Differential diagnosis of benign and malignant liver tumors in children using the method of quantitative assessment of multiparametric magnetic resonance imaging data

Abstract

Introduction. Liver tumors account for 1.1 % of all newly diagnosed neoplasms in children. The rarity of this pathology causes difficulties in differential diagnosis. Currently, magnetic resonance imaging (MRI) is the main and most promising method for diagnosing liver diseases. In our work, we decided to quantify the data from this study.Purpose of the study – determination of the possibilities of quantitative assessment of multiparametric MRI data in the differential diagnosis of benign and malignant liver tumors in children.Material and methods. 133 patients with 307 liver lesions aged from 5 months to 20 years were examined. All patients underwent MRI on high-field MRI machines using an extracellular contrast agent, which included T2 weighted images with and without suppression of the signal from adipose tissue, diffusion-weighted images with automatic calculation of maps of the apparent diffusion coefficient (ADC), T1 weighted images with suppression of the signal from adipose tissue before and after the introduction of a contrast agent (in the arterial, portal, venous and delayed phases). Quantitative characteristics of changes in signal intensity in the lesion, intact liver parenchyma, spleen, kidney, aorta, and inferior vena cava (IVC) were obtained. To level the influence of external factors, we used not the absolute values of the signal intensity, but the ratios: lesion/intact liver parenchyma, lesion/kidney, lesion/aorta, lesion/spleen, lesion/IVC. For each lesion, 5 coefficients were calculated in each of the sequences, with the exception of patients (n = 4) after splenectomy, in whom 4 coefficients were calculated. In addition, for images obtained after the injection of a contrast agent, the ratios of the signal on post-contrast images to the native phase were calculated. Quantitative parameters such as the maximum size of the tumor, its volume and the age of the patient were included in the calculation. Tumors were represented by benign (n = 139) and malignant (n = 169) formations. The diagnosis of all malignant neoplasms and some benign ones was confirmed morphologically, benign ones – using MRI with intravenous contrast and dynamic observation.Results. A mathematical model was built:A = 1/(1+e-Z),where Z = 6,25019 + 1,03132 × S + 1,30077 × P2le/li – 0,00459 × DCle + 4,01375 × P1le/a – 2,05533 × Part le/li – 2,55823 × Pport le/k + 7,56980 × Pdel5 le/k – 15,91047 × Pdel5 le/a.The model is informative and statistically significant (p < 0.001). If A > 0.5, it should be considered that the studied focus is of a malignant nature, if A ≤ 0.5, the formation is benign. Model sensitivity and specificity were, respectively, 0.947 and 0.917.Conclusion. The mathematical model makes it possible to differentiate between malignant and benign formations with a high degree of informativeness, which is a priority task in detecting a mass formation in the liver.