Preoperative MR imaging is very important for oral tongue cancer evaluation. MRI could help define tumor extent and detect cervical metastasis due to its excellent soft-tissue contrast. Preoperative tumor TNM or histological grade evaluation is still with great challenge. Non-specific signal characteristics were seen for OTSCC on preoperative MRI (hypointense in T1WI, hyperintense in T2WI, and significantly enhanced in CE-T1WI). Numerous studies working on tumor thickness and volume have found out tumor thickness on MRI had high coincidence with that on histopathology [
31,
32]. Meanwhile, tumor thickness on MRI was a great predictive parameter for occult lymph node metastasis [
10‐
12]. Although methods for tumor thickness measurement varied, three-dimensional measurement of tumor thickness was important for oral cancer treatment. In our study, we accessed the tumor thickness by counting slice numbers where the lesion was visible and then multiplying the slice numbers by the slice thickness. We applied this method instead of measurement on sagittal images since marginal tumor with the diameter less than 1 cm might be hard to access in sagittal view. In our study, no statistic difference was shown between low- and high-grade group either in tumor thickness or in tumor volume.
In this study, we applied SWI in OTSCC to facilitate its preoperative prediction of tumor histological grade. SWI was a fairly new technique having advantages in tumor diagnosis and grading. It could maximize the sensitivity to susceptibility effects by combining a high-resolution, fully flow-compensated, 3D gradient-echo sequence with filtered phase information to enhance the contrast in magnitude images [
17]. SWI can noninvasively visualize more internal characteristics in tumors and therefore to improve the visibility of tumors and be helpful for depicting hemorrhage, calcification and increased vascularity, which may reflect tumor grade [
27,
29]. ITSS score was a semi-quantitative measurement on SWI. Our results showed mean ITSS score was greatly higher in high-grade tumors than that in low-grade OTSCCs. Value of ITSSs and ITSS score in grading gliomas has been described in many studies [
19,
22‐
24], linear or dot-like ITSSs were more frequently observed in high-grade gliomas than in low-grade ones, so that high-grade gliomas got higher ITSS scores. Considering effect of tumor size on ITSS score, we defined ITSS ratio in this study. It was the ratio of ITSSs to lesion involving area, representing for the ITSSs in unit area. Univariate ROC analysis showed among the six selected parameters, three ITSS related parameters (ITSSs, ITSS score, ITSS ratio) performed well in predicting tumor histological grade while tumor thickness and tumor volume did not. ITSSs did the best performance in predicting tumor histological grade, with the area under ROC curve of 0.79. To further access the combined effect of multi-parameters, several multi-parametric models have been tried. Our data demonstrated that although tumor thickness or tumor volume was not strong enough to be a predictive parameter, the combination use with ITSSs exactly improve the predictive capability of tumor histological grade than using ITSSs only. Multi-parametric model using combination of ITSSs and tumor thickness yielded an area under ROC curve of 0.84(0.69–0.99) and allowed great differentiation between different histological groups. Similar results have been reported in recent study working on imaging parameters from SWI, diffusion and perfusion to improve differential diagnosis primary central nervous system lymphomas (PCNSLs) and glioblastomas [
25]. ITSSs only yielded an area under ROC curve of 0.753 for differentiation PCNSLS from glioblastomas, while the combination use of ITSSs and ADC mean value or ITSSs and relative regional blood volume (rCBV) mean value would greatly increase the probability for differentiation, and the three-parameter model containing ITSSs, ADC mean value and rCBV performed best. Results form both studies presented superiority of SWI in depiction of intratumoral characteristics. Increased vascularity usually implied a higher tumor grade, as neovascularity was a reflection of fast tumor growth. Malignant tumors usually relied on newly developed vasculature in order to survive or sustain high rates of proliferation, and they are generally associated with increased microvessel density, high vascularity, and cellularity [
33]. Semi-quantitative SWI parameters were valuable in tumor differentiation and tumor histological grade prediction, but regarding multiple influence factors such as susceptibility effect and image quality of SWI images, multi-parametric models containing SWI and other parameters would improve accuracy in the area of tumor grade prediction and differentiation.
To further understand the effects of CE-SWI, five of 30 subjects were performed on CE-SWI. Similar to many other studies [
34‐
36], tumors exhibited obvious contrast enhancement on CE-SWI along with ITSSs clearer post contrast agent administration. Admittedly, CE-SWI could help to distinguish hemorrhage from abnormal venous vasculature since blood vessels would change signal intensities whereas hemorrhage would not [
37,
38]. However, in our limited subjects with CE-SWI, none of the signal changes were observed in this way and therefore requiring studies with a larger population to be conducted.
Our study had several limitations. One major limitation is that only single axial slice used per patient for SWI analysis may lead to selection bias. In addition, due to small numbers of subjects with high-grade SCCs in our study, for statistic purpose, SCCs were only sub-grouped subjects with SCCs into high- and low-grade groups. Regarding SWI technique, it was noted that paramagnetic object would be enlarged due to “amplified blooming effect” [
39], and ITSS might be affected to some degree. Therefore, quantitative method like quantitative susceptibility measurement might be a potential choice for this application. However, our study is the first report of SWI application in OTSCC. Additionally, we did no correlation for scales when using MRI parameters to predict historical grade. Tumor thickness maybe underestimated by measurement of multiplying the number of axial T2WI slices by the slice thickness.