- Email: [email protected]
Diffusion-weighted magnetic resonance imaging in the characterization of testicular germ cell neoplasms: Effect of ROI methods on apparent diffusion coefficient values and interobserver variability
European Journal of Radiology 89 (2017) 1–6
Contents lists available at ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
Editorial Musings
Diffusion-weighted magnetic resonance imaging in the characterization of testicular germ cell neoplasms: Effect of ROI methods on apparent diffusion coefficient values and interobserver variability Athina C. Tsili a,∗ , Alexandra Ntorkou a , Loukas Astrakas b , Vasilis Xydis a , Stavros Tsampalas c , Nikolaos Sofikitis c , Maria I. Argyropoulou a a
Department of Clinical Radiology, Medical School, University of Ioannina, University Campus, 45110, Ioannina, Greece Department of Medical Physics, Medical School, University of Ioannina, University Campus, 45110, Ioannina, Greece c Department of Urology, Medical School, University of Ioannina, University Campus, 45110, Ioannina, Greece b
a r t i c l e
i n f o
Article history: Received 2 November 2016 Received in revised form 9 January 2017 Accepted 17 January 2017 Keywords: MR imaging Diffusion Apparent diffusion coefficient ADC Testicular germ cell neoplasms ROI
a b s t r a c t Introduction: To evaluate the difference in apparent diffusion coefficient (ADC) measurements at diffusion-weighted (DW) magnetic resonance imaging of differently shaped regions-of-interest (ROIs) in testicular germ cell neoplasms (TGCNS), the diagnostic ability of differently shaped ROIs in differentiating seminomas from nonseminomatous germ cell neoplasms (NSGCNs) and the interobserver variability. Materials and methods: Thirty-three TGCNs were retrospectively evaluated. Patients underwent MR examinations, including DWI on a 1.5-T MR system. Two observers measured mean tumor ADCs using four distinct ROI methods: round, square, freehand and multiple small, round ROIs. The interclass correlation coefficient was analyzed to assess interobserver variability. Statistical analysis was used to compare mean ADC measurements among observers, methods and histologic types. Results: All ROI methods showed excellent interobserver agreement, with excellent correlation (P < 0.001). Multiple, small ROIs provided the lower mean ADC in TGCNs. Seminomas had lower mean ADC compared to NSGCNs for each ROI method (P < 0.001). Round ROI proved the most accurate method in characterizing TGCNS. Conclusion: Interobserver variability in ADC measurement is excellent, irrespective of the ROI shape. Multiple, small round ROIs and round ROI proved the more accurate methods for ADC measurement in the characterization of TGCNs and in the differentiation between seminomas and NSGCNs, respectively. © 2017 Elsevier B.V. All rights reserved.
1. Introduction Diffusion-weighted (DW) magnetic resonance imaging (MRI) assesses the random motion of water molecules. Water movement is completely random in a totally unrestricted environment, and this phenomenon is called Brownian motion or free diffusion [1,2]. Within biologic tissues, the movement of water is not completely random, but rather, is impeded by the interaction with tissue compartments, cell membranes, and intracellular organelles [1,2].
∗ Corresponding author. E-mail addresses: a [email protected], [email protected] (A.C. Tsili), [email protected] (A. Ntorkou), [email protected] (L. Astrakas), [email protected] (V. Xydis), [email protected] (S. Tsampalas), [email protected] (N. Sofikitis), [email protected] (M.I. Argyropoulou). http://dx.doi.org/10.1016/j.ejrad.2017.01.017 0720-048X/© 2017 Elsevier B.V. All rights reserved.
DWI has the potential to improve tissue characterization, especially when interpreted in combination with conventional MRI findings. Lesion detection and characterization mainly depends on the extent of tissue cellularity, and increased cellularity is associated with restricted diffusion and reduced apparent diffusion coefficient (ADC) [1,2]. Recent work exploring the role of DWI and ADC mapping in the assessment of testicular diseases includes the characterization of testicular mass lesions, the diagnosis of testicular torsion, the detection and localization of impalpable testes and the detection of testicular fibrosis in patients with varicocele [3–12]. Testicular carcinoma represents approximately 1% of male neoplasms and 5% of urologic malignancies, with 8.720 new cases of testicular cancer estimated to occur in the USA during 2016 and 380 deaths related to the disease [13,14]. Most (95%) testicular carcinomas are germ cell neoplasms (TGCNs), arising from the
2
A.C. Tsili et al. / European Journal of Radiology 89 (2017) 1–6
germinal epithelium of the seminiferous tubules [15,16]. TGCNs are fairly evenly split between seminomas and nonseminomatous GCNs (NSGCNs). The gross morphology and histopathologic characteristics of these two categories of tumors are different [15,16]. Conventional MRI and DWI characteristics have been reported closely to correlate with histopathologic differences [7,17,18]. Radical orchiectomy is the treatment of choice in cases of TGCNs and should be carried out without any delay [19]. However, in cases of disseminated disease and/or life-threatening metastases, it is recommended to start with chemotherapy, and orchiectomy should follow when clinical stabilization occurs [19]. Histology by means of biopsy provides the definite diagnosis in these patients [19]. A non-invasive method helping in the preoperative characterization of the histologic type of TGCN would be extremely useful in these cases. Measurements of ADC values may be subject to several sources of variability, such as patient-related factors, image-related factors and systematic reader errors, the latter including the choice of region-of-interest (ROI), intraobserver and interobserver repeatability [20,21]. There are a few reports regarding the influence of manual ROI shape on interobserver variability of visceral tumor ADC measurements [22–28]. As to our knowledge, there is no study evaluating the effect of using ROIs of different shapes to measure ADC of TGCNs. The purpose of the present study was to evaluate the difference in ADC measurements at DW MRI of differently shaped ROIs in TGCNS, to investigate the diagnostic ability of differently shaped ROIs in differentiating seminomas from NSGCNs and to assess interobserver variability. 