- Email: [email protected]
Diffusion Kurtosis Imaging in the Assessment of Cervical Carcinoma
ARTICLE IN PRESS
Original Investigation
Diffusion Kurtosis Imaging in the Assessment of Cervical Carcinoma Mandi Wang, MMedSci, Jose A.U. Perucho, BEng, Queenie Chan, PhD, Jianqing Sun, PhD, Philip Ip, MBChB, FRCPath, FHKCPath, FHKAM (Pathology), Ka Yu Tse, MBBS, MMedSci, FHKCOG, FHKAM (O&G), Elaine Y.P. Lee, BMedSci, BMBS, FRCR
Abbreviations DKI diffusion kurtosis imaging MD mean diffusivity MK mean kurtosis DWI diffusion-weighted imaging ADC apparent diffusion coefficient ICC intraclass correlation coefficient VOI volume of interest ROI region of interest SCC squamous cell carcinoma ACA adenocarcinoma
Rationale and Objectives: To evaluate the additional value of diffusion kurtosis imaging (DKI) in the characterization of cervical carcinoma. Materials and Methods: Seventy-five patients (56.9 § 13.4 years) with histologic-confirmed cervical carcinoma were included. Diffusion-weighted imaging (DWI) was acquired on a 3T MRI with five b values (0, 500, 800, 1000, and 1500 s/mm2). Data were analyzed based on DKI model (5 b values) and conventional DWI (0 and 1000 s/mm2). Largest single-slice region of interest (ROI) and volume of interest (VOI) were drawn around the tumor. Mean diffusivity (MD), mean kurtosis (MK), and apparent diffusion coefficient (ADC) of cervical carcinoma and normal myometrium were measured and compared. MD, MK, and ADC of cervical carcinoma were compared among histologic subtypes, tumor grades, and FIGO stages. Results: ROI- and VOI-derived DKI parameters and ADC were all in excellent consistency (intraclass correlation coefficient, ICC > 0.90, respectively). Cervical carcinoma had significantly lower MD, ADC, and higher MK than normal myometrium (p < 0.001). MD and ADC showed significant differences between histologic subtypes and FIGO stages, lower in squamous cell carcinoma than adenocarcinoma and higher in FIGO III than FIGO IIIIV (p < 0.050), but not with tumor grade. No difference was observed in MK for different clinicopathologic features tested. Conclusion: ROI and VOI analyses were in excellent consistency. MD and ADC were able to distinguish histologic subtypes and separating FIGO stages, MK could not. DKI showed no clear added value over conventional DWI in the characterization of cervical carcinoma. Key Words: Diffusion kurtosis imaging; Diffusion-weighted imaging; Cervical carcinoma; Histologic subtypes, FIGO. © 2019 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.
INTRODUCTION
D
iffusion-weighted imaging (DWI) has emerged as an important and promising tool in the assessment of cervical carcinoma (1). DWI combined with T2weighted (T2W) imaging increases the accuracy of determining parametrial invasion (2). Quantitative DWI using the apparent
Acad Radiol 2019; &:1–8 From the Department of Diagnostic Radiology, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Pok Fu Lam, Hong Kong (M.W., J.A.U.P., E.Y.P.L.); Philips Healthcare, Greater China, China (Q.C.); Philips Healthcare, Shanghai, China (J.S.); Department of Pathology, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong (P.I.); Department of Obstetrics and Gynaecology, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong (K.Y.T.). Received June 6, 2019; revised June 27, 2019; accepted June 27, 2019. Declarations of Interest: None. Address correspondence to: E.Y.P.L. e-mail: [email protected] © 2019 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.acra.2019.06.022
diffusion coefficient (ADC) could differentiate malignant from benign lymph nodes in cervical carcinoma (3). Conventionally, it is assumed that the water diffusion is Gaussian. In fact, the probability distribution function of water diffusion in biologic tissues is affected by cell membranes and cellular compartment, both intracellular and extracellular compartments. Therefore, Gaussian diffusion model may no longer precisely represent the actual diffusion in biologic tissues. Diffusion kurtosis imaging (DKI) is an extension of DWI, which quantifies the non-Gaussian water diffusion when multiple and larger b values are acquired (4,5). The non-Gaussian behavior occurs in an irregular and heterogeneous environment with several large interfaces, including the increased nuclearcytoplasmic ratio of the tumor cell (6). DKI can potentially be more sensitive to tissue heterogeneity and water exchange. MD and MK are two quantitative parameters derived from DKI model. MD is a corrected apparent diffusion coefficient, and MK describes the deviation of the signal decay from a
1
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
WANG ET AL
monoexponential model, which reflects the interaction of water molecules and intracellular compounds (7). DKI has been successfully applied in the evaluation of carcinoma in various organs, mostly in brain and prostate (811). However, only a few studies reported the application of DKI in cervical carcinoma, ranging from tissue differentiation to tumor grading (6,12,13). The purpose of this study was to explore the additional value of DKI over conventional DWI in the characterization of cervical carcinoma, specifically in differentiating histologic subtypes, tumor grades, and FIGO stages; and compared between single-slice and volumetric assessment in quantitative DKI analyses. MATERIALS AND METHODS Patients
This retrospective study was approved by the local institutional review board and in accordance with the Helsinki Declaration. Informed consent was waived. The study was conducted between January 2016 and November 2018. Inclusion criteria included histologic confirmed treatmentnaived cervical carcinoma and without prior history of other malignancy. Exclusion criteria were those with contraindications to magnetic resonance imaging (MRI) and maximum tumor volume smaller than 1.5 cm3. Cervical biopsies were reported and reviewed by pathologist specialized in gynecologic malignancy and results were discussed at multidisciplinary team (MDT) meetings. Assessed pathologic parameters included histologic subtypes and tumor grades according to the WHO Classification of Tumors of Female Reproductive Organs (14). All cases were restaged using the revised 2018/9 FIGO staging for cervical carcinoma (15). Imaging Acquisition
All recruited patients had MRI with DWI sequence before surgery or chemoradiation. All MRI examinations were performed on a 3.0T platform (Achieva 3T TX, Philips Healthcare, Best, the Netherlands) with a 16-channel phased array torso coil. All the patients fasted for 6 hours and received 20 mg intravenous hyoscine butylbromide
(Buscopan, Boehringer Ingelheim, Germany) before MRI examinations to reduce the peristaltic artifacts. Imaging sequences were standardized for all patients (Table 1). DWI was acquired once using five b values (0, 500, 800, 1000, and 1500 s/mm2). Image Analysis
DKI is an extension of DWI and DKI analysis was based on the 5 acquired b values as aforementioned. In conventional DWI analysis, ADC was generated by selecting the 2 b values (0 and 1000 s/mm2) out of the acquired 5 b values. DKI was evaluated using the following equation to generate MD and MK: In½SðbÞ ¼ In½Sð0Þb ¢ D þ 16 b2 ¢ D2 ¢ K (16), where S represented the signal intensity, b was the b-value, D was the corrected apparent diffusion coefficient derived from non-Gaussian model, and K is a unitless parameter of apparent kurtosis coefficient. MD and MK were the averages of D and K among 3 distributed directions. MD, MK and ADC were calculated with in-house software written in MATLAB 2018b (Mathworks, Natick, MA, USA). MD and MK maps were generated by Philips DKI & DTI software (Philips Healthcare, China). First radiologist (2 years’ experience in pelvic MRI) determined the volume of interest (VOI) of the cervical carcinoma by strictly delineating the border of the tumor from surrounding normal tissue on every slice that the tumor existed. The VOI was subsequently verified by a second radiologist (>10 years’ experience in pelvic MRI). Both radiologists were blinded to the clinicopathologic results. The VOI of the cervical carcinoma was contoured on the b = 1000 s/mm2 images with reference to T2W images. The slice with the largest region of interest (ROI) was determined by MATLAB from the VOI. The DKI parameters from the ROI (MDR, MKR) and the VOI (MDV, MKV), ADC values from the ROI (ADCR) and the VOI (ADCV) were calculated. A 1.0 cm2 fixed-sized ROI was placed on the normal myometrium on b = 1000 s/ mm2 for each patient and the ROI was placed at least 2.0 cm away from the cervical tumor. All contouring was performed on ImageJ software (1.45s, National institutes of Health, USA).
