Table 1.
Demographic, clinical and ultrasound characteristics of 913 patients.
Fig 1.
(a) Flow diagram of feature selection and validation of clinical data. Cohort 1 comprised 771 samples that were randomly divided into the training (539 samples) and test (232 samples) sets. Two additional independent data sets, Cohorts 2 and 3, included 70 and 72 samples, respectively, from Renji and Xinhua Hospital and were also employed as test sets to validate the predictive accuracy of the classification based on clinical data. (b) Flow diagram for the comparison between the classifier models based on the three gene expression levels, the clinical information, and integrating the gene expression with clinical data.
Table 2.
Comparison of the characteristics of benign and malignant tumors in the training set following histologic classification of thyroid nodules.
Table 3.
Predictive performance of three independent data sets using the clinical information model.
Fig 2.
Histogram of the relative gene expression levels of DPP4, SCG5 and CA12 in malignant and benign thyroid nodules.
**P<0.01 by two-tailed t test between benign and malignant thyroid tumor types. *P<0.05 by two-tailed t test between benign and malignant thyroid tumor types.
Table 4.
Comparison of thyroid cancer predictive performance based on the gene expression, clinical data, or integrated model.