基于超声影像组学预测甲状腺结节分类研究

打开文本图片集

Predictionofthyroid nodulesclassificationbasedonultrasound radiomics

Hou Yumei

InternationalHealthanagemententer，HenanUniversityAfiatedZhengzhouYiheHospital，Zhengzhou404 China

【Abstract】 ObjectiveTo explore the application value of ultrasound radiomics in diferentiating benign andmalignant thyroid nodulesclasified byTBSRTC.MethodsAretrospectiveanalysis of145 patients with definite pathological resultsof thyroid fine needleaspiration cytology or postoperativepathological examination in our hospital from June 2O21 to December 2O23，and transcervical ultrasonographywasperformed before treatment. According to theratioof7：3，thepatients with nodules wererandomlydivided into95training setsand 5Ovalidation sets.High-throughput dataacquisition，feature extraction and dimensionalityreduction were performedonthe region of interest （ROI） on theultrasound image，including grayscale，texture，shape，edgeand other data.Then，a prediction model basedon multi-clasification logisticregression was constructed and the prediction effect of the model was evaluated bymultiple diagnosticeficiencyparameters.Finaly，theprediction model is verified.Results Fromthe1317 extracted featureparameters，12radiomics feature variables wereselected to identify the three classifications of thyroid nodules TBSRTC.The accuracy of the prediction model in the training set was 98.9 % ，the AUC of TBSRTC II was 0.991， the sensitivity was 100% ，and the specificity was 98.1% .The AUC of TBSRTCI and I was O.974，the sensitivity was 94.7% ，and the specificity was 100% . The AUC of TBSRTC IV-VI was1.000. The accuracy of the validation set was 0.940，the AUC of TBSRTC I was O.950，the sensitivity was 100% ，and the specificity was 90.0% . The AUC of TBSRTC Iand II was O.9OO，the sensitivity was 80.0% ，and the specificity was （20号 100% .The AUC of TBSRTC IV-VI was 0.975，the sensitivity was 95.0% ，and the specificity was 100% . Conclusion This study confirms the feasibilityand effectiveness of ultrasound radiomicsinidentifying TBSRTC clasification of thyroid nodules，and provides a simple，low-cost and non-invasive screening method for clinical practice

【Keywords】 Thyroid nodules;Ultrasound;RadiomicsDOI：10.19522/j.cnki.1671-5098.2025.01.005

甲状腺结节大多数是良性的，研究表明大约85%～90% 的结节是良性的，但剩下约 10%～15% 是恶性的，良、恶性甲状腺结节的鉴别对患者的治疗、预后至关重要[1]。（剩余10784字）