A predictive model of thyroid nodules: based on the CAD software and multi-disciplinary indicators(PDF)
《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]
- Issue:
- 2021年第1期
- Page:
- 54-60
- Research Field:
- 医学影像物理
- Publishing date:
Info
- Title:
- A predictive model of thyroid nodules: based on the CAD software and multi-disciplinary indicators
- Author(s):
- YE Fengying1; YANG Wenmin1; LI Shangqing1; SU Qichen1; YOU Jianhong2; WANG Kangjian3; LYU Guorong1; 4
- 1. Department of Ultrasound, the Second Affiliated Hospital of Fujian Medical University, Quanzhou 362000, China 2. Department of Ultrasound, Zhongshan Hospital Affiliated to Xiamen University, Xiamen 361004, China 3. Department of Ultrasound, Zhangzhou Affiliated Hospital of Fujian Medical University, Zhangzhou 363000, China 4. Collaborative Innovation Center for Maternal and Infant Health Service Application Technology, Quanzhou 362011, China
- Keywords:
- Keywords: computer-aided diagnosis thyroid nodule ultrasonography logistic regression prediction model
- PACS:
- R318；R581
- DOI:
- DOI:10.3969/j.issn.1005-202X.2021.01.010
- Abstract:
- Abstract: Objective To quantitatively analyze the ultrasonic features of thyroid nodules using computer-aided diagnosis (CAD) software, to establish a predictive model of thyroid nodules based on CAD result, clinical information and laboratory indicators, and to compare the diagnostic efficacy of the predictive model with that of sonographers of different seniority. Methods The model-building data set prospectively included 364 patients (a total of 388 nodules) who underwent thyroid surgery and preoperative ultrasound examination in three hospitals (the Second Affiliated Hospital of Fujian Medical University, Zhongshan Hospital Affiliated to Xiamen University, Zhangzhou Affiliated Hospital) from January 2019 to September 2019. The CAD analytic results, clinical information and laboratory indicators were collected. Using pathology as the gold standard, the differences of 21 related factors between benign and malignant thyroid groups was compared. Finally, 11 factors were selected for logistic regression analysis, and 6 related factors with statistical significance for the prediction of benign and malignant nodules were selected for model establishment. The model validation data set included 105 patients (a total of 105 nodules) who underwent fine needle aspiration (FNA) in the three hospitals during the same period. The predictive model and three sonographers of different seniority made their judgments of the thyroid nodules separately. Using the pathology as the gold standard, the receiver operating characteristics curve (ROC) was drawn, the area under the curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) were calculated. The diagnostic efficacy of predictive model and sonographers of different seniority was compared with each other. Result A predictive model was built as follow: Logit（p）=-5.218+2.601×（hypoechogenicity index）+1.981×（echogenic foci index）+3.079×（edge blurring index）+1.267×（aspect ratio>1）+0.614×（TSH）-0.071×（tumor maximum diameter）. The AUC of the model was 0.884, the sensitivity was 85.50%, the specificity was 81.97%, the PPV was 91.1%, and the NPV was 72.5%. The AUC, sensitivity of the model were between that of senior sonographer and intermediate sonographer, The specificity of the model were between that of junior sonographer and intermediate sonographer. Conclusion The predictive model can predict the malignance risk of thyroid nodules well, and the overall diagnostic efficacy of the model was lower than the senior sonographer, but higher than the junior and intermediate sonographers.
Last Update: 2021-01-29