Introduction

MRI plays an important role in gynecological imaging and DWI also gives clinically useful information. ADC derived from DWI would be one of promising biomarkers not only in assessing tumor aggressiveness but also predicting prognosis. However, no consensus has been reached regarding which b-values should be utilized to calculate ADC and which histogram metrics related to ADC should be used. Indeed, conflicted results were reported about the relationship between lesion aggressiveness and ADC skewness and kurtosis.^1,2 A clinical biomarker also should satisfy the followings; easy-to-use, reliable (repeatable and reproducible) and clinically useful. The present study was performed to clarify which b-value combination to calculate ADC and which ADC histogram metrics are appropriate as a clinical biomarker with respect to data repeatability and a correlation with histological grading of endometrioid adenocarcinoma.

Methods

Consecutive 50 patients underwent pelvic MRI using whole-body clinical system (1.5T and 3T of Ingenia, Philips Medical Systems, Best, the Netherlands) to evaluate uterine endometrial lesions in our institute from July 2017 to July 2018. All patients gave informed consent to participate in the study. Of them, 9 patients were excluded (no histology in 2, neoadjuvant chemotherapy in 2, metal artifact in one, difficulty in assigning ROI in 3 and sequence error in one). Finally, 41 patients with 19 endometrioid adenocarcinoma, one carcinosarcoma, one clear cell adenocarcinoma and 20 benign endometrial lesion or normal endometrium were analyzed. Two sequential free-breathing single shot diffusion-weighted spin-echo echo-planar imaging with b-values of 0, 100, 1000 and 1500 s/mm² and identical positioning were obtained during examination. The ROIs were assigned at suspected endometrial lesions on the 1st DWI dataset and pasted them onto the 2nd dataset holding their size and location. Each voxel ADC was calculated by fitting signal intensity of each voxel into mono-exponential curve. ADCs calculated from several combinations of b-values were expresses as follows: ADC (0, 1000), ADC (0, 1500), ADC (100, 1000), ADC (100, 1500), ADC (100, 1000, 1500), ADC (0, 100, 1000, 1500) and ADC (1000, 1500) were calculated from b-values of 0 and 1000 s/mm², 0 and 1500 s/mm², 100 and 1000 s/mm², 100 and 1500 s/mm², 100, 1000 and 1500 s/mm², 0, 100, 1000 and 1500 s/mm² and 1000 and 1500 s/mm², respectively. The repeatability of ADC histogram metrics of minimum, 10%, 25%, median, mean, 75% 90%, maximum, skewness, kurtosis and entropy within ROI was evaluated using intraclass correlation coefficient (ICC) and Bland-Altman plot (%). The correlation between ADC and grading of endometrioid adenocarcinoma was also analyzed with Spearman’s rank-order correlation with setting benign lesion and normal endometrium as G0

Results

The mean of ADC (0, 1500) and ADC(0, 100, 1000, 1500) showed highest repeatability (ICC of 0.981 and 0.983 for ADC (0, 1500) and ADC (0, 100, 1000, 1500), respectively and standard deviation of bias in Bland-Altman plot (%) of 6.21 and 6.59 for ADC (0, 1500) and ADC (0, 100, 1000, 1500), respectively) and a significant correlation with grading (ρ of -0.57 and p<0.001 for ADC (0, 1500) and ρ of -0.56 and p<0.001 for ADC (0, 100, 1000, 1500), respectively). The skewness, kurtosis and entropy showed low repeatability and no significant correlation with grading.

Conclusion

ADC mean based on a b-value combination of 0 and 1500 s/mm² show high repeatability and a significant correlation with histological grading in uterine endometrial adenocarcinoma and would be appropriate as a clinical biomarker.

Acknowledgements

No acknowledgement found.