Differentiation of malignant from benign soft tissue tumors using radiomics based on pharmacokinetic parameter maps obtained from dynamic contrast-enhanced magnetic resonance imaging data

Preoperative differentiation between benign and malignant soft tissue tumors (STTs) is vital for clinical treatment decisions. To develop and test a dynamic contrast-enhanced magnetic resonance imaging (DCE MRI)-based radiomics nomogram for differentiating STTs.

Ninety-seven patients with pathologically confirmed STTs (training set, n = 67; validation set, n = 30). The volume transfer constant (Ktrans), microvascular permeability reflux constant (Kep), and distribution volume per tissue volume unit (Ve) parameter maps were acquired using a Siemens workstation, and DCE-MRI was analyzed on a Tofts-Kety model. Radiomics features were then extracted from these parametric maps and selected by minimum redundancy maximum relevance (mRMR) rules and least absolute shrinkage and selection operator (LASSO) regression analysis. Using the selected features, four radiomics signatures were constructed (Ktrans, Kep, Ve, and Combined radiomics signature). Clinical factors, MRI morphologic features, and three quantitative parameter values underwent univariate and multivariate logistics regression. Significant risk factors and the radiomics signatures were incorporated to formulate the radiomics nomogram.

The AUCs of Ktrans, Kep, Ve, and Combined radiomics signatures were 0.651, 0.703, 0.847, and 0.809 in the validation set. And their accuracies were 69.2%, 76.9%, 86.5%, and 75.0%, respectively. The AUCs of the clinical and parameter-value models were 0.732 and 0.632 in the validation set. And their accuracies were 75% and 73%, respectively. The AUCs of the nomogram were 0.948 (training set) and 0.861 (validation set), showing good calibration and clinical utility.

The novel radiomics nomogram based on DCE-MRI parametric maps can distinguish between benign and malignant STTs accurately, efficiently, and noninvasively.