2. Materials and methods 2.1. Patient population In this retrospective study, 41 consecutive patients with pathologically proven TGCNs were included. Patients underwent MR examinations including DWI using a 1.5.T MR system from July 2008 to February 2016. Of these, eight patients were excluded due to the following: presence of artifacts (n = 2); mainly hemorrhagic tumor (n = 3); necrotic tumor, not histologically characterized (n = 1); and mainly multicystic tumor, without obvious areas of restricted diffusion (n = 2). Therefore, the study cohort included 33 patients (mean age, 34 years; range, 20–85 years). Histology reported the presence of 20 seminomas and 13 NSGCNS (embryonal carcinoma, n = 7; seminoma and embryonal carcinoma, n = 2; seminoma, polyembryoma and teratoma, n = 1; embryonal carcinoma, teratoma and yolk sac tumor, n = 2; seminoma, embryonal carcinoma, teratoma and yolk sac tumor, n = 1). In all cases, the interval between MRI and radical orchiectomy was less than two weeks. Institutional Review Board approval was obtained for the study, and the requirement for informed consent was waived. 2.2. MR technique All MR examinations were performed on a 1.5-T scanner (Philips Medical Systems, Cleveland, OH, USA), with the use of a circular surface coil. Patients were examined in the supine position, with the testes placed at a similar distance from the coil, by placing a towel beneath them, and the penis draped on the lower anterior abdominal wall. Transverse T1-weighted (T1WI) sequences and axial, sagittal and coronal T2-weighted (T2WI) were used for data analysis. DWI was performed along the axial plane during quiet breathing, with b-values of 0 and 900 s/mm−2 . The orientation and location of DWI images were identical to the conventional
transverse images. A total acquisition time of 29 s was obtained to cover the entire scrotal area. No parallel imaging was used. Dynamic contrast-enhanced (DCE) subtraction MRI in the coronal plane was followed, using a three-dimensional (3D) fast field-echo (FFE) technique. Peripheral intravenous tubing with a 22-gauge catheter placed in a subcutaneous vein of the antecubital fossa was performed. Dynamic images were obtained before and after a rapid injection of 0.2 mmol of gadopentetic acid (Magnevist; Bayer Healthcare, Berlin, Germany) per kilogram of bodyweight, performed manually and followed by a flush of 20 mL of physiologic saline solution. Seven consecutive imaging sets were acquired immediately after the start of contrast injection, with no interval between them. The data set obtained before administration of contrast material was used as a mask for subsequent image subtraction. Each of the seven data sets obtained after contrast medium administration was subtracted section by section, using commercially available software (Philips Medical Systems, Cleveland, OH, USA). The MRI protocol used in this study is described in Table 1. 2.3. Image analysis The MR images were independently analyzed by two radiologists (ACT and AN, with 12 and three years of experience, each in scrotal MRI). Both readers were blinded to the histopathologic data. Post processing was performed on the institutional picture archiving and communication system (PACS). ADC maps were created by the MRI system software and sent to the hospital PACS. The maximal tumor diameter was measured on T2WI. The presence of carcinoma on the ADC maps was diagnosed as areas of low signal intensity, with the aid of the corresponding T2WI and subtracted dynamic contrast-enhanced T1WI. Measurements were performed on the ADC map that contained the largest tumor crosssection, using four differently shaped regions-of-interest (ROIs): round, square, freehand ROI and five small, round ROIs at the same time, using the same section (Fig. 1). Mean ADCs were obtained for each ROI in each case. Round and square ROIs were defined to be as large as possible. The freehand ROI was drawn along the borders of the entire neoplasm. The five small round ROIs were placed within the same slice, with care not to overlap each other and the mean value was subsequently calculated for mean ADC. All ROIs were carefully drawn to exclude artifacts and areas of hemorrhage and/or necrosis, with the aid of the corresponding T1WI, T2WI and subtracted dynamic contrast-enhanced T1WI. 2.4. Statistical analyses The Mann-Whitney U Test was used to compare size differences between seminomas and NSGCNs. Intraclass correlation coefficient (ICC, 0.00–0.20 poor, 0.21–0.40 fair, 0.41–0.60 moderate, 0.61–0.80 good and 0.81–1.00 excellent correlation) was used to assess interobserver reliability for tumor ADC measurements from the two readers for each individual ROI method. A general linear model was used to compare seminoma versus NSGCN ADC measurements. In this statistical approach, measurements of the two observers and the four methods were treated as “repeated measures”. Equivalently observer and method were considered within-subjects factors and pathology type (either seminoma or NSGCN) was considered a within-subject factor, which defines the two conditions within each subject. Repeated measures ANOVA and the Bonferonni correction for post hoc testing were applied to compare ADC measurements among observers, methods and histologic types. For each observer, a stepwise backward binary logistic analysis was performed to assess independent predictors of pathology type among the four different methods of measurements. ROC analysis, using measurements from both observers, was performed and the area under the curve (AUC) was calculated to assess how
A.C. Tsili et al. / European Journal of Radiology 89 (2017) 1–6
3
Table 1 MRI protocol (Abbreviations: TR: time repetition; TE: time echo; FOV: field of view; T1WI: T1-weighted imaging; T2WI: T2-weighted imaging; DWI: diffusion-weighted imaging; 3D FFE: three-dimensional fast field-echo; c.m. i.v., contrast material intravenously). Sequences
spin-echo T1WI
fast-spin echo T2WI
DWI
3D FFE post-contrast
Plane TR (msec) TE (msec) Slice thickness (mm) Gap (mm) FOV (mm) Matrix (mm) b value (s m−2 ) Flip angle Dynamic scans c.m i.v. (ml kg−1 )
transverse 500–650 13–15 3–4 0.5 240 × 270 180 × 256
transverse, sagittal, coronal 4000 100–120 3–4 0.5 240 × 270 180 × 256
transverse 3900 115 3–4 0.5 240 × 270 180 × 256 0, 900
coronal 9 4.1 4 0 219 × 219 256 × 256
35◦ 7 (every 60 sec) 0.2
Fig. 1. ADC measurements were performed on the ADC map using four distinct ROI methods: (a) round ROI; (b) square ROI; (c) freehand ROI; (d) and five, small round ROIs. Care was taken to avoid areas of necrosis with the aid of the corresponding (e) T2WI and (f) subtracted dynamic contrast-enhanced T1WI.