TABLE 1. MRI Protocols
2
Sequences
T2W TSE
T2W SPAIR
T2W TSE
DWI/DKI
CE 3D T1W FFE
Plane TR/TE (ms) Turbo factor SENSE factor FOV (mm) Matrix size Slice thickness (mm) Intersection gap (mm) Bandwidth (Hz/pixel) Number of excitations
Sagittal 4000/80 30 2 240 £ 240 480 £ 298 4 0 230 2
Coronal 3500/80 21 2 230 £ 230 352 £ 300 4 0 186 1
Axial 2800/100 12 2 402 £ 300 787 £ 600 4 0 169 1
Axial 2000/54 NA 2 406 £ 300 168 £ 124 4 0 15.3 2
3D 3/1.4 NA 2 370 £ 203 248 £ 134 1.5 0 724 1
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
Statistical Analysis
All statistical analyses were performed on SPSS (version 25, USA) and MedCalc (version 18.2.1, MedCalc Software bvba, Ostend, Belgium). Normality test was performed. Intraclass correlation coefficient (ICC) and Bland-Altman plots were used to assess the consistency of the DKI parameters and ADC derived from ROI and VOI derivations. An ICC value of 0.000.20 indicated poor consistency; 0.210.40, fair consistency; 0.410.60, moderate consistency; 0.610.80, good consistency; and 0.811.00, excellent consistency. Pearson correlation was applied to evaluate the correlation between MD and MK. A paired t test was used to evaluate the differences in DKI parameters between cervical carcinoma and normal myometrium, as well as between ROI and VOI. Student’s t test and one-way ANOVA were used to compare the DKI parameters and ADC in different clinicopathologic groups (histologic subtypes, tumor grading, and FIGO staging). A two-tailed p value <0.05 was considered as statistical significance. RESULTS Clinical Characteristics
In total, 75 patients were enrolled in this study. Their demographics and clinical characteristics were showed in Table 2. Among them, 19 patients had thin myometrium or the normal myometrium was less than 2.0 cm from the boundary of the tumor; therefore, only 56 patients were included in the contouring of the normal myometrium. Comparison Between Single-Slice and Volumetric Quantification of Cervical Cancer
MD, MK, and ADC derived from both methods (ROI and VOI) were all in excellent consistency (MD: ICC = 0.91; TABLE 2. Clinical Characteristics Characteristics Number Age (y) Histology Squamous cell carcinoma (SCC) Adenocarcinoma (ACA) Others* Tumor grade# G1 G2 G3 FIGO stage I II III IV
Patients (n) 75 56.9 § 13.4 (range 2184) 55 17 3 5 18 49 7 21 41 6
* Others included one clear cell carcinoma, one malignant-mixed mullerian tumor, and one mucinous carcinoma. # Tumor grading was not possible in three patients.
IMAGING ASSESSMENT OF CERVICAL CARCINOMA
MK: ICC = 0.92; ADC: ICC = 0.93). Bland-Altman plots were graphically presented in Figure 1. DKI Parameters and ADC of Cervical Carcinoma
MD was negatively correlated with MK (ROI: r = 0.766, p < 0.001; VOI: r = 0.786, p < 0.001). MD was positively correlated with ADC (ROI: r = 0.974, p < 0.001; VOI: r = 0.983, p < 0.001). MD and ADC of cervical carcinoma were significantly lower than that of normal myometrium (MD and ADC, p < 0.001, Table 3) MK of cervical carcinoma was significantly higher than that of normal myometrium (p < 0.001, Table 3). MD and ADC showed significant differences between histologic subtypes. Squamous cell carcinoma (SCC) had lower MD and ADC than adenocarcinoma (ACA) (p < 0.050, Table 3). However, both MKR and MKV were not significantly different between these two histologic subtypes (Figs 2 and 3). No significant differences were observed in MD, MK, and ADC among different tumor grades, regardless of method of derivations (ROI and VOI) (Table 3). As with the FIGO staging, MD and ADC of FIGO III was significant higher than that of FIGO IIIIV. (p < 0.050, Table 3). MK showed no difference between FIGO III and FIGO IIIIV. DKI parameters and ADC were not discriminative when the FIGO stages were individually considered. DISCUSSION In this study, we showed that DKI parameters and ADC of cervical carcinoma were significantly different from normal myometrium. Additionally, MD and ADC were different between SCC and ACA, their corresponding FIGO stages, but MK was not contributory. Furthermore, single-slice ROI and VOI derivations were consistent with similar discriminative ability, suggesting that ROI derivation would be sufficient for quantitative DKI and ADC analysis in cervical carcinoma. Previous studies have shown DKI would allow tissue differentiation, specifically in separating cervical carcinoma from adjacent normal uterine tissues (6,13), as well as in other uterine malignancies such as endometrial carcinoma (17,18). However, the comparison between ROI and VOI placements in quantification of DKI parameters was not evaluated. Herein, we showed the ROI- and VOI-derived DKI parameters were comparable and consistent, more importantly the ROI-derived DKI parameters had similar discriminative ability as those derived from VOI in terms of tissue differentiation. With this, we believe ROI quantification for DKI will be sufficient for cervical carcinoma assessment, with the benefit of being easier to perform and time-saving. The ability of DKI to differentiate cervical carcinoma from myometrium corroborated with previous studies (6,13). DKI studies in other cancers also exhibited an advantage of DKI in differentiating benign from malignant tumors, especially in prostate, aiding in confidence in diagnosis and tumor delineation 3
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
WANG ET AL
Figure 1. Bland-Altman plots for MD (A), MK (B), and ADC (C) generated from different measurement of ROI or VOI.
