To determine the effect of region of interest (ROI) on tumor’s apparent diffusion coefficient (ADC) and interobserver variability in thyroid nodules. Thirty-three individuals with 45 pathologically-confirmed thyroid nodules were assessed by preoperative diffusion-weighted imaging (DWI) with b values of 0 and 400 s/mm2, respectively. Two readers evaluated the ADC values of lesions based on three ROI techniques: whole-volume, single-slice and small solid-sample groups. Interobserver variability was analyzed for all ROI techniques, and the mean ADCs of benign and cancerous thyroid nodules were compared. For the mean ADCs of non-cancerous thyroid nodules, average differences and limits of agreement (LOAs) between readers were 0.00 [− 0.17–0.17] × 10− 3 mm2/s for whole-volume ROI (ICC = 0.967), 0.00 [− 0.26–0.26] × 10− 3 mm2/s for single-slice ROI (ICC = 0.932) and − 0.02 [− 0.38–0.41] × 10− 3 mm2/s for small solid-sample ROI (ICC = 0.823). For the mean ADCs of cancerous thyroid nodules, average differences and LOAs between readers were − 0.05 [− 0.23–0.13] × 10− 3 mm2/s (ICC = 0.885), 0.01 [− 0.23–0.25] × 10− 3 mm2/s (ICC = 0.839) and − 0.07 [− 0.52–0.39] × 10− 3 mm2/s (ICC = 0.579) for the three ROI methods, respectively. The mean ADC values were more scattered in the small solid-sample ROI group in comparison with the whole-volume and single-slice groups, in noncancerous and cancerous specimens. Of all three ROI techniques, whole-volume ROI-determined ADC had the highest combined sensitivity (80.0%), specificity (88.3%) and Youden index (0.683), with a cut-off of 1.84 × 10− 3 mm2/s. The ROI method overtly affects ADC measurements in benign and cancerous thyroid nodules. Small solid-sample ROI yielded the worst interobserver variability of average ADC measurements.
This study aims to explore the value of a machine learning (ML) model based on radiomics features and clinical features in predicting the outcome of spontaneous supratentorial intracerebral hemorrhage (sICH) 90 days after surgery. A total of 348 patients with sICH underwent craniotomy evacuation of hematoma from three medical centers. One hundred and eight radiomics features were extracted from sICH lesions on baseline CT. Radiomics features were screened using 12 feature selection algorithms. Clinical features included age, gender, admission Glasgow Coma Scale (GCS), intraventricular hemorrhage (IVH), midline shift (MLS), and deep ICH. Nine ML models were constructed based on clinical feature, and clinical features + radiomics features, respectively. Grid search was performed on different combinations of feature selection and ML model for parameter tuning. The averaged receiver operating characteristics (ROC) area under curve (AUC) was calculated and the model with the largest AUC was selected. It was then tested using multicenter data. The combination of lasso regression feature selection and logistic regression model based on clinical features + radiomics features had the best performance (AUC: 0.87). The best model predicted an AUC of 0.85 (95%CI, 0.75–0.94) on the internal test set and 0.81 (95%CI, 0.64–0.99) and 0.83 (95%CI, 0.68–0.97) on the two external test sets, respectively. Twenty-two radiomics features were selected by lasso regression. The second-order feature gray level non-uniformity normalized was the most important radiomics feature. Age is the feature with the greatest contribution to prediction. The combination of clinical features and radiomics features using logistic regression models can improve the outcome prediction of patients with sICH 90 days after surgery.