accurately each method can discriminate between seminoma and NSGCN cases. The maximum Youden Index (J) was used to derive the optimal cut-off ADC value which maximizes the sum of sensitivity and specificity. All statistical analysis was performed using SPSS v.20.0 (IBM Corp., Armonk, NY, USA). Two-tailed values of P < 0.05 were regarded as statistically significant. 3. Results The median size of seminomas and nonseminomas was 4.1 cm (range: 1.2–9.9 cm) and 4.7 cm (range: 1.5–11.5 cm), respectively. There was no size difference between the two groups (P = 0.73). The intraclass correlation coefficient between the two observers was high (ICC = 0.963, P < 0.001) assuring the reliability of the measurements. ICCs were 0.968 for round ROIs, 0.964 for square ROIs, 0.966 for freehand ROIs, and 0.951 for the multiple small, round ROIs. All ROI methods showed excellent correlation (P < 0.001). Similarly, ANOVA showed no difference (P = 0.658) between mean ADC of the first (mean ADC ± std error = 0.742 ± 0.027 × 10−3 mm2 /s) and the second observer (mean ADC ± std err = 0.747 ± 0.024 × 10−3 mm2 /s). Mean ADC ± std (×10−3 mm2 /s) was 0.749 ± 0.025 for round ROI, 0.746 ± 0.025 for square ROI, 0.765 ± 0.025 for freehand ROI and 0.718 ± 0.024 for multiple, small round ROIs (Table 2). Post hoc
comparisons between mean values of the measurement methods showed that multiple, small round ROIs provides lower ADC (Table 2) compared to the others. Also, the mean ADC ± std (×10−3 mm2 /s) of seminoma (0.608 ± 0.031) was lower (P < 0.001) compared to the mean ADC of NSGCN (0.88 ± 0.038). Similarly, separately for each ROI method, the mean ADC of seminoma was lower (P < 0.001 in all cases) compared to the mean ADC of NSGCN (Table 3, Fig. 2). Logistic regression analysis, both for the first and the second observer, showed that among the various ROI methods, only the round ROI is independent predictor of the pathology type (P = 0.008 in both cases). Similarly, ROC analysis showed that round ROI has the best discriminative accuracy with AUC = 0.910 (square ROI, AUC = 0.897; freehand ROI, AUC = 0.888; and, multiple small ROIs, AUC = 0.872) (Fig. 3). The optimal cut-off ADC using the round ROI was found 0.73 × 10−3 mm2 /s (J = 0.746). 4. Discussion Our results demonstrate that all ROI methods provide excellent interobserver agreement with excellent correlation when measuring ADC in testicular germ cell neoplasms. Using four distinct ROI methods, ADC of TGCNs was within the range reported in previous publications [3–6]. However, mean ADC of TGCNs was lower for multiple, small round ROIs than that measured with round, square
4
A.C. Tsili et al. / European Journal of Radiology 89 (2017) 1–6
Table 2 Mean ADC measurements for each method and their post hoc comparison. ROI methods
ADC (mean ± std. error 10−3 mm2 /s)
P(comparison to circular)
P(comparison to square)
P(comparison to freehand)
round square freehand multiple, small round ROIs
0.749 ± 0.025 0.746 ± 0.025 0.765 ± 0.025 0.718 ± 0.024
1.000 0.377 0,006
0.149 0,013
<0.001
Table 3 Mean ADC measurements for each method and histologic type. ROI methods
Histologic type
ADC (mean ± std. error10−6 mm2 /s)
P
round
seminoma NSGCN seminoma NSGCN seminoma NSGCN seminoma NSGCN
0.603 ± 0.010 0.894 ± 0.194 0.606 ± 0.099 0.885 ± 0.198 0.634 ± 0.091 0.896 ± 0.201 0.59 ± 0.082 0.846 ± 0.201
<0.001
square freehand multiple, small round ROIs
Fig. 2. Boxplots of ADC measured by both observers, for each histologic type and for each of the four ROI methods. Dark horizontal lines represent the mean values; the box represents the 25th and 75th percentiles and the whiskers the 5th and 95th percentiles.