TABLE 3. DKI Parameters and ADC in Cervical Carcinoma and Normal Myometrium Characteristic
Tissue Cervical carcinoma Normal myometrium p Value Histologic subtypes SCC ACA p Value Tumor grades G12 (well-moderately differentiated) G3 (poorly differentiated) p Value FIGO stages III IIIIV p Value
4
MD (£ 103 mm2/s)
ADC (£ 103 mm2/s)
MK
MDR
MDV
ADCR
ADCV
MKR
MKV
1.39 § 0.23 1.81 § 0.31 <0.001
1.39 § 0.19 1.81 § 0.31 <0.001
1.02 § 0.17 1.25 § 0.17 <0.001
1.10 § 0.15 1.25 § 0.17 <0.001
1.05 § 0.13 0.89 § 0.10 <0.001
1.03 § 0.12 0.89 § 0.10 <0.001
1.36 § 0.20 1.59 § 0.32 0.010
1.37 § 0.16 1.57 § 0.27 0.011
1.01 § 0.14 1.16 § 0.24 0.019
1.08 § 0.13 1.24 § 0.23 0.014
1.04 § 0.13 0.97 § 0.16 0.078
1.03 § 0.11 0.97 § 0.16 0.092
1.42 § 0.25 1.41 § 0.23 0.878
1.47 § 0.22 1.40 § 0.20 0.227
1.04 § 0.16 1.03 § 0.17 0.872
1.15 § 0.17 1.11 § 0.16 0.342
1.02 § 0.12 1.03 § 0.14 0.885
1.00 § 0.11 1.02 § 0.13 0.562
1.50 § 0.25 1.36 § 0.22 0.017
1.49 § 0.21 1.38 § 0.19 0.025
1.09 § 0.20 1.01 § 0.15 0.040
1.17 § 0.18 1.09 § 0.15 0.038
1.00 § 0.14 1.03 § 0.13 0.321
1.00 § 0.13 1.02 § 0.12 0.380
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
IMAGING ASSESSMENT OF CERVICAL CARCINOMA
Figure 2. An 82-year-old patient with squamous cell carcinoma. (A). DWI at b = 1000 s/mm2 shows a large hyperintense mass arising from the ectocervix with marked restricted diffusion (white arrow); (B) ADC map shows a hypointense mass, ADCR = 0.91 £ 103 mm2/s, ADCV = 0.92 £ 103 mm2/s; (C) MD map, MDR = 1.20 £ 103 mm2/s, MDV = 1.26 £ 103 mm2/s; (D) MK map, MKR = 1.06, MKV = 1.14.
(10,11,13,1921). The lower MD in cervical carcinoma was likely related to the restricted of free water diffusion in more cellular packed tumor environment, whereas the higher MK would suggest greater deviation of Gaussian diffusion in cervical carcinoma than normal myometrium, as shown in Dia et al. and in other tumors (11,13,17,19,22). Nevertheless, similar diagnostic performance of ADC was observed in tissue differentiation in our study, and had been widely reported previously (2329). In this study, we had chosen myometrium from the same patient as a reference tissue as it was easily accessible on the same scan. Most of the patients in our cohort had bulky tumors without normal cervix, hence the normal cervix could not be used as a
comparison reference. Using the normal myometrial ROI could avoid discrepancies in individual factors that may affect quantification, for example, age, menopausal status, hormonal status to name a few. MD and ADC in SCC were lower than those in ACA, allowing histologic subtyping, and our result was in line with the previous DKI and conventional DWI studies (12,13,25,30,31). However, MK was not discriminative in this aspect, similar to findings reported Dia et al. (13). It could be that both SCC and ACA affected the non-Gaussian diffusion in a similar extent as depicted by MK. This was contrary to what was found in Winfield et al. that MK was statistically 5
ARTICLE IN PRESS WANG ET AL
Academic Radiology, Vol &, No &&, && 2019
Figure 3. A 67-year-old patient with adenocarcinoma. (A) DWI at b = 1000 s/mm2 shows an irregular hyperintense cervical mass extending to the uterine cavity with marked restricted diffusion (white arrow); (B) ADC map shows hypointense mass, ADCR = 0.99 £ 103 mm2/s, ADCV = 0.99 £ 103 mm2/s; (C) MD map, MDR = 1.33 £ 103 mm2/s, MDV = 1.24 £ 103 mm2/s; (D) MK map, MKR = 1.05, MKV = 1.06.
higher in SCC than ACA. The discrepancy could be related to the different b values used in their study; with the highest b value of 800 s/mm2 in the study by Winfield et al., which may not sufficiently high enough to capture the nonGaussian diffusion (12,32). Furthermore, MK quantification is more susceptible to imaging noise, motion, and artifacts, which could have affected the discriminative power of MK in histologic subtyping (4). Both MD and ADC observed significant decrease from low FIGO stage (FIGO III) to high FIGO stage (FIGO IIIIV), while MK did not. None of the parameters were significantly different between tumor grades. Contrary to the 6
literature, ADC was used for distinguishing tumor grades in cervical carcinoma (23,25,30,31,33). Additionally, DKI was shown to be helpful in tumor grading of endometrial carcinoma, more superior than ADC (17,18,34). Nevertheless, results were mixed in other extracranial cancers (10,11,19,21,3537). Hence, further research is required in this area. Our study had included a high proportion of poorly differentiated (G3) cervical carcinoma, while others have more balanced distribution of the cases between wellmoderately differentiated tumors (G12) and poorly differentiated tumors, which could have affected the results here (6,12). Besides, none of the authorities has stated histologic
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
criteria for grading cervical carcinoma so far, thus histologic grade has largely been subjective and may not be an accurate histologic marker to compare with DKI parameters and ADC (38). The high intraobserver and interobserver variability in tumor grading could further confound the finding in this study as compared to others (39). Overall, MD and ADC exhibited similar performance and ability in discriminating cervical carcinoma and normal myometrium, histologic subtypes, and FIGO stages. MK was not able to differentiate these clinicopathologic signatures above. Consequently, no added value had been observed over conventional DWI for characterization of cervical carcinoma in our DKI study. Our study had several limitations. First, the maximal b value in our study might not be high enough to capture the nonGaussian diffusion. Other DKI studies on cervical carcinoma were able to differentiate benign and malignant tumors, as well as the tumor grades when the highest b value of 2000 or 2500 s/mm2 was used (6,13). Rosenkrantz et al. (32) suggested that at least three b values should be acquired for DKI analysis outside of the brain, at least two of these should be both above and below b = 1000 s/mm2 to allow robust estimates of MD and MK. In our study, we used five b values (0, 500, 800, 1000, and 1500 s/mm2), and the highest b value was at least 1500 s/mm2, which met the criteria suggested. Second, this was a retrospective study and the sample size of patients was not substantial large, for this reason, there were only five cases of G1, seven cases of FIGO I, and six cases of FIGO IV, these could reduce the accuracy of results. Further study with more patients would be needed to validate our findings and generalizability of our results. Third, the DKI parameters between malignant and benign tumors were not compared, we only compared the difference between cervical carcinoma and normal tissue. However, there are very few benign entities in the cervix, which could not be differentiated from carcinoma based on conventional morphologic imaging on MRI, making this exercise less meaningful. In conclusion, largest single-slice ROI analysis was sufficient for both DKI and ADC quantification in cervical carcinoma. Although MD and MK allowed tissue differentiating, they had limited role in separating different clinicopathologic signatures of cervical carcinoma, with only MD able to distinguish the histologic subtypes and FIGO stages (FIGO III and FIGO IIIIV). DKI showed no clear added value compared to conventional DWI in the assessment of cervical carcinoma. REFERENCES 1.
2.
3.
4.
Dappa E, Elger T, Hasenburg A, et al. The value of advanced MRI techniques in the assessment of cervical cancer: a review. Insights Imaging 2017; 8(5):471–481. Park JJ, Kim CK, Park SY, et al. Parametrial invasion in cervical cancer: fused T2-weighted imaging and high-b-value diffusion-weighted imaging with background body signal suppression at 3T. Radiology 2015; 274 (3):734–741. Liu Y, Liu H, Bai X, et al. Differentiation of metastatic from non-metastatic lymph nodes in patients with uterine cervical cancer using diffusionweighted imaging. Gynecol Oncol 2011; 122(1):19–24. Jensen JH, Helpern JA. MRI quantification of non-Gaussian water diffusion by kurtosis analysis. NMR Biomed 2010; 23(7):698–710.
