The mean SNRs of DKI images at low b values were higher than that of high b values (14.2 ± 2.1 vs. 9.1 ± 1, p < 0.05). The scanning time of low b values DKI was shorter than that of high b values DKI (8.6 ± 1.6 mins vs.17.6 ± 2.3 mins, p < 0.05). The MD values of ccRCCs were higher than those of non-ccRCCs at low and high b values (p < 0.05), MK, KA and RK values of ccRCCs were lower than those of non-ccRCCs at low and high b values (p < 0.05), while comparable FA values were found between ccRCCs and non-ccRCCs at low and highb values (p > 0.05). ROC curve analyses showed that MD values at high b values had the highest diagnostic efficacy in differentiating ccRCC from non-ccRCC. For pairwise comparisons of ROC curves and diagnostic efficacy, DKI at low b values was worse than high b values analysis (p < 0.05).
DKI plays an essential role for the study of RCC and these parameters are used as an important image marker for the presence of altered kidney tissue[7]. However, an optimal b-value [8, 9] has not been found to lead to the detection of high-risk carcinomas concerning kidney tissue. In this work, high b-value DKI showed a superior diagnostic efficacy to low b-value DKI in differentiating ccRCC from non-ccRCC except for SNR and scanning time. Raising the b-value increases the degree of diffusion weighting[10], which increases the contrast between tissues with different diffusion coefficients while also decreasing the overall signal intensity and SNR[11].
On account of this signal decay, low b-value DW-MRI becomes less qualitative and more quantitative, since it must be based on complex ADC calculations[12]. Therefore, as low b-value DW-MRI does not facilitate qualitative detection of malignancies which may adversely affect diagnostic accuracy[13, 14]. However, as the b-value increases, the SNR decreases[15], so the optimal high b-value may depend on the strength of the magnetic field, the software, and the manufacturer [16]. However, the acquisition of high b-values in clinical practice is complicated by technical problems and long scan times [17]. Therefore, there is currently no widely accepted optimal “high b-value” beyond the requirement for a set of DWI images with a b-value ≥ 1400s/mm2. However, DWI with higher b-value yields more significant suppression of benign tissue and thus potentially better tumor visualization. Direct acquisition of such b-values is technically challenging due to issues related to the reduced SNR: noise and an increase in anatomical distortion and resulting artifacts [18]. There also have been numerous studies that have examined the effect of varying b-values on the diagnostic accuracy of detecting malignant lesions in other abdominal tissues[19]. For example, Wu et al. analyzed DW-MRI in combination with conventional MRI and found that a b-value of 1500 s/mm2 significantly improved the specificity, but not the sensitivity, in diagnosing upper urinary tract cancer compared to a b-value of 500 s/mm2 [20]. Our study exclusively focused on the effects of low and high b-values on the diagnostic accuracy of detecting RCC tumors in the kidney. Doganay et al[14] and Erbay et al [21] also demonstrated that detection of malignant renal tumors improves at b-values of greater than 600 s/mm2. These studies indicate that varying b-values can significantly affect the diagnostic accuracy of DW-MRI's detection of malignant lesions.
Retest reliability was measured by an independent repeat assessment by two observers with 5 and 10 years’ experience and shown to be excellent. In addition, there was no statistically significant difference in retest reliability between the two observers, suggesting that the stability of DKI evaluation of microstructural differences in RCC is not affected by work experience which is conducive to the clinical popularization and application of DKI technology.
We acknowledge some limitations to the current study. First, the number of included studies was relatively small, especially for non-ccRCC. Recommend further studies with larger population to validate the results of our study. Second, ROIs were selected in the most solid portion of each tumor (approximately 2/3 of the tumor), rather than the entire renal tumor, predisposing to some selection bias due to tissue heterogeneity. Third, our study was based on single-center, retrospective study, the reliability of the results should be confirmed by prospective studies and multicenter study. Fourth, in the current study, hand-drawn ROIs used for each group were not absolutely identical, which may lead to a sampling bias. Finally, we did not evaluate comparison with benign renal tumors. Therefore, it may be too exploratory to apply our findings to the differentiation of RCC from benign renal tumors.