A study on the repeatability of radiomics parameters under repeated CT scans

Abstract Background To investigate the repeatability and diagnostic ability of radiomics parameters of lung lesions under repeated CT scan. Methods A retrospective analysis of 101 patients who underwent CT-guided percutaneous biopsy in our hospital and were pathologically confirmed as lung cancer and chronic inflammation. Two images of repeated CT scan in the puncture process were selected. The gray histogram and gray-level co-occurrence matrix of the lesion area were extracted based on ImageJ: mean gray value (mean), gray variance (SD), minimum gray value (minor), maximum gray value (max), Skewness, kurtosis, contrast, correlation, angular second moment (ASM), inverse different moment(IDM), and entropy. Intraclass correlation (ICC) was used to evaluate the repeatability of the measurement results of the two images. Results The parameters with ICC greater than 0.8 were selected, and the average values of the two measurements were taken. The independent sample t test and Mann–Whitney U test were used to compare the difference of radiomics parameters between benign and malignant lesions. Application subjects: Nine parameters with excellent consistency between the two measurements were entropy (ICC: 0.982), IDM (0.975), ASM (0.956), mean (0.902), max (0.841), SD (0.894), correlation (0.878), and minor (0.897). The parameter with good consistency was kurtosis (ICC: 0.784). The parameter with poor consistency was contrast (ICC: 0.577).There were significant differences in IDM, correlation, and entropy between the benign group and the malignant group ( P < 0.05). In ROC curve analysis, correlation had the best diagnostic effect (AUC: 0.749), followed by entropy (AUC: 0.642) and IDM (AUC: 0.617). Conclusion The radiomics parameters of quantitative calculation of CT images under repeated scanning showed good repeatability. The values of entropy, contrast, and correlation have certain clinical diagnostic value in differentiating benign and malignant lung lesions.