IMAGING ASSESSMENT OF CERVICAL CARCINOMA
5. Jensen JH, Helpern JA, Ramani A, et al. Diffusional kurtosis imaging: the quantification of non-Gaussian water diffusion by means of magnetic resonance imaging. Magn Reson Med 2005; 53(6):1432–1440. 6. Wang P, Thapa D, Wu G, et al. A study on diffusion and kurtosis features of cervical cancer based on non-Gaussian diffusion weighted model. Magn Reson Imaging 2018; 47:60–66. 7. Le Bihan D. Apparent diffusion coefficient and beyond: what diffusion MR imaging can tell us about tissue structure. Radiology 2013; 268 (2):318–322. 8. Minosse S, Marzi S, Piludu F, et al. Correlation study between DKI and conventional DWI in brain and head and neck tumors. Magn Reson Imaging 2017; 42:114–122. 9. Falk Delgado A, Nilsson M, van Westen D, et al. Glioma grade discrimination with MR diffusion kurtosis imaging: a meta-analysis of diagnostic accuracy. Radiology 2018; 287(1):119–127. 10. Wang X, Tu N, Qin T, et al. Diffusion kurtosis imaging combined with DWI at 3-T MRI for detection and assessment of aggressiveness of prostate cancer. AJR Am J Roentgenol 2018; 211(4):797–804. 11. Tamada T, Prabhu V, Li J, et al. Prostate cancer: diffusion-weighted MR imaging for detection and assessment of aggressiveness-comparison between conventional and kurtosis models. Radiology 2017; 284(1):100–108. 12. Winfield JM, Orton MR, Collins DJ, et al. Separation of type and grade in cervical tumours using non-mono-exponential models of diffusionweighted MRI. Eur Radiol 2017; 27(2):627–636. 13. Dia AA, Hori M, Onishi H, et al. Application of non-Gaussian water diffusional kurtosis imaging in the assessment of uterine tumors: a preliminary study. PLoS One 2017; 12(11):e0188434. 14. Kurman RJ, Carcangiu ML, Herrington CS, et al. WHO classification of tumours of female reproductive organs. France: Lyon: International Agency for Research on Cancer, 2014. 15. Bhatla N, Berek JS, Cuello Fredes M, et al. Revised FIGO staging for carcinoma of the cervix uteri. Int J Gynaecol Obstet 2019; 145(1):129–135. 16. Moseley ME, Cohen Y, Kucharczyk J, et al. Diffusion-weighted MR imaging of anisotropic water diffusion in cat central nervous system. Radiology 1990; 176(2):439–445. 17. Yue W, Meng N, Wang J, et al. Comparative analysis of the value of diffusion kurtosis imaging and diffusion-weighted imaging in evaluating the histological features of endometrial cancer. Cancer Imaging 2019; 19(1):9. 18. Yamada I, Sakamoto J, Kobayashi D, et al. Diffusion kurtosis imaging of endometrial carcinoma: correlation with histopathological findings. Magn Reson Imaging 2019; 57:337–346. 19. Sun K, Chen X, Chai W, et al. Breast cancer: diffusion kurtosis MR imaging-diagnostic accuracy and correlation with clinical-pathologic factors. Radiology 2015; 277(1):46–55. 20. Tamura C, Shinmoto H, Soga S, et al. Diffusion kurtosis imaging study of prostate cancer: preliminary findings. J Magn Reson Imaging 2014; 40 (3):723–729. 21. Quentin M, Pentang G, Schimmoller L, et al. Feasibility of diffusional kurtosis tensor imaging in prostate MRI for the assessment of prostate cancer: preliminary results. Magn Reson Imaging 2014; 32(7):880–885. 22. Barrett T, McLean M, Priest AN, et al. Diagnostic evaluation of magnetization transfer and diffusion kurtosis imaging for prostate cancer detection in a re-biopsy population. Eur Radiol 2018; 28(8):3141–3150. 23. Exner M, Kuhn A, Stumpp P, et al. Value of diffusion-weighted MRI in diagnosis of uterine cervical cancer: a prospective study evaluating the benefits of DWI compared to conventional MR sequences in a 3T environment. Acta Radiol 2016; 57(7):869–877. 24. Guan Y, Shi H, Chen Y, et al. Whole-lesion histogram analysis of apparent diffusion coefficient for the assessment of cervical cancer. J Comput Assist Tomogr 2016; 40(2):212–217. 25. Kuang F, Ren J, Zhong Q, et al. The value of apparent diffusion coefficient in the assessment of cervical cancer. Eur Radiol 2013; 23(4):1050–1058. 26. Hoogendam JP, Klerkx WM, de Kort GA, et al. The influence of the bvalue combination on apparent diffusion coefficient based differentiation between malignant and benign tissue in cervical cancer. J Magn Reson Imaging 2010; 32(2):376–382. 27. Payne GS, Schmidt M, Morgan VA, et al. Evaluation of magnetic resonance diffusion and spectroscopy measurements as predictive biomarkers in stage 1 cervical cancer. Gynecol Oncol 2010; 116(2):246–252. 28. Chen J, Zhang Y, Liang B, et al. The utility of diffusion-weighted MR imaging in cervical cancer. Eur J Radiol 2010; 74(3):e101–e106. 29. Kuang F, Yan Z, Li H, et al. Diagnostic accuracy of diffusion-weighted MRI for differentiation of cervical cancer and benign cervical lesions at
7
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
WANG ET AL
30.
31.
32.
33.
34.
8
3.0T: comparison with routine MRI and dynamic contrast-enhanced MRI. J Magn Reson Imaging 2015; 42(4):1094–1099. Liu Y, Ye Z, Sun H, et al. Clinical application of diffusion-weighted magnetic resonance imaging in uterine cervical cancer. Int J Gynecol Cancer 2015; 25(6):1073–1078. Karunya RJ, Tharani P, John S, et al. Role of functional magnetic resonance imaging derived parameters as imaging biomarkers and correlation with clinicopathological features in carcinoma of uterine cervix. J Clin Diagn Res 2017; 11(8):Xc06–xc11. Rosenkrantz AB, Padhani AR, Chenevert TL, et al. Body diffusion kurtosis imaging: basic principles, applications, and considerations for clinical practice. J Magn Reson Imaging 2015; 42(5):1190–1202. Liu Y, Ye Z, Sun H, et al. Grading of uterine cervical cancer by using the ADC difference value and its correlation with microvascular density and vascular endothelial growth factor. Eur Radiol 2013; 23 (3):757–765. Chen T, Li Y, Lu SS, et al. Quantitative evaluation of diffusion-kurtosis imaging for grading endometrial carcinoma: a comparative study
35.
36.
37.
38. 39.
with diffusion-weighted imaging. Clin Radiol 2017; 72(11):995.e911– 995.e920. Zhu L, Pan Z, Ma Q, et al. Diffusion kurtosis imaging study of rectal adenocarcinoma associated with histopathologic prognostic factors: preliminary findings. Radiology 2017; 284(1):66–76. Wen Z, Chen Y, Yang X, et al. Application of magnetic resonance diffusion kurtosis imaging for distinguishing histopathologic subtypes and grades of rectal carcinoma. Cancer Imaging 2019; 19(1):8. Sun Y, Xiao Q, Hu F, et al. Diffusion kurtosis imaging in the characterisation of rectal cancer: utilizing the most repeatable region-of-interest strategy for diffusion parameters on a 3T scanner. Eur Radiol 2018; 28 (12):5211–5220. McCluggage WG. Towards developing a meaningful grading system for cervical squamous cell carcinoma. J Pathol Clin Res 2018; 4(2):81–85. Noviello MB, Silva-Filho AL, Traiman P, et al. Inter- and intraobserver variability in the assessment of tumor grade and lymphovascular space invasion in patients with squamous cell carcinoma of the cervix. Eur J Obstet Gynecol Reprod Biol 2008; 138(2):246–248.
