Related Experiment Video
Updated: Jun 12, 2025
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
Application of microvascular ultrasound-assisted thyroid imaging report and data system in thyroid nodule risk
Guangrong Ma1,2, Libin Chen3, Yong Wang4
1Department of Ultrasound in Medicine, The Second Affiliated Hospital of Zhejiang University School of Medicine, Hangzhou, P.R. China.
Objectives:
To establish superb microvascular imaging (SMI) based thyroid imaging reporting and data system (SMI TI-RADS) for risk stratification of malignancy in thyroid nodules.
Methods:
In total, 471 patients, comprising 643 thyroid nodules, who received conventional ultrasound (US), SMI, and a final diagnosis were extensively analyzed. A qualitative assessment of US features of the nodules was performed followed by univariable and multivariable logistic regression analyses, leading to the construction of the SMI TI-RADS, which was further verified using internal and external validation cohorts.
Results:
Among the stand-alone US, predictive factors were the shape and margins of the nodules, echogenicity and echogenic foci, vascularity, extrathyroidal extension, ring-SMI patterns, penetrating vascularity, flow-signal enlarged, and vascularity area ratio. SMI TI-RADS depicted an enhanced area under the receiver operating characteristic curve (AUC) of 0.94 (95% CI: 0.92, 0.96; p < 0.001 relative to other stratification systems), a 79% biopsy yield of malignancy (BYM, 189/240 nodules), and a 21% unnecessary biopsy rate (UBR, 51/240 nodules). In the verification cohorts, we demonstrated AUCs, malignancy biopsy yields, and unnecessary biopsy rates of 0.88 (95% CI: 0.83, 0.94), 79% (59/75 nodules), and 21% (16/75 nodules) for the internal cohort, respectively, and 0.91 (95% CI: 0.85, 0.96), 72% (31/43 nodules), and 28% (12/43 nodules) for the external cohort, respectively.
Conclusion:
SMI TI-RADS was found to be superior in diagnostic sensitivity, specificity, and efficiency than existing TI-RADSs, showing better stratification of the malignancy risk, and thus decreasing the rate of unnecessary needle biopsy.
Critical Relevance Statement:
To develop an imaging and data system based on conventional US and SMI features for stratifying the malignancy risk in thyroid nodules.
Key Points:
SMI features could improve thyroid nodule risk stratification. SMI TI-RADS showed superior diagnostic efficiency and accuracy for biopsy guidance. SMI TI-RADS can provide better guidance for clinical diagnosis and treatment of thyroid nodules.