Fig. 3. ROC curves for each ROI method.
<0.001 <0.001 <0.001
and freehand ROI. The multiple, small round ROIs include only the most viable, solid part of the neoplasm, seen as areas of highest restriction on the ADC maps, resulting in lower ADC values. In these measurements, areas of hemorrhage and/or necrosis along with the edges of the tumor are more likely to be excluded, than with the other three methods. In cases of TGCNs that require immediate chemotherapy instead of radical orchiectomy, the characterization of the histologic type of testicular tumor is of great importance [19]. Conventional MRI findings have been found closely to correlate with the histopathologic characteristics of TGCNs [17,18]. Recently, measurements of ADC provided an additional tool in differentiating seminomas from NSGCNs [7]. In a previous report of 26 histologically proven TGCNs, we found ADC useful in the preoperative discrimination of seminomas from nonseminomatous tumors. Specifically, the mean ADC ± std (×10−3 mm2 /s) of seminomas (0.59 ± 0.009) was significantly lower compared to that of NSGCTs (0.90 ± 0.33) in this report [7]. This is in accordance with present results. Histopathologic characteristics of seminomatous tumors, including presence of large, uniform cells with large nuclei, prominent nucleoli and abundant cytoplasm, arranged in nests, outlined by lymphocytebearing fibrovascular septa probably account for the significant restriction in the extracellular movement of water protons in cases of seminomas when compared to NSGCNs [15,16]. Mean ADCs from all ROI methods showed significant differences between seminomas and NSGCNs, with round ROI providing the most accurate technique for the characterization of the histologic subtype of TGCNs, with an optimal cut-off value of 0.73 × 10−3 mm2 /s. This is in discrepancy with the first part of our study, where multiple, small round ROIs proved the most useful technique to characterize TGCNs. The frequent presence of areas of necrosis and/or hemorrhage in nonseminomas and the coexistence of fibrovascular septa in seminomas might probably create difficulties in the delineation of small, round ROIs when measuring ADC values, and the therefore in the histologic characterization of TGCNs. Various publications on DWI have established its role in the detection and characterization of malignancies [1,2]. In most studies, ADC measurements are performed by one observer and a variety of methods for ROI shape, including round, square, polygonal and small round ROIs. Methods for ROI placement also vary, including whole tumor volume, a single-slice or small tumor samples [20,21]. There are prior reports that have assessed the interobserver reliability and the effect of different ROI methods in ADC measurements in various malignancies [22–28]. In a retrospective study of 69 patients with endometrial carcinoma, Inoue
A.C. Tsili et al. / European Journal of Radiology 89 (2017) 1–6
et al. investigated the influence of four manual ROI methods, similar to ours, that is round, square, freehand and multiple, small ROIs on both the minimum and mean ADC tumor measurements and also assessed interobserver variability [22]. The authors found that all four ROI methods had excellent interobserver agreement, with excellent correlation, as proved also in our study. Each minimum ADC was significantly different, except between square and round ROI. Mean ADCs showed significant differences only between freehand ROI and the other methods. This study concluded that mean ADC provides more stable results, with round or square ROI representing a more suitable method for ADC measurements [22]. In a recent study of 54 women with ovarian tumors with a solid component, the effect of similar ROI methods on ADC values and their diagnostic ability in differentiating benign from malignant tumors were assessed [23]. The interobserver reliability in ADC measurement was good or excellent, with the exception of the square ROI. Freehand ROI showed the lowest and round ROI showed the highest minimum ADC. Freehand ROI showed the highest mean ADC, and round and square ROIs showed the lowest mean ADC. The authors concluded that ROI shape influences the ADC measurements and the optimal cutoff ADC used to differentiate benign from malignant ovarian tumors [23]. In another study, the influence of three ROI protocols, including ‘whole-volume’, ‘single-slice’, using a single freehand ROI and ‘small solid samples’, using three round/oval-shaped ROIs placed within the most solid tumoral part of three independent tumourcontaining slices was assessed, in 46 patients with locally advanced rectal carcinoma [25]. The authors concluded that variations in ROI size and positioning had a significant effect on tumour ADC and interobserver variability. ADC of the whole tumour volume provided the most reproducible results in the same report [25]. Ahlawat et al. assessed the interobserver reliability of three different manual ROI protocols for determining the mean and minimum ADC in 73 musculoskeletal soft tissue masses compared with whole tumor volume ADC [27]. The authors showed that a fast selective observer-based method provided similar results to whole tumor volume analysis with high interobserver agreement and a significant decrease in measurement time [27]. Another report by Nogueira et al. compared two different methods of ROIs on ADC measurement in 39 breast lesions, using a small area of highest restriction and a large ROI, including the whole lesion. Although, interobserver variability was low for both methods, small ROIs showed less overlap in ADC and higher reproducibility, suggesting that this method might improve lesion characterization [28]. A considerable influence of the ROI method on mean ADC and the interobserver variability in pancreatic adenocarcinoma has been reported by Liu et al. [26]. The authors in a prospective