Original Investigation
Diffusion Kurtosis Imaging in the Assessment of Cervical Carcinoma Mandi Wang, MMedSci, Jose A.U. Perucho, BEng, Queenie Chan, PhD, Jianqing Sun, PhD, Philip Ip, MBChB, FRCPath, FHKCPath, FHKAM (Pathology), Ka Yu Tse, MBBS, MMedSci, FHKCOG, FHKAM (O&G), Elaine Y.P. Lee, BMedSci, BMBS, FRCR
Abbreviations DKI diffusion kurtosis imaging MD mean diffusivity MK mean kurtosis DWI diffusion-weighted imaging ADC apparent diffusion coefficient ICC intraclass correlation coefficient VOI volume of interest ROI region of interest SCC squamous cell carcinoma ACA adenocarcinoma
Rationale and Objectives: To evaluate the additional value of diffusion kurtosis imaging (DKI) in the characterization of cervical carcinoma. Materials and Methods: Seventy-five patients (56.9 § 13.4 years) with histologic-confirmed cervical carcinoma were included. Diffusion-weighted imaging (DWI) was acquired on a 3T MRI with five b values (0, 500, 800, 1000, and 1500 s/mm2). Data were analyzed based on DKI model (5 b values) and conventional DWI (0 and 1000 s/mm2). Largest single-slice region of interest (ROI) and volume of interest (VOI) were drawn around the tumor. Mean diffusivity (MD), mean kurtosis (MK), and apparent diffusion coefficient (ADC) of cervical carcinoma and normal myometrium were measured and compared. MD, MK, and ADC of cervical carcinoma were compared among histologic subtypes, tumor grades, and FIGO stages. Results: ROI- and VOI-derived DKI parameters and ADC were all in excellent consistency (intraclass correlation coefficient, ICC > 0.90, respectively). Cervical carcinoma had significantly lower MD, ADC, and higher MK than normal myometrium (p < 0.001). MD and ADC showed significant differences between histologic subtypes and FIGO stages, lower in squamous cell carcinoma than adenocarcinoma and higher in FIGO III than FIGO IIIIV (p < 0.050), but not with tumor grade. No difference was observed in MK for different clinicopathologic features tested. Conclusion: ROI and VOI analyses were in excellent consistency. MD and ADC were able to distinguish histologic subtypes and separating FIGO stages, MK could not. DKI showed no clear added value over conventional DWI in the characterization of cervical carcinoma. Key Words: Diffusion kurtosis imaging; Diffusion-weighted imaging; Cervical carcinoma; Histologic subtypes, FIGO. © 2019 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.
INTRODUCTION
D
iffusion-weighted imaging (DWI) has emerged as an important and promising tool in the assessment of cervical carcinoma (1). DWI combined with T2weighted (T2W) imaging increases the accuracy of determining parametrial invasion (2). Quantitative DWI using the apparent
Acad Radiol 2019; &:1–8 From the Department of Diagnostic Radiology, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Pok Fu Lam, Hong Kong (M.W., J.A.U.P., E.Y.P.L.); Philips Healthcare, Greater China, China (Q.C.); Philips Healthcare, Shanghai, China (J.S.); Department of Pathology, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong (P.I.); Department of Obstetrics and Gynaecology, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong (K.Y.T.). Received June 6, 2019; revised June 27, 2019; accepted June 27, 2019. Declarations of Interest: None. Address correspondence to: E.Y.P.L. e-mail: [email protected] © 2019 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.acra.2019.06.022
diffusion coefficient (ADC) could differentiate malignant from benign lymph nodes in cervical carcinoma (3). Conventionally, it is assumed that the water diffusion is Gaussian. In fact, the probability distribution function of water diffusion in biologic tissues is affected by cell membranes and cellular compartment, both intracellular and extracellular compartments. Therefore, Gaussian diffusion model may no longer precisely represent the actual diffusion in biologic tissues. Diffusion kurtosis imaging (DKI) is an extension of DWI, which quantifies the non-Gaussian water diffusion when multiple and larger b values are acquired (4,5). The non-Gaussian behavior occurs in an irregular and heterogeneous environment with several large interfaces, including the increased nuclearcytoplasmic ratio of the tumor cell (6). DKI can potentially be more sensitive to tissue heterogeneity and water exchange. MD and MK are two quantitative parameters derived from DKI model. MD is a corrected apparent diffusion coefficient, and MK describes the deviation of the signal decay from a
1
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
WANG ET AL
monoexponential model, which reflects the interaction of water molecules and intracellular compounds (7). DKI has been successfully applied in the evaluation of carcinoma in various organs, mostly in brain and prostate (811). However, only a few studies reported the application of DKI in cervical carcinoma, ranging from tissue differentiation to tumor grading (6,12,13). The purpose of this study was to explore the additional value of DKI over conventional DWI in the characterization of cervical carcinoma, specifically in differentiating histologic subtypes, tumor grades, and FIGO stages; and compared between single-slice and volumetric assessment in quantitative DKI analyses. MATERIALS AND METHODS Patients
This retrospective study was approved by the local institutional review board and in accordance with the Helsinki Declaration. Informed consent was waived. The study was conducted between January 2016 and November 2018. Inclusion criteria included histologic confirmed treatmentnaived cervical carcinoma and without prior history of other malignancy. Exclusion criteria were those with contraindications to magnetic resonance imaging (MRI) and maximum tumor volume smaller than 1.5 cm3. Cervical biopsies were reported and reviewed by pathologist specialized in gynecologic malignancy and results were discussed at multidisciplinary team (MDT) meetings. Assessed pathologic parameters included histologic subtypes and tumor grades according to the WHO Classification of Tumors of Female Reproductive Organs (14). All cases were restaged using the revised 2018/9 FIGO staging for cervical carcinoma (15). Imaging Acquisition
All recruited patients had MRI with DWI sequence before surgery or chemoradiation. All MRI examinations were performed on a 3.0T platform (Achieva 3T TX, Philips Healthcare, Best, the Netherlands) with a 16-channel phased array torso coil. All the patients fasted for 6 hours and received 20 mg intravenous hyoscine butylbromide
(Buscopan, Boehringer Ingelheim, Germany) before MRI examinations to reduce the peristaltic artifacts. Imaging sequences were standardized for all patients (Table 1). DWI was acquired once using five b values (0, 500, 800, 1000, and 1500 s/mm2). Image Analysis
DKI is an extension of DWI and DKI analysis was based on the 5 acquired b values as aforementioned. In conventional DWI analysis, ADC was generated by selecting the 2 b values (0 and 1000 s/mm2) out of the acquired 5 b values. DKI was evaluated using the following equation to generate MD and MK: In½SðbÞ ¼ In½Sð0Þb ¢ D þ 16 b2 ¢ D2 ¢ K (16), where S represented the signal intensity, b was the b-value, D was the corrected apparent diffusion coefficient derived from non-Gaussian model, and K is a unitless parameter of apparent kurtosis coefficient. MD and MK were the averages of D and K among 3 distributed directions. MD, MK and ADC were calculated with in-house software written in MATLAB 2018b (Mathworks, Natick, MA, USA). MD and MK maps were generated by Philips DKI & DTI software (Philips Healthcare, China). First radiologist (2 years’ experience in pelvic MRI) determined the volume of interest (VOI) of the cervical carcinoma by strictly delineating the border of the tumor from surrounding normal tissue on every slice that the tumor existed. The VOI was subsequently verified by a second radiologist (>10 years’ experience in pelvic MRI). Both radiologists were blinded to the clinicopathologic results. The VOI of the cervical carcinoma was contoured on the b = 1000 s/mm2 images with reference to T2W images. The slice with the largest region of interest (ROI) was determined by MATLAB from the VOI. The DKI parameters from the ROI (MDR, MKR) and the VOI (MDV, MKV), ADC values from the ROI (ADCR) and the VOI (ADCV) were calculated. A 1.0 cm2 fixed-sized ROI was placed on the normal myometrium on b = 1000 s/ mm2 for each patient and the ROI was placed at least 2.0 cm away from the cervical tumor. All contouring was performed on ImageJ software (1.45s, National institutes of Health, USA).
