Diagnostic performance of multi-parametric MRI to differentiate benign sex cord stromal tumors from malignant (non-stromal and stromal) testicular neoplasms.

Abstract

Testicular stromal tumors are uncommon, although mostly benign. The purpose of this study is to assess the role of multi-parametric MRI in differentiating benign testicular stromal tumors from malignant testicular neoplasms (non-stromal and stromal). A single-center retrospective study comparing benign stromal tumors (STs) to malignant testicular neoplasms (MTNs) was conducted. MR imaging assessment included tumor size, T2- and T1-weighted signal intensity, T2- and T1-weighted texture pattern, diffusion restriction, presence of hemorrhage and/or necrosis, and measurement of apparent diffusion coefficient and dynamic contrast enhancement (DCE). Inter-observer agreement was assessed using Cohen's kappa and Bland-Altman and data were compared using independent t-tests or χ2. Receiver operating characteristic curve analysis was used to test models incorporating various imaging features. Radical orchiectomy and histopathology revealed 20 testicular neoplasms: seven STs (35%) and thirteen MTNs (65%). MTNs were significantly larger in size than STs (5.1 ± 2.36cm vs. 1.27 ± 0.56cm; p-value < 0.001). STs demonstrated more hypointense T2W signal (85.7% vs. 46.2%; p-value < 0.09), less T2W heterogeneous texture (14.3% vs. 61.5%; p-value < 0.04), and less diffusion restriction (16.7% vs. 83.3%; p-value < 0.01) in comparison to MTNs. STs demonstrated mainly homogenous post-contrast enhancement pattern (71.4% vs. 7.7%; p-value < 0.004), while MTNs showed primarily heterogeneous enhancement pattern (77% vs. 14.3%; p-value < 0.02). STs revealed greater corrected venous phase enhancement (STs: 0.59 ± 0.29; MTNs: 0.25 ± 0.25; p-value < 0.03). STs showed higher ADC values, though the difference was not statistically significant (p-value < 0.25). A model combining T2W, DWI, and DCE features showed the best overall diagnostic accuracy with area under ROC curve of 0.99 and confidence interval ranging from 0.94 to 1. Multi-parametric MRI can potentially differentiate benign stromal tumors from malignant testicular neoplasms, which can help to avoid radical orchiectomy. However, future studies using larger sample sizes are needed to validate our results.