiert (Crossref) und nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Medical Innovation Capability Improvement Plan of Capital Medical University, National Natural Science Foundation of China, Natural Science Foundation of Sichuan Province.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Thyroid cancer is one of the most common cancers in clinical practice, and accurate classification of thyroid nodule ultrasound images is crucial for computer‐aided diagnosis. Models based on a convolutional neural network (CNN) or a transformer struggle to integrate local and global features, which impacts the recognition accuracy.

Purpose

Our method is designed to capture both the key local fine‐grained features and the global spatial features essential for thyroid nodule diagnosis simultaneously. It adapts to the irregular morphology of thyroid nodules, dynamically focuses on the key pixel‐level regions of thyroid nodules, and thereby improves the model's recognition accuracy and generalization ability.

Methods

The proposed multi‐scale fusion model, the local and global feature fusion network (LGF‐Net), inspired by the dual‐path mechanism of human visual diagnosis, consists of two branches: a CNN branch and a Transformer branch. The CNN branch integrates the wavelet transform and deformable convolution module (WTDCM) to enhance the model's ability to capture discriminative local features and recognize fine‐grained textures. By introducing the aggregated attention (AA) mechanism, which mimics biological vision, into the Transformer branch, spatial features are effectively captured. The adaptive feature fusion module (FFM) is then utilized to integrate the multi‐scale features of thyroid nodules, further improving classification performance.

Results

We evaluated our model on the public thyroid nodule classification dataset (TNCD) and a private clinical dataset using accuracy, recall, precision, and F1‐score. On TNCD, the model achieved 81.50%, 79.51%, 79.92%, and 79.70%, respectively. On the private dataset, it reached 91.24%, 88.90%, 90.73%, and 89.73%, respectively. These results outperformed state‐of‐the‐art methods. We also conducted ablation studies and visualization analysis to validate the model's components and interpretability.

Conclusions

The experiments demonstrate that our method improves the accuracy of thyroid nodule recognition, shows its strong generalization ability and potential for clinical application, and provides interpretability for clinicians' diagnoses.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Thyroid NoduleNeural Networks, ComputerImage Processing, Computer-AssistedImage Interpretation, Computer-AssistedDiagnosis, Computer-AssistedThyroid NeoplasmsHumansUltrasonographyAlgorithms

Stichwörter der Autor:innen

feature fusionhybrid networkthyroid nodulesultrasound image

Autor:innen und Institutionen

  • College of Biomedical Engineering, Sichuan University, Chengdu, China.
  • Department of Ultrasound, West China Hospital, Sichuan University, Chengdu, China.
  • Department of General Surgery, China-Japan Friendship Hospital, Beijing, China.

ORCID-Profile: Jiangli Lin

Förderung

  • Medical Innovation Capability Improvement Plan of Capital Medical University
  • National Natural Science Foundation of China
  • Natural Science Foundation of Sichuan Province

Zitieren

APA

Xiao, Y., Zhuang, Y., Ling, W., Jiang, S., Chen, K., Liao, G., Xie, Y., Hou, Y., Han, L., Hua, Z., Luo, Y., & Lin, J. (2025). LGF‐Net: A multi‐scale feature fusion network for thyroid nodule ultrasound image classification. Journal of Applied Clinical Medical Physics, 26(8). https://doi.org/10.1002/acm2.70149

Vancouver

Xiao Y, Zhuang Y, Ling W, Jiang S, Chen K, Liao G, et al. LGF‐Net: A multi‐scale feature fusion network for thyroid nodule ultrasound image classification. Journal of Applied Clinical Medical Physics. 2025;26(8). doi:10.1002/acm2.70149

BibTeX
@article{Xiao2025,
  author = {Xiao, Yao and Zhuang, Yan and Ling, Wenwu and Jiang, Shouyu and Chen, Ke and Liao, Guoliang and Xie, Yuhua and Hou, Yao and Han, Lin and Hua, Zhan and Luo, Yan and Lin, Jiangli},
  title = {LGF‐Net: A multi‐scale feature fusion network for thyroid nodule ultrasound image classification},
  journal = {Journal of Applied Clinical Medical Physics},
  year = {2025},
  volume = {26},
  number = {8},
  doi = {10.1002/acm2.70149}
}
Kostenlos starten →