229P - The application of diffusion kurtosis imaging in predicting muscle invasion of bladder cancer: A comparison with conventional DWI

Background

Differences in treatment paradigms and prognosis based on tumor aggressiveness emphasized the urgent need to accurately detect muscle invasion of bladder cancer. DWI, based on the hypothesis that diffusion follows by Gaussian distribution, has shown great promise in differentiating MIBC and NMIBC. DKI is believed to better reflect the deviation from a Gaussian distribution due to irregularity and heterogeneity of cell microstructure and tissue components. Therefore, the present study was designed to quantitatively investigate the diagnostic performances of DKI in predicting aggressiveness of bladder cancer and to compare the potential of parameters obtained from DWI and DKI.

Methods

Multiple b value DWIs were performed using a 3.0-T magnetic resonance imaging unit in 61 patients with bladder cancer including MIBC and NMIBC confirmed by histopathological findings. DWI data were postprocessed using mono-exponential and DKI models to calculate the apparent diffusion coefficient, apparent diffusional kurtosis, and kurtosis-corrected diffusion coefficient. Receiver-operating characteristic analysis was performed to compare the diagnostic efficacy of all diffusion parameters. SPSS and MedCalc were used to perform the statistical analyses. ADC and DKI parameters were measured and processed using IMAge/enGINE MR_Diffusion (an open-source software).

Results

Both ADC and DKI values differed significantly between MIBC and NMIBC. The ADC and D_app values of the MIBC group were lower than the NMIBC group (all p<0.001), whilst K_app values were higher than those of the NMIBC group (p<0.001). The AUC values of ADC, D_app and K_app were 0.833, 0.859, and 0.920 for differentiating MIBC from NMIBC, respectively. The combination of D_app and K_app value had the highest AUC of 0.944. For pairwise comparisons of ROC curves, ADC was worse than K_app (p=0.030), and worse than the combination of D_app and K_app (p=0.007). ADC was not significantly different from D_app (p=0.528).

Conclusions

Both conventional DWI and DKI models are beneficial in differentiating between MIBC and NMIBC, whilst K_app and the combination of D_app and K_app could produce more robust values than conventional ADC in evaluating aggressiveness of bladder cancer.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

Zhongshan Hospital Affiliated to Fudan University.

Funding

Natural Science Foundation of China.

Disclosure

All authors have declared no conflicts of interest.