The detection rate of thyroid nodule has increased significantly in recent years,although it is generally accepted that papillary thyroid carcinoma is less invasive,there are still some patients with postoperative recurrence or distant metastasis,the 2015 American Thyroid Association(ATA)guidelines clearly state that their main goals are to reduce the risk of most disease-related deaths and relapses,and to reduce the potential harm of over-treatment of patients,high-risk patients should be given appropriate treatment and monitoring.Therefore,the differential diagnosis between benign and malignant thyroid nodule is very important,but there is overlap or even identical between them,which brings great difficulties and challenges to conventional ultrasound diagnosis.There are many versions of Thyroid Imaging,Reporting and Data System(TI-RADS)classification of thyroid at home and abroad.At present,ultrasound-guided thyroid nodule biopsy is still the gold standard for preoperative diagnosis,but some nodules can not be identified,especially for nodules less than 1 cm in diameter,it is also difficult to successfully puncture or obtain accurate pathological results.The specificity of BRAFV600E gene detection is high,but the sensitivity is low.Ultrasound-based radiomics is an emerging technology that can extract and quantitatively analyze imaging features from ultrasound images in a high-throughput manner,allowing for features such as the shape,texture and wavelets of a tumor.Machine learning is used to construct class or predictive models to objectively evaluate the relationship between tumor malignancy with clinic,pathology,gene and protein,and to provide valuable information for clinicians.This article reviews the progress of multimodal ultrasound and ultrasound-based radiomics in thyroid nodule diagnosis.
NETL
NSTL
万方数据