TABLE 1. MRI Protocols
2
Sequences
T2W TSE
T2W SPAIR
T2W TSE
DWI/DKI
CE 3D T1W FFE
Plane TR/TE (ms) Turbo factor SENSE factor FOV (mm) Matrix size Slice thickness (mm) Intersection gap (mm) Bandwidth (Hz/pixel) Number of excitations
Sagittal 4000/80 30 2 240 £ 240 480 £ 298 4 0 230 2
Coronal 3500/80 21 2 230 £ 230 352 £ 300 4 0 186 1
Axial 2800/100 12 2 402 £ 300 787 £ 600 4 0 169 1
Axial 2000/54 NA 2 406 £ 300 168 £ 124 4 0 15.3 2
3D 3/1.4 NA 2 370 £ 203 248 £ 134 1.5 0 724 1
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
Statistical Analysis
All statistical analyses were performed on SPSS (version 25, USA) and MedCalc (version 18.2.1, MedCalc Software bvba, Ostend, Belgium). Normality test was performed. Intraclass correlation coefficient (ICC) and Bland-Altman plots were used to assess the consistency of the DKI parameters and ADC derived from ROI and VOI derivations. An ICC value of 0.000.20 indicated poor consistency; 0.210.40, fair consistency; 0.410.60, moderate consistency; 0.610.80, good consistency; and 0.811.00, excellent consistency. Pearson correlation was applied to evaluate the correlation between MD and MK. A paired t test was used to evaluate the differences in DKI parameters between cervical carcinoma and normal myometrium, as well as between ROI and VOI. Student’s t test and one-way ANOVA were used to compare the DKI parameters and ADC in different clinicopathologic groups (histologic subtypes, tumor grading, and FIGO staging). A two-tailed p value <0.05 was considered as statistical significance. RESULTS Clinical Characteristics
In total, 75 patients were enrolled in this study. Their demographics and clinical characteristics were showed in Table 2. Among them, 19 patients had thin myometrium or the normal myometrium was less than 2.0 cm from the boundary of the tumor; therefore, only 56 patients were included in the contouring of the normal myometrium. Comparison Between Single-Slice and Volumetric Quantification of Cervical Cancer
MD, MK, and ADC derived from both methods (ROI and VOI) were all in excellent consistency (MD: ICC = 0.91; TABLE 2. Clinical Characteristics Characteristics Number Age (y) Histology Squamous cell carcinoma (SCC) Adenocarcinoma (ACA) Others* Tumor grade# G1 G2 G3 FIGO stage I II III IV
Patients (n) 75 56.9 § 13.4 (range 2184) 55 17 3 5 18 49 7 21 41 6
* Others included one clear cell carcinoma, one malignant-mixed mullerian tumor, and one mucinous carcinoma. # Tumor grading was not possible in three patients.
IMAGING ASSESSMENT OF CERVICAL CARCINOMA
MK: ICC = 0.92; ADC: ICC = 0.93). Bland-Altman plots were graphically presented in Figure 1. DKI Parameters and ADC of Cervical Carcinoma
MD was negatively correlated with MK (ROI: r = 0.766, p < 0.001; VOI: r = 0.786, p < 0.001). MD was positively correlated with ADC (ROI: r = 0.974, p < 0.001; VOI: r = 0.983, p < 0.001). MD and ADC of cervical carcinoma were significantly lower than that of normal myometrium (MD and ADC, p < 0.001, Table 3) MK of cervical carcinoma was significantly higher than that of normal myometrium (p < 0.001, Table 3). MD and ADC showed significant differences between histologic subtypes. Squamous cell carcinoma (SCC) had lower MD and ADC than adenocarcinoma (ACA) (p < 0.050, Table 3). However, both MKR and MKV were not significantly different between these two histologic subtypes (Figs 2 and 3). No significant differences were observed in MD, MK, and ADC among different tumor grades, regardless of method of derivations (ROI and VOI) (Table 3). As with the FIGO staging, MD and ADC of FIGO III was significant higher than that of FIGO IIIIV. (p < 0.050, Table 3). MK showed no difference between FIGO III and FIGO IIIIV. DKI parameters and ADC were not discriminative when the FIGO stages were individually considered. DISCUSSION In this study, we showed that DKI parameters and ADC of cervical carcinoma were significantly different from normal myometrium. Additionally, MD and ADC were different between SCC and ACA, their corresponding FIGO stages, but MK was not contributory. Furthermore, single-slice ROI and VOI derivations were consistent with similar discriminative ability, suggesting that ROI derivation would be sufficient for quantitative DKI and ADC analysis in cervical carcinoma. Previous studies have shown DKI would allow tissue differentiation, specifically in separating cervical carcinoma from adjacent normal uterine tissues (6,13), as well as in other uterine malignancies such as endometrial carcinoma (17,18). However, the comparison between ROI and VOI placements in quantification of DKI parameters was not evaluated. Herein, we showed the ROI- and VOI-derived DKI parameters were comparable and consistent, more importantly the ROI-derived DKI parameters had similar discriminative ability as those derived from VOI in terms of tissue differentiation. With this, we believe ROI quantification for DKI will be sufficient for cervical carcinoma assessment, with the benefit of being easier to perform and time-saving. The ability of DKI to differentiate cervical carcinoma from myometrium corroborated with previous studies (6,13). DKI studies in other cancers also exhibited an advantage of DKI in differentiating benign from malignant tumors, especially in prostate, aiding in confidence in diagnosis and tumor delineation 3
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
WANG ET AL
Figure 1. Bland-Altman plots for MD (A), MK (B), and ADC (C) generated from different measurement of ROI or VOI.
TABLE 3. DKI Parameters and ADC in Cervical Carcinoma and Normal Myometrium Characteristic
Tissue Cervical carcinoma Normal myometrium p Value Histologic subtypes SCC ACA p Value Tumor grades G12 (well-moderately differentiated) G3 (poorly differentiated) p Value FIGO stages III IIIIV p Value
4
MD (£ 103 mm2/s)
ADC (£ 103 mm2/s)
MK
MDR
MDV
ADCR
ADCV
MKR
MKV
1.39 § 0.23 1.81 § 0.31 <0.001
1.39 § 0.19 1.81 § 0.31 <0.001
1.02 § 0.17 1.25 § 0.17 <0.001
1.10 § 0.15 1.25 § 0.17 <0.001
1.05 § 0.13 0.89 § 0.10 <0.001
1.03 § 0.12 0.89 § 0.10 <0.001
1.36 § 0.20 1.59 § 0.32 0.010
1.37 § 0.16 1.57 § 0.27 0.011
1.01 § 0.14 1.16 § 0.24 0.019
1.08 § 0.13 1.24 § 0.23 0.014
1.04 § 0.13 0.97 § 0.16 0.078
1.03 § 0.11 0.97 § 0.16 0.092
1.42 § 0.25 1.41 § 0.23 0.878
1.47 § 0.22 1.40 § 0.20 0.227
1.04 § 0.16 1.03 § 0.17 0.872
1.15 § 0.17 1.11 § 0.16 0.342
1.02 § 0.12 1.03 § 0.14 0.885
1.00 § 0.11 1.02 § 0.13 0.562
1.50 § 0.25 1.36 § 0.22 0.017
1.49 § 0.21 1.38 § 0.19 0.025
1.09 § 0.20 1.01 § 0.15 0.040
1.17 § 0.18 1.09 § 0.15 0.038
1.00 § 0.14 1.03 § 0.13 0.321
1.00 § 0.13 1.02 § 0.12 0.380
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
IMAGING ASSESSMENT OF CERVICAL CARCINOMA
Figure 2. An 82-year-old patient with squamous cell carcinoma. (A). DWI at b = 1000 s/mm2 shows a large hyperintense mass arising from the ectocervix with marked restricted diffusion (white arrow); (B) ADC map shows a hypointense mass, ADCR = 0.91 £ 103 mm2/s, ADCV = 0.92 £ 103 mm2/s; (C) MD map, MDR = 1.20 £ 103 mm2/s, MDV = 1.26 £ 103 mm2/s; (D) MK map, MKR = 1.06, MKV = 1.14.
