Dual-source dual-energy computed tomography-derived quantitative parameters combined with machine learning for the differential diagnosis of benign and malignant thyroid nodules.

Abstract

This study aimed to investigate the ability of quantitative parameter-derived dual-source dual-energy computed tomography (DS-DECT) combined with machine learning to distinguish between benign and malignant thyroid nodules. Patients with thyroid nodules and pathological surgical results who underwent preoperative DS-DECT were selected. Quantitative parameter-derived DS-DECT was applied to classify benign and malignant nodules. Then, machine learning and binary logistic regression analysis models were constructed using the DS-DECT quantitative parameters to distinguish between benign and malignant nodules. The receiver operating characteristic curve was used to assess the diagnostic performance. The DeLong test was used to compare the diagnostic efficacy. One hundred and thirty patients with 139 confirmed thyroid nodules were involved in the study. The malignant group had a significantly higher iodine concentrationnodule (arterial phase) (P=0.001), normalized iodine concentration (arterial phase) (P=0.002), iodine concentration difference (P<0.001), spectral curve slope (nonenhancement) (P=0.007), spectral curve slope (arterial phase) (P=0.001), effective atomic number (nonenhancement) (P<0.001), and effective atomic number (arterial phase) (P=0.039) than the benign group. The binary logistic regression analysis model had an AUC (area under the curve) of 0.76, a sensitivity of 0.821, and a specificity of 0.667. The machine learning model had an AUC of 0.86, a sensitivity of 0.822, specificity of 0.791 in the training cohort, an AUC of 0.84, a sensitivity of 0.727, and specificity of 0.750 in the testing cohort. Multiple quantitative parameters of DS-DECT combined with machine learning could differentiate between benign and malignant thyroid nodules.