study of 21 patients with surgically-proven pancreatic adenocarcinoma compared the influence of three techniques of ROI placement for ADC measurements, including whole-volume, single-slice, and small solid samples of tumor [26]. The study concluded that ADC based on the small solid samples of tumor provided the highest diagnostic performance in assessing pancreatic adenocarcinoma, and this was in accordance with our results [26]. DWI is usually performed before the intravenous administration of a gadolinium chelate contrast agent, as it was in the present study. Up to now, there are limited data regarding the effectiveness of DWI after gadolinium administration, with studies reporting no significant change of ADC values after contrast material administration [29,30]. This is important in cases when DWI fails or data interpretation is difficult due to the presence of artifacts, allowing DW sequences to be repeated after contrast administration [29,30]. This study has some limitations to be acknowledged. First, it is a retrospective review of relatively small patient numbers. No subgroup analysis of the different histologic subtypes of NSGCNS was performed due to the small number of cases. Further studies
5
are required to validate the present results in a larger study cohort. Nonseminomas comprise a heterogeneous group of neoplasms and histologic diversity of these tumors could result in ADC differences. Second, we did not use the whole tumor volume ROI in our report. Although, the whole-volume ROI is considered a reliable method, this technique is time-consuming and not easily applicable in routine clinical practice. Another limitation is that the measurement of ADC by the four ROI methods was performed at the same time, and therefore, there might be some bias. Furthermore, each reader independently had drawn the largest cross-section of the tumor. Therefore, ROIs might be placed in different areas by each reader. Finally, intraobserver agreement was not assessed. 5. Conclusion In concluding, ROI shape has no marked effect on interobserver variability in ADC measurements in testicular germ cell neoplasms. Multiple, small round ROIs and round ROI have been proved more useful methods for ADC measurement in characterizing TGCNs and in differentiating seminomas from nonseminomatous tumors, respectively. Although histology remains the gold standard for the characterization of TGCNs, larger prospective studies could probably justify the potential role of MRI, including DWI as a non-invasive tool, helpful in the histologic differentiation of these neoplasms. Conflict of interest I declare no financial or commercial interests related to this paper. References [1] T.C. Kwee, T. Takahara, R. Ochiai, et al., Whole-body diffusion-weighted magnetic resonance imaging, Eur. J. Radiol. 70 (2009) 409–417. [2] R. Bammer, Basic principles of diffusion-weighted imaging, Eur. J. Radiol. 45 (2003) 169–184. [3] A.C. Tsili, M.I. Argyropoulou, D. Giannakis, S. Tsampalas, N. Sofikitis, K. Tsampoulas, Diffusion-weighted MR imaging of normal and abnormal scrotum: preliminary results, Asian J. Androl. 14 (2012) 649–654. [4] A.M. Algebally, H.I. Tantawy, R.R. Yousef, W. Szmigielski, A. Darweesh, Advantage of adding diffusion weighted imaging to routine MRI examinations in the diagnostics of scrotal lesions, Pol. J. Radiol. 80 (2015) 442–449. [5] A.C. Tsili, D. Giannakis, A. Sylakos, et al., Apparent diffusion coefficient values of normal testis and variations with age, Asian J. Androl. 16 (2014) 493–497. [6] A.C. Tsili, A. Ntorkou, D. Baltogiannis, et al., The role of apparent diffusion coefficient values in detecting testicular intraepithelial neoplasia: preliminary results, Eur. J. Radiol. 84 (2015) 828–833. [7] A.C. Tsili, A. Sylakos, A. Ntorkou, et al., Apparent diffusion coefficient values and dynamic contrast enhancement patterns in differentiating seminomas from nonseminomatous testicular neoplasms, Eur. J. Radiol. 84 (2015) 1219–1226. [8] D. Maki, Y. Watanabe, M. Nagayama, et al., Diffusion-weighted magnetic resonance imaging in the detection of testicular torsion: feasibility study, J. Magn. Reson. Imaging 34 (2011) 1137–1142. [9] Y. Watanabe, M. Nagayama, A. Okumura, et al., MR imaging of testicular torsion: features of testicular hemorrhagic necrosis and clinical outcomes, J. Magn. Reson. Imaging 26 (2007) 100–108. [10] M. Kantarci, S. Doganay, A. Yalcin, Y. Aksoy, B. Yilmaz-Cankaya, B. Salman, Diagnostic performance of diffusion-weighted MRI in the detection of nonpalpable undescended testes: comparison with conventional MRI and surgical findings, AJR Am. J. Roentgenol. 195 (2010) W268–73. [11] S. Emad-Eldin, N. Abo-Elnagaa, S.A. Hanna, A.H. Abdel-Satar, The diagnostic utility of combined diffusion-weighted imaging and conventional magnetic resonance imaging for detection and localization of non palpable undescended testes, J. Med. Imaging Radiat. Oncol. 60 (2016) 344–351. [12] E. Karakas, O. Karakas, N. Cullu, et al., Diffusion-weighted MRI of the testes in patients with varicocele: a preliminary study, AJR Am. J. Roentgenol. 202 (2014) 324–328. [13] P. Albers, W. Albrecht, F. Algaba, et al., Guidelines on Testicular Cancer, European Association of Urology, 2015, Available at: http://uroweb.org/wpcontent/uploads/11-Testicular-Cancer LR1.pdf. [14] American Cancer Society. Cancer facts and figures 2016. Available at: http:// www.cancer.org/acs/groups/content/@research/documents/document/acspc047079.pdf. [15] T.M. Ulbright, D.M. Berney, et al., Testicular and paratesticular tumors, in: S.E. Mills, D. Carter, J.K. Greenson (Eds.), Sternberg’s Diagnostic Surgical
6
[16]
[17]
[18]
[19]
[20]
[21]
[22]
