Construction and Validation of Two-Level CT-Based Radiomics Models Used for Thyroid Cancer Screening in the Population

Abstract

Background: With the increasing incidence of thyroid cancer, the most clinically significant emphasis on the early detection of nodules and accurate determination of malignant ones in the population. However, this diagnostic process is still a big challenge for the radiologist by scanning non-enhanced CT images. Methods: We retrospectively collected CT images of 1372 patients (357 benign, 514 malignant, and 501 healthy control cases respectively), and 396 radiomic features were extracted from the CT images of each patient. The data were randomly divided into a training set and validation set at a ratio of 7:3. The training group was employed to identify features and train the classifiers, and the validation group was independently used to evaluate classifiers’ predictive performance. Finally, two-level models were built to detect thyroid nodule and to predict benign or malignant of the nodule retrospectively. Findings: At the first level, an SVM model built to the thyroid nodule recognition presented good discrimination in both the training group (AUC, 1.00;95%CI, 1.00–1.00) and validation group (AUC, 1.00; 95% CI, 0.99–1.00). At the second level, seven different models were built to predict benign or malignant of the nodule. Random forest (RF) showed the highest robustness, and its diagnostic performance was also feasible in the validation set (AUC 0.82, 95% CI, 0.77-0.88). Interpretation: CT-based radiomic models have great value in thyroid nodule recognition and benign or malignant thyroid nodules prediction. Funding Statement: This research did not receive any funding from public agencies, commercial, or not-for-profit sectors. Declaration of Interests: None. Ethics Approval Statement: This research obtained approval from the institutional review board, and the requirement for patients’ consent was exempted.