A Thyroid Nodule Differentiation Model for Benign-Malignant Identification by Fusing Transfer Learning and Gradient Boosting Decision Tree.

Authors

Bo Li, Tingxue Li, Hao Ju, Dingnan Zhang, Yanzhu Zhang, Longxiao Guo

Published in

Journal of imaging informatics in medicine. Jul 29, 2026. Epub Jul 29, 2026.

Abstract

Accurate identification of benign and malignant thyroid nodules is a core link in clinical diagnosis and treatment decision-making, which directly affects the selection of subsequent treatment plans for patients. Aiming at the problems such as the strong subjectivity of traditional ultrasound diagnosis and the insufficient generalization ability of a single artificial intelligence model in small-sample medical scenarios, this study proposes an optimized computer-aided diagnosis model on the basis of the published hybrid diagnosis framework. The model extracts high-level semantic features of ultrasound images through the ResNet50 network with transfer learning and constructs a two-stage diagnosis architecture combined with the XGBoost classifier with optimized parameters, giving full play to the advantages of deep learning in feature representation and the high-efficiency classification ability of gradient boosting decision trees. Experimental results based on a multi-center clinical ultrasound dataset show that the diagnostic accuracy of the optimized model reaches $95\%$ , and it outperforms the original hybrid model and traditional single algorithms in key indicators such as precision, recall, and F1-score. The proposed framework demonstrated promising diagnostic performance on the internal retrospective dataset. However, further validation using independent external cohorts is required before broader clinical application. It provides an objective and reliable diagnostic reference for clinicians and has important clinical value for reducing the unnecessary rate of fine-needle aspiration biopsy and improving the efficiency of early diagnosis of thyroid diseases.

PMID:
42527788
Bibliographic data and abstract were imported from PubMed on 30 Jul 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.