INTRODUCTION

The tumor stage and grade of bladder cancer are primarily associated with its clinical treatment and prognosis.^[1] Therefore, it is thus essential to accurately diagnose tumor stage and histological grade of bladder cancer in the clinic.
Intravoxel-incoherent-motion diffusion-weighted-imaging (IVIM-DWI) can reflect both perfusion and diffusion-related effects in tumors, thus accurately characterizing the molecule diffusion pattern of tumors without contrast agent injection.^[2] Meanwhile, diffusion kurtosis imaging (DKI), developed based on an advanced diffusion model, is able to provide additional information regarding the deviation of water diffusion away from Gaussian distribution, which is related to the irregularity and heterogeneity of cellular microstructures.^[3] Several previous studies have separately used IVIM or DKI in evaluating the diagnostic values in grading and staging bladder cancers.^[4,5] Although good results were obtained, distinctive physiological and pathological features of bladder cancer, including tumor cellularity, heterogeneity, and perfusion, cannot be comprehensively evaluated with either single technique. The combination of both techniques might hold more clinical potential in differentiating bladder cancers but has not been studied so far.
Therefore, the purpose of this study was to investigate if combined diffusion parameters from IVIM and DKI models were feasible in grading and staging bladder cancers.

METHODS

Subjects
42 patients (median age, 65 years; range, 21-80 years) known or suspected to have bladder cancers were recruited in this study. Certain exclusion criteria were defined as follows: (1) chemotherapy performed before MRI examination (2) lesion smaller than 5 mm (3) patient motion or inadequate image quality.
MRI acquisition
All MR experiments were performed at a 3.0T-MR system (Discovery 750w, GE Healthcare, USA) with a 16-channel coil employed. IVIM and DKI imaging were performed on the bladder after acquiring the corresponding T2-weighted anatomical images.
For IVIM, a single shot spin-echo echo-planar imaging sequence was applied in axial view with 11 b-values (0, 10, 25, 50, 75, 100, 150, 200, 400, 800, and 1000s/mm2). TR/TE，4000/80.4ms; field-of-view, 32cm×32cm; matrix, 128×128; slice thickness, 4mm; scan time, 2mins 46s.
For DKI, a separate single-shot spin-echo echo-planar imaging sequence with 6 b-values (0, 400, 800, 1200, 1600, and 2000s/mm²) with 15 diffusion directions per b value was applied. Other scan parameters were TR/TE，6000/126ms; field-of-view, 32cm×32cm; matrix, 128×128; slice thickness, 4mm; scan time, 7mins 43s.
Data analysis
IVIM and DKI data were separately analyzed using IVIM and DKI post-processing software developed in the GE ADW4.6 workstation. The IVIM related parameters including true diffusion coefficient (D), perfusion related pseudo-diffusion coefficient (D*) and perfusion fraction (f), and DKI parameters of mean diffusivity (MD) and mean kurtosis (MK), were obtained accordingly.
Two senior radiologists were employed to place regions of interest (ROI)s on the slice showing the largest cross-section section of the tumor that was greater than 5mm. All parameter values of ROIs were measured three times and the corresponding mean values were obtained for further data analysis.
SPSS 22.0 statistic software was used for statistic analyses. Intra-class correlation coefficients analysis (ICC) was applied to assess the inter-observer agreement of parameter measurements over two radiologists. The Mann-Whitney U test was used for comparing each parameter between high-grade and low-grade, and between non–muscle-invasive and muscle-invasive bladder cancers. Receiver operating characteristic (ROC) analyses were further performed to separately evaluate the diagnostic performances of IVIM and DKI parameters with significant differences, and the combined model with these parameters was included. The significance threshold was set as P<0.05.

RESULTS

In total 12 low-grade and 18 high-grade patients, and 17 non–muscle-invasive and 13 muscle-invasive patients were included for data analysis.
Excellent inter-observer agreements of diffusion parameter measurements were confirmed by high ICCs, ranging between 0.879 and 0.956.
D and MD were significantly higher in the non-muscle-invasive than muscle-invasive bladder cancers (1.18±0.27×10⁻³mm²/s vs 0.82±0.33×10⁻³mm²/s, 2.03±0.89×10⁻³mm²/s vs 1.35±0.69×10⁻³mm²/s; P<0.05)(Tab.1, Fig.1) and in low-grade than high-grade ones (1.33±0.23×10⁻³mm² /s vs 0.83±0.29×10⁻³mm²/s，2.45±0.97×10⁻³mm²/s vs 1.30±0.35×10⁻³mm²/s； P<0.05)(Tab.2, Fig.2). For MK, an opposite pattern was found between non–muscle-invasive and muscle-invasive bladder cancers (0.57±0.25vs0.90±0.21, P<0.05) and between low-grade bladder than high-grade bladder cancers (0.47±0.16vs0.84±0.26, P<0.05). Comparable D* and f values were observed between each two groups (P>0.05).
The areas under the ROC curve of D, MD, and MK were 0.838, 0.850, and 0.829 in staging bladder cancers(Fig.3a); and 0.911, 0.875, and 0.933 in grading bladder cancers, respectively(Fig.3b). The AUCs of the combined D, MD and MK were improved to 0.949 in staging and 0.938 in grading bladder cancers.

DISCUSSION

Our results demonstrated that the diffusion parameters (D, MD, MK) were significantly different in bladder cancers with various stages and grades. This may partly be explained by their histopathological characteristics. As the pathological grade and stage increased, the number of tumor cells increased, tissue disturbance increased and extracellular space curvature decreased, contributing to the altered diffusion behaviors of water molecules.

CONCLUSION

In conclusion, combined IVIM and DKI related parameters may be effective imaging biomarkers for predicting bladder cancer stage and grade.