A.C. Tsili et al. / European Journal of Radiology 89 (2017) 1–6 Pathology, 5th edition, Lippincott Williams & Wilkins, Philadelphia, 2010, pp. 1944–2004. WHO histological classification of testis tumours, in: J.N. Eble, G. Sauter, J.I. Epstein, I.A. Sesterhenn (Eds.), Pathology & Genetics. Tumours of the Urinary System and Male Genital Organs, IARC Press, Lyon, 2004, pp. 212–274. J.O. Johnson, R.F. Mattrey, J. Phillipson, Differentiation of seminomatous from nonseminomatous testicular tumors with MR imaging, AJR Am. J. Roentgenol. 154 (1990) 539–543. A.C. Tsili, C. Tsampoulas, X. Giannakopoulos, et al., MRI in the histologic characterization of testicular neoplasms, AJR Am. J. Roentgenol. 189 (2007) W331–7. J. Oldenburg, S.D. Fosså, J. Nuver, et al., Testicular seminoma and non-seminoma: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up, Ann. Oncol. 24 (Suppl. 6) (2013) vi125–vi132. E. Giannotti, S. Waugh, L. Priba, Z. Davis, E. Crowe, S. Vinnicombe, Assessment and quantification of sources of variability in breast apparent diffusion coefficient (ADC) measurements at diffusion weighted imaging, Eur. J. Radiol. 84 (2015) 1729–1736. P. Clauser, M. Marcon, M. Maieron, C. Zuiani, M. Bazzocchi, P.A. Baltzer, Is there a systematic bias of apparent diffusion coefficient (ADC) measurements of the breast if measured on different workstations? An inter- and intra-reader agreement study, Eur. Radiol. 26 (2016) 2291–2296. C. Inoue, S. Fujii, S. Kaneda, et al., Apparent diffusion coefficient (ADC) measurement in endometrial carcinoma: effect of region of interest methods on ADC values, J. Magn. Reson. Imaging 40 (2014) 157–161.
[23] N. Mukuda, S. Fujii, C. Inoue, et al., Apparent diffusion coefficient (ADC) measurement in ovarian tumor: effect of region-of-interest methods on ADC values and diagnostic ability, J. Magn. Reson. Imaging 43 (2016) 720–725. [24] I.M. Blazic, G.B. Lilic, M.M. Gajic, Quantitative assessment of rectal cancer response to neoadjuvant combined chemotherapy and radiation therapy: comparison of three methods of positioning region of interest for ADC measurements at diffusion-weighted MR imaging, Radiology 2 (2016) 151908 [Epub ahead of print]. [25] D.M.J. Lambregts, G.L. Beets, M. Maas, et al., Tumour ADC measurements in rectal cancer: effect of ROI methods on ADC values and interobserver variability, Eur. Radiol. 21 (2011) 2567–2574. [26] L. Liu, C. Ma, J. Li, et al., Comparison of the diagnostic performances of three techniques of ROI placement for ADC measurements in pancreatic adenocarcinoma, Acad. Radiol. 22 (2015) 1385–1392. [27] S. Ahlawat, P. Khandheria, F. Del Grande, et al., Interobserver variability of selective region-of-interest measurement protocols for quantitative diffusion weighted imaging in soft tissue masses: comparison with whole tumor volume measurements, J. Magn. Reson. Imaging 43 (2016) 446–454. [28] L. Nogueira, S. Brandão, E. Matos, et al., Region of interest demarcation for quantification of the apparent diffusion coefficient in breast lesions and its interobserver variability, Diagn. Interv. Radiol. 21 (2015) 123–127. [29] C.L. Wang, Y.W. Chea, D.T. Boll, et al., Effect of gadolinium chelate contrast agents on diffusion weighted MR imaging of the liver, spleen, pancreas and kidney at 3 T, Eur. J. Radiol. 80 (2011) e1–7. [30] S. Kinner, L. Umutlu, S. Blex, et al., Diffusion weighted MR imaging in patients with HCC and liver cirrhosis after administration of different gadolinium contrast agents: is it still reliable, Eur. J. Radiol. 81 (2012) e625–8.
Contents lists available at ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
Editorial Musings
Diffusion-weighted magnetic resonance imaging in the characterization of testicular germ cell neoplasms: Effect of ROI methods on apparent diffusion coefficient values and interobserver variability Athina C. Tsili a,∗ , Alexandra Ntorkou a , Loukas Astrakas b , Vasilis Xydis a , Stavros Tsampalas c , Nikolaos Sofikitis c , Maria I. Argyropoulou a a
Department of Clinical Radiology, Medical School, University of Ioannina, University Campus, 45110, Ioannina, Greece Department of Medical Physics, Medical School, University of Ioannina, University Campus, 45110, Ioannina, Greece c Department of Urology, Medical School, University of Ioannina, University Campus, 45110, Ioannina, Greece b
a r t i c l e
i n f o
Article history: Received 2 November 2016 Received in revised form 9 January 2017 Accepted 17 January 2017 Keywords: MR imaging Diffusion Apparent diffusion coefficient ADC Testicular germ cell neoplasms ROI
a b s t r a c t Introduction: To evaluate the difference in apparent diffusion coefficient (ADC) measurements at diffusion-weighted (DW) magnetic resonance imaging of differently shaped regions-of-interest (ROIs) in testicular germ cell neoplasms (TGCNS), the diagnostic ability of differently shaped ROIs in differentiating seminomas from nonseminomatous germ cell neoplasms (NSGCNs) and the interobserver variability. Materials and methods: Thirty-three TGCNs were retrospectively evaluated. Patients underwent MR examinations, including DWI on a 1.5-T MR system. Two observers measured mean tumor ADCs using four distinct ROI methods: round, square, freehand and multiple small, round ROIs. The interclass correlation coefficient was analyzed to assess interobserver variability. Statistical analysis was used to compare mean ADC measurements among observers, methods and histologic types. Results: All ROI methods showed excellent interobserver agreement, with excellent correlation (P < 0.001). Multiple, small ROIs provided the lower mean ADC in TGCNs. Seminomas had lower mean ADC compared to NSGCNs for each ROI method (P < 0.001). Round ROI proved the most accurate method in characterizing TGCNS. Conclusion: Interobserver variability in ADC measurement is excellent, irrespective of the ROI shape. Multiple, small round ROIs and round ROI proved the more accurate methods for ADC measurement in the characterization of TGCNs and in the differentiation between seminomas and NSGCNs, respectively. © 2017 Elsevier B.V. All rights reserved.
