The aims of this study were, firstly, to provide a formula (neurogenic index) based on MR characteristics used in daily routine for predicting whether a soft tissue tumor is neurogenic or not, secondly, to test prospectively the performance of this formula, and thirdly, to compare this performance with that of radiologists experienced in MR imaging of soft tissue tumors. Retrospectively, MR images of 70 neurogenic and 70 non-neurogenic soft tissue tumors were evaluated in random order by two teams of two observers each. A neurogenic index (NI) was calculated based on those MR parameters that showed no or minor interobserver variability. Subsequently, three investigators in concert used the NI in a validation group of 15 neurogenic and 22 nonneurogenic soft tissue tumors. The same team, based on their own experience, tried to differentiate in the same validation group neurogenic from non-neurogenic soft tissue tumors. This was expressed in a subjective score (SS). Sensitivity, specificity, and predictive values were calculated. NI comprised spread (intra- or extracompartmental), distribution, fluid-fluid levels, homogeneity on T2-weighted images (WI), highest signal intensity (SI) on T1WI, lowest SI on T2WI, and delineation on T2WI. In the validation group, NI had a sensitivity of 88.6%, a specificity of 52.0%, a positive predictive value (PPV) of 54.1%, and a negative predictive value (NPV) of 84.6% for neurogenic tumors. The subjective score SS was superior and had a sensitivity of 93.3%, a specificity of 77.2%, a PPV of 73.7%, and a NPV of 94.4%. Our NI was less accurate than the SS; however, the low number of false-negative diagnoses for neurogenic tumors warrants continued efforts in development of neural networks.