Related Experiment Video
Updated: Mar 28, 2026
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
[Evaluation of wall configuration ultrasonogrophicin diagnosis of thyroid small nodules using binary logistic
Qiaomei Fu1, Pengxi Wu2, Yan Ding
1Department of Ultrasound Medicine, the children's Hospital of Wuxi, Wuxi 214023, China.
Objective:
To screen out the sonogram features for the differential diagnosis of benign and malignant thyroid small nodules (≤ 1.0 cm) by Logistics regression analysis, to establish the binary Logistic regression model of sonogram features as independent variable and investigate the value of wall configuration of ultrasonogrophic nodules in the differential diagnosis of benign and malignant thyroid small nodules.
Methods:
A total of 208 thyroid nodules ≤ 1.0 cm in diameter in 190 patients were evaluated. With postoperative pathological examination or fine needle aspiration biopsy, 106 nodules were confirmed as benign and 102 as malignant. Ultrasonic features of thyroid nodules were evaluated for the differential diagnosis of benign and malignant small thyroid nodules based on pathological diagnosis as a gold standard, a Logistic model was obtained, and the odds ratio of variables were compared. The margin of thyroid nodule was divided into regular or irregular margin, and the latter was divided further into four subtypes: strip, triangular, antler and papillary. The border was divided into clear, fuzzy or both. The periphery was divided into those with normal and abnormal echo;. The calcification included no calcification, microcalcification and non-microcalcification.
Results:
Four statistically significant features were obtained finally by Logistics regression analysis, including margin, border, periphery and calcification. A formula was constructed by two-variables logistic regression analysis and probability of malignancy = 1/(1 + e - z), in which z = 5.026 × margin + 4.218 × border + 4.024 × periphery + 3.892 × calcification - 15.247. The odds ratio of margin was higher than the other independent variables.
Conclusions:
Logistics regression analysis indicates that the calcification, border, periphery, and especially margin of thyroid nodules are significant features for differentiating benign and malignant thyroid nodules. The margin score was more intuitionistic for the differentialtion of benign and malignant thyroid nodules.