(10,11,13,1921). The lower MD in cervical carcinoma was likely related to the restricted of free water diffusion in more cellular packed tumor environment, whereas the higher MK would suggest greater deviation of Gaussian diffusion in cervical carcinoma than normal myometrium, as shown in Dia et al. and in other tumors (11,13,17,19,22). Nevertheless, similar diagnostic performance of ADC was observed in tissue differentiation in our study, and had been widely reported previously (2329). In this study, we had chosen myometrium from the same patient as a reference tissue as it was easily accessible on the same scan. Most of the patients in our cohort had bulky tumors without normal cervix, hence the normal cervix could not be used as a
comparison reference. Using the normal myometrial ROI could avoid discrepancies in individual factors that may affect quantification, for example, age, menopausal status, hormonal status to name a few. MD and ADC in SCC were lower than those in ACA, allowing histologic subtyping, and our result was in line with the previous DKI and conventional DWI studies (12,13,25,30,31). However, MK was not discriminative in this aspect, similar to findings reported Dia et al. (13). It could be that both SCC and ACA affected the non-Gaussian diffusion in a similar extent as depicted by MK. This was contrary to what was found in Winfield et al. that MK was statistically 5
ARTICLE IN PRESS WANG ET AL
Academic Radiology, Vol &, No &&, && 2019
Figure 3. A 67-year-old patient with adenocarcinoma. (A) DWI at b = 1000 s/mm2 shows an irregular hyperintense cervical mass extending to the uterine cavity with marked restricted diffusion (white arrow); (B) ADC map shows hypointense mass, ADCR = 0.99 £ 103 mm2/s, ADCV = 0.99 £ 103 mm2/s; (C) MD map, MDR = 1.33 £ 103 mm2/s, MDV = 1.24 £ 103 mm2/s; (D) MK map, MKR = 1.05, MKV = 1.06.
higher in SCC than ACA. The discrepancy could be related to the different b values used in their study; with the highest b value of 800 s/mm2 in the study by Winfield et al., which may not sufficiently high enough to capture the nonGaussian diffusion (12,32). Furthermore, MK quantification is more susceptible to imaging noise, motion, and artifacts, which could have affected the discriminative power of MK in histologic subtyping (4). Both MD and ADC observed significant decrease from low FIGO stage (FIGO III) to high FIGO stage (FIGO IIIIV), while MK did not. None of the parameters were significantly different between tumor grades. Contrary to the 6
literature, ADC was used for distinguishing tumor grades in cervical carcinoma (23,25,30,31,33). Additionally, DKI was shown to be helpful in tumor grading of endometrial carcinoma, more superior than ADC (17,18,34). Nevertheless, results were mixed in other extracranial cancers (10,11,19,21,3537). Hence, further research is required in this area. Our study had included a high proportion of poorly differentiated (G3) cervical carcinoma, while others have more balanced distribution of the cases between wellmoderately differentiated tumors (G12) and poorly differentiated tumors, which could have affected the results here (6,12). Besides, none of the authorities has stated histologic
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
criteria for grading cervical carcinoma so far, thus histologic grade has largely been subjective and may not be an accurate histologic marker to compare with DKI parameters and ADC (38). The high intraobserver and interobserver variability in tumor grading could further confound the finding in this study as compared to others (39). Overall, MD and ADC exhibited similar performance and ability in discriminating cervical carcinoma and normal myometrium, histologic subtypes, and FIGO stages. MK was not able to differentiate these clinicopathologic signatures above. Consequently, no added value had been observed over conventional DWI for characterization of cervical carcinoma in our DKI study. Our study had several limitations. First, the maximal b value in our study might not be high enough to capture the nonGaussian diffusion. Other DKI studies on cervical carcinoma were able to differentiate benign and malignant tumors, as well as the tumor grades when the highest b value of 2000 or 2500 s/mm2 was used (6,13). Rosenkrantz et al. (32) suggested that at least three b values should be acquired for DKI analysis outside of the brain, at least two of these should be both above and below b = 1000 s/mm2 to allow robust estimates of MD and MK. In our study, we used five b values (0, 500, 800, 1000, and 1500 s/mm2), and the highest b value was at least 1500 s/mm2, which met the criteria suggested. Second, this was a retrospective study and the sample size of patients was not substantial large, for this reason, there were only five cases of G1, seven cases of FIGO I, and six cases of FIGO IV, these could reduce the accuracy of results. Further study with more patients would be needed to validate our findings and generalizability of our results. Third, the DKI parameters between malignant and benign tumors were not compared, we only compared the difference between cervical carcinoma and normal tissue. However, there are very few benign entities in the cervix, which could not be differentiated from carcinoma based on conventional morphologic imaging on MRI, making this exercise less meaningful. In conclusion, largest single-slice ROI analysis was sufficient for both DKI and ADC quantification in cervical carcinoma. Although MD and MK allowed tissue differentiating, they had limited role in separating different clinicopathologic signatures of cervical carcinoma, with only MD able to distinguish the histologic subtypes and FIGO stages (FIGO III and FIGO IIIIV). DKI showed no clear added value compared to conventional DWI in the assessment of cervical carcinoma. REFERENCES 1.
2.
3.
4.
Dappa E, Elger T, Hasenburg A, et al. The value of advanced MRI techniques in the assessment of cervical cancer: a review. Insights Imaging 2017; 8(5):471–481. Park JJ, Kim CK, Park SY, et al. Parametrial invasion in cervical cancer: fused T2-weighted imaging and high-b-value diffusion-weighted imaging with background body signal suppression at 3T. Radiology 2015; 274 (3):734–741. Liu Y, Liu H, Bai X, et al. Differentiation of metastatic from non-metastatic lymph nodes in patients with uterine cervical cancer using diffusionweighted imaging. Gynecol Oncol 2011; 122(1):19–24. Jensen JH, Helpern JA. MRI quantification of non-Gaussian water diffusion by kurtosis analysis. NMR Biomed 2010; 23(7):698–710.