1. Introduction Diffusion-weighted (DW) magnetic resonance imaging (MRI) assesses the random motion of water molecules. Water movement is completely random in a totally unrestricted environment, and this phenomenon is called Brownian motion or free diffusion [1,2]. Within biologic tissues, the movement of water is not completely random, but rather, is impeded by the interaction with tissue compartments, cell membranes, and intracellular organelles [1,2].
∗ Corresponding author. E-mail addresses: a [email protected], [email protected] (A.C. Tsili), [email protected] (A. Ntorkou), [email protected] (L. Astrakas), [email protected] (V. Xydis), [email protected] (S. Tsampalas), [email protected] (N. Sofikitis), [email protected] (M.I. Argyropoulou). http://dx.doi.org/10.1016/j.ejrad.2017.01.017 0720-048X/© 2017 Elsevier B.V. All rights reserved.
DWI has the potential to improve tissue characterization, especially when interpreted in combination with conventional MRI findings. Lesion detection and characterization mainly depends on the extent of tissue cellularity, and increased cellularity is associated with restricted diffusion and reduced apparent diffusion coefficient (ADC) [1,2]. Recent work exploring the role of DWI and ADC mapping in the assessment of testicular diseases includes the characterization of testicular mass lesions, the diagnosis of testicular torsion, the detection and localization of impalpable testes and the detection of testicular fibrosis in patients with varicocele [3–12]. Testicular carcinoma represents approximately 1% of male neoplasms and 5% of urologic malignancies, with 8.720 new cases of testicular cancer estimated to occur in the USA during 2016 and 380 deaths related to the disease [13,14]. Most (95%) testicular carcinomas are germ cell neoplasms (TGCNs), arising from the
2
A.C. Tsili et al. / European Journal of Radiology 89 (2017) 1–6
germinal epithelium of the seminiferous tubules [15,16]. TGCNs are fairly evenly split between seminomas and nonseminomatous GCNs (NSGCNs). The gross morphology and histopathologic characteristics of these two categories of tumors are different [15,16]. Conventional MRI and DWI characteristics have been reported closely to correlate with histopathologic differences [7,17,18]. Radical orchiectomy is the treatment of choice in cases of TGCNs and should be carried out without any delay [19]. However, in cases of disseminated disease and/or life-threatening metastases, it is recommended to start with chemotherapy, and orchiectomy should follow when clinical stabilization occurs [19]. Histology by means of biopsy provides the definite diagnosis in these patients [19]. A non-invasive method helping in the preoperative characterization of the histologic type of TGCN would be extremely useful in these cases. Measurements of ADC values may be subject to several sources of variability, such as patient-related factors, image-related factors and systematic reader errors, the latter including the choice of region-of-interest (ROI), intraobserver and interobserver repeatability [20,21]. There are a few reports regarding the influence of manual ROI shape on interobserver variability of visceral tumor ADC measurements [22–28]. As to our knowledge, there is no study evaluating the effect of using ROIs of different shapes to measure ADC of TGCNs. The purpose of the present study was to evaluate the difference in ADC measurements at DW MRI of differently shaped ROIs in TGCNS, to investigate the diagnostic ability of differently shaped ROIs in differentiating seminomas from NSGCNs and to assess interobserver variability. 2. Materials and methods 2.1. Patient population In this retrospective study, 41 consecutive patients with pathologically proven TGCNs were included. Patients underwent MR examinations including DWI using a 1.5.T MR system from July 2008 to February 2016. Of these, eight patients were excluded due to the following: presence of artifacts (n = 2); mainly hemorrhagic tumor (n = 3); necrotic tumor, not histologically characterized (n = 1); and mainly multicystic tumor, without obvious areas of restricted diffusion (n = 2). Therefore, the study cohort included 33 patients (mean age, 34 years; range, 20–85 years). Histology reported the presence of 20 seminomas and 13 NSGCNS (embryonal carcinoma, n = 7; seminoma and embryonal carcinoma, n = 2; seminoma, polyembryoma and teratoma, n = 1; embryonal carcinoma, teratoma and yolk sac tumor, n = 2; seminoma, embryonal carcinoma, teratoma and yolk sac tumor, n = 1). In all cases, the interval between MRI and radical orchiectomy was less than two weeks. Institutional Review Board approval was obtained for the study, and the requirement for informed consent was waived. 2.2. MR technique All MR examinations were performed on a 1.5-T scanner (Philips Medical Systems, Cleveland, OH, USA), with the use of a circular surface coil. Patients were examined in the supine position, with the testes placed at a similar distance from the coil, by placing a towel beneath them, and the penis draped on the lower anterior abdominal wall. Transverse T1-weighted (T1WI) sequences and axial, sagittal and coronal T2-weighted (T2WI) were used for data analysis. DWI was performed along the axial plane during quiet breathing, with b-values of 0 and 900 s/mm−2 . The orientation and location of DWI images were identical to the conventional
