Multi-parametric effect in predicting tumor histological grade by using susceptibility weighted magnetic resonance imaging in tongue squamous cell carcinoma.

BACKGROUND

Susceptibility weighted imaging (SWI) is helpful for depicting hemorrhage, calcification, and increased vascularity in some neoplasms, which may reflect tumor grade. In this study, we aimed to apply SWI in patients with oral tongue squamous cell carcinomas (OTSCCs) and relate multi-parametric effect to tumor histological grade prediction.

METHODS

Preoperative MR examinations were performed on a 1 .5T MRI scanner with T1-, T2- and contrast-enhanced (CE) T1-weighted imaging. In addition to routine head and neck MRI sequences, SWI was performed. Tumor thickness and volume were measured. Intratumoral susceptibility signal intensities (ITSSs), ITSS score and ITSS ratio on SWI were evaluated and recorded. Subjects were sub-grouped into low- and high-grade according to the histological findings post operation. Parameters such as tumor thickness, tumor volume and three ITSS related parameters were compared between low- and high-grade groups. ROC analysis was performed on above parameters to access the capability in predicting tumor histological grade. Different multi-parametric models were run to access multi-parametric combination effect.

RESULTS

Thirty patients with OTSCC were finally included in the study. Twenty of them were categorized as low-grade SCC and the other ten subjects were high-grade SCC according to the pathologic findings. No significant difference was seen for tumor thickness or tumor volume between two sub-groups. ITSSs were seen in 23/30 patients. Significant difference of ITSS scores between low- and high-grade OTSCCs was observed, with mean value of 0.95 ± 0.83 and 1.70 ± 0.95, respectively. Univariate ROC analysis demonstrated ITSSs, ITSS score and ITSS ratio were valuable parameters for predicting tumor histological grade and ITSSs was superior to the other two parameters, with an area under ROC curve of 0.790. Multi-parametric model using combination of ITSSs and tumor thickness would greatly improve the predictive capability in comparison with a univariate approach, yielding the area under ROC curve of 0.84(0.69,0.99). On contrast-enhanced SWI (CE-SWI), ITSSs were shown more clearly delineated in comparison with non-contrast enhanced SWI.

CONCLUSIONS

In conclusion, SWI was superior in depiction of internal characteristics of OTSCCs, which would potentially provide more diagnostic information. Multi-parametric model using combination of ITSSs and tumor thickness would be valuable in predicting tumor histological grade.