IMAGING ASSESSMENT OF CERVICAL CARCINOMA
5. Jensen JH, Helpern JA, Ramani A, et al. Diffusional kurtosis imaging: the quantification of non-Gaussian water diffusion by means of magnetic resonance imaging. Magn Reson Med 2005; 53(6):1432–1440. 6. Wang P, Thapa D, Wu G, et al. A study on diffusion and kurtosis features of cervical cancer based on non-Gaussian diffusion weighted model. Magn Reson Imaging 2018; 47:60–66. 7. Le Bihan D. Apparent diffusion coefficient and beyond: what diffusion MR imaging can tell us about tissue structure. Radiology 2013; 268 (2):318–322. 8. Minosse S, Marzi S, Piludu F, et al. Correlation study between DKI and conventional DWI in brain and head and neck tumors. Magn Reson Imaging 2017; 42:114–122. 9. Falk Delgado A, Nilsson M, van Westen D, et al. Glioma grade discrimination with MR diffusion kurtosis imaging: a meta-analysis of diagnostic accuracy. Radiology 2018; 287(1):119–127. 10. Wang X, Tu N, Qin T, et al. Diffusion kurtosis imaging combined with DWI at 3-T MRI for detection and assessment of aggressiveness of prostate cancer. AJR Am J Roentgenol 2018; 211(4):797–804. 11. Tamada T, Prabhu V, Li J, et al. Prostate cancer: diffusion-weighted MR imaging for detection and assessment of aggressiveness-comparison between conventional and kurtosis models. Radiology 2017; 284(1):100–108. 12. Winfield JM, Orton MR, Collins DJ, et al. Separation of type and grade in cervical tumours using non-mono-exponential models of diffusionweighted MRI. Eur Radiol 2017; 27(2):627–636. 13. Dia AA, Hori M, Onishi H, et al. Application of non-Gaussian water diffusional kurtosis imaging in the assessment of uterine tumors: a preliminary study. PLoS One 2017; 12(11):e0188434. 14. Kurman RJ, Carcangiu ML, Herrington CS, et al. WHO classification of tumours of female reproductive organs. France: Lyon: International Agency for Research on Cancer, 2014. 15. Bhatla N, Berek JS, Cuello Fredes M, et al. Revised FIGO staging for carcinoma of the cervix uteri. Int J Gynaecol Obstet 2019; 145(1):129–135. 16. Moseley ME, Cohen Y, Kucharczyk J, et al. Diffusion-weighted MR imaging of anisotropic water diffusion in cat central nervous system. Radiology 1990; 176(2):439–445. 17. Yue W, Meng N, Wang J, et al. Comparative analysis of the value of diffusion kurtosis imaging and diffusion-weighted imaging in evaluating the histological features of endometrial cancer. Cancer Imaging 2019; 19(1):9. 18. Yamada I, Sakamoto J, Kobayashi D, et al. Diffusion kurtosis imaging of endometrial carcinoma: correlation with histopathological findings. Magn Reson Imaging 2019; 57:337–346. 19. Sun K, Chen X, Chai W, et al. Breast cancer: diffusion kurtosis MR imaging-diagnostic accuracy and correlation with clinical-pathologic factors. Radiology 2015; 277(1):46–55. 20. Tamura C, Shinmoto H, Soga S, et al. Diffusion kurtosis imaging study of prostate cancer: preliminary findings. J Magn Reson Imaging 2014; 40 (3):723–729. 21. Quentin M, Pentang G, Schimmoller L, et al. Feasibility of diffusional kurtosis tensor imaging in prostate MRI for the assessment of prostate cancer: preliminary results. Magn Reson Imaging 2014; 32(7):880–885. 22. Barrett T, McLean M, Priest AN, et al. Diagnostic evaluation of magnetization transfer and diffusion kurtosis imaging for prostate cancer detection in a re-biopsy population. Eur Radiol 2018; 28(8):3141–3150. 23. Exner M, Kuhn A, Stumpp P, et al. Value of diffusion-weighted MRI in diagnosis of uterine cervical cancer: a prospective study evaluating the benefits of DWI compared to conventional MR sequences in a 3T environment. Acta Radiol 2016; 57(7):869–877. 24. Guan Y, Shi H, Chen Y, et al. Whole-lesion histogram analysis of apparent diffusion coefficient for the assessment of cervical cancer. J Comput Assist Tomogr 2016; 40(2):212–217. 25. Kuang F, Ren J, Zhong Q, et al. The value of apparent diffusion coefficient in the assessment of cervical cancer. Eur Radiol 2013; 23(4):1050–1058. 26. Hoogendam JP, Klerkx WM, de Kort GA, et al. The influence of the bvalue combination on apparent diffusion coefficient based differentiation between malignant and benign tissue in cervical cancer. J Magn Reson Imaging 2010; 32(2):376–382. 27. Payne GS, Schmidt M, Morgan VA, et al. Evaluation of magnetic resonance diffusion and spectroscopy measurements as predictive biomarkers in stage 1 cervical cancer. Gynecol Oncol 2010; 116(2):246–252. 28. Chen J, Zhang Y, Liang B, et al. The utility of diffusion-weighted MR imaging in cervical cancer. Eur J Radiol 2010; 74(3):e101–e106. 29. Kuang F, Yan Z, Li H, et al. Diagnostic accuracy of diffusion-weighted MRI for differentiation of cervical cancer and benign cervical lesions at
7
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
WANG ET AL
30.
31.
32.
33.
34.
8
3.0T: comparison with routine MRI and dynamic contrast-enhanced MRI. J Magn Reson Imaging 2015; 42(4):1094–1099. Liu Y, Ye Z, Sun H, et al. Clinical application of diffusion-weighted magnetic resonance imaging in uterine cervical cancer. Int J Gynecol Cancer 2015; 25(6):1073–1078. Karunya RJ, Tharani P, John S, et al. Role of functional magnetic resonance imaging derived parameters as imaging biomarkers and correlation with clinicopathological features in carcinoma of uterine cervix. J Clin Diagn Res 2017; 11(8):Xc06–xc11. Rosenkrantz AB, Padhani AR, Chenevert TL, et al. Body diffusion kurtosis imaging: basic principles, applications, and considerations for clinical practice. J Magn Reson Imaging 2015; 42(5):1190–1202. Liu Y, Ye Z, Sun H, et al. Grading of uterine cervical cancer by using the ADC difference value and its correlation with microvascular density and vascular endothelial growth factor. Eur Radiol 2013; 23 (3):757–765. Chen T, Li Y, Lu SS, et al. Quantitative evaluation of diffusion-kurtosis imaging for grading endometrial carcinoma: a comparative study
35.
36.
37.
38. 39.
with diffusion-weighted imaging. Clin Radiol 2017; 72(11):995.e911– 995.e920. Zhu L, Pan Z, Ma Q, et al. Diffusion kurtosis imaging study of rectal adenocarcinoma associated with histopathologic prognostic factors: preliminary findings. Radiology 2017; 284(1):66–76. Wen Z, Chen Y, Yang X, et al. Application of magnetic resonance diffusion kurtosis imaging for distinguishing histopathologic subtypes and grades of rectal carcinoma. Cancer Imaging 2019; 19(1):8. Sun Y, Xiao Q, Hu F, et al. Diffusion kurtosis imaging in the characterisation of rectal cancer: utilizing the most repeatable region-of-interest strategy for diffusion parameters on a 3T scanner. Eur Radiol 2018; 28 (12):5211–5220. McCluggage WG. Towards developing a meaningful grading system for cervical squamous cell carcinoma. J Pathol Clin Res 2018; 4(2):81–85. Noviello MB, Silva-Filho AL, Traiman P, et al. Inter- and intraobserver variability in the assessment of tumor grade and lymphovascular space invasion in patients with squamous cell carcinoma of the cervix. Eur J Obstet Gynecol Reprod Biol 2008; 138(2):246–248.