transverse images. A total acquisition time of 29 s was obtained to cover the entire scrotal area. No parallel imaging was used. Dynamic contrast-enhanced (DCE) subtraction MRI in the coronal plane was followed, using a three-dimensional (3D) fast field-echo (FFE) technique. Peripheral intravenous tubing with a 22-gauge catheter placed in a subcutaneous vein of the antecubital fossa was performed. Dynamic images were obtained before and after a rapid injection of 0.2 mmol of gadopentetic acid (Magnevist; Bayer Healthcare, Berlin, Germany) per kilogram of bodyweight, performed manually and followed by a flush of 20 mL of physiologic saline solution. Seven consecutive imaging sets were acquired immediately after the start of contrast injection, with no interval between them. The data set obtained before administration of contrast material was used as a mask for subsequent image subtraction. Each of the seven data sets obtained after contrast medium administration was subtracted section by section, using commercially available software (Philips Medical Systems, Cleveland, OH, USA). The MRI protocol used in this study is described in Table 1. 2.3. Image analysis The MR images were independently analyzed by two radiologists (ACT and AN, with 12 and three years of experience, each in scrotal MRI). Both readers were blinded to the histopathologic data. Post processing was performed on the institutional picture archiving and communication system (PACS). ADC maps were created by the MRI system software and sent to the hospital PACS. The maximal tumor diameter was measured on T2WI. The presence of carcinoma on the ADC maps was diagnosed as areas of low signal intensity, with the aid of the corresponding T2WI and subtracted dynamic contrast-enhanced T1WI. Measurements were performed on the ADC map that contained the largest tumor crosssection, using four differently shaped regions-of-interest (ROIs): round, square, freehand ROI and five small, round ROIs at the same time, using the same section (Fig. 1). Mean ADCs were obtained for each ROI in each case. Round and square ROIs were defined to be as large as possible. The freehand ROI was drawn along the borders of the entire neoplasm. The five small round ROIs were placed within the same slice, with care not to overlap each other and the mean value was subsequently calculated for mean ADC. All ROIs were carefully drawn to exclude artifacts and areas of hemorrhage and/or necrosis, with the aid of the corresponding T1WI, T2WI and subtracted dynamic contrast-enhanced T1WI. 2.4. Statistical analyses The Mann-Whitney U Test was used to compare size differences between seminomas and NSGCNs. Intraclass correlation coefficient (ICC, 0.00–0.20 poor, 0.21–0.40 fair, 0.41–0.60 moderate, 0.61–0.80 good and 0.81–1.00 excellent correlation) was used to assess interobserver reliability for tumor ADC measurements from the two readers for each individual ROI method. A general linear model was used to compare seminoma versus NSGCN ADC measurements. In this statistical approach, measurements of the two observers and the four methods were treated as “repeated measures”. Equivalently observer and method were considered within-subjects factors and pathology type (either seminoma or NSGCN) was considered a within-subject factor, which defines the two conditions within each subject. Repeated measures ANOVA and the Bonferonni correction for post hoc testing were applied to compare ADC measurements among observers, methods and histologic types. For each observer, a stepwise backward binary logistic analysis was performed to assess independent predictors of pathology type among the four different methods of measurements. ROC analysis, using measurements from both observers, was performed and the area under the curve (AUC) was calculated to assess how
A.C. Tsili et al. / European Journal of Radiology 89 (2017) 1–6
3
Table 1 MRI protocol (Abbreviations: TR: time repetition; TE: time echo; FOV: field of view; T1WI: T1-weighted imaging; T2WI: T2-weighted imaging; DWI: diffusion-weighted imaging; 3D FFE: three-dimensional fast field-echo; c.m. i.v., contrast material intravenously). Sequences
spin-echo T1WI
fast-spin echo T2WI
DWI
3D FFE post-contrast
Plane TR (msec) TE (msec) Slice thickness (mm) Gap (mm) FOV (mm) Matrix (mm) b value (s m−2 ) Flip angle Dynamic scans c.m i.v. (ml kg−1 )
transverse 500–650 13–15 3–4 0.5 240 × 270 180 × 256
transverse, sagittal, coronal 4000 100–120 3–4 0.5 240 × 270 180 × 256
transverse 3900 115 3–4 0.5 240 × 270 180 × 256 0, 900
coronal 9 4.1 4 0 219 × 219 256 × 256
35◦ 7 (every 60 sec) 0.2
Fig. 1. ADC measurements were performed on the ADC map using four distinct ROI methods: (a) round ROI; (b) square ROI; (c) freehand ROI; (d) and five, small round ROIs. Care was taken to avoid areas of necrosis with the aid of the corresponding (e) T2WI and (f) subtracted dynamic contrast-enhanced T1WI.
accurately each method can discriminate between seminoma and NSGCN cases. The maximum Youden Index (J) was used to derive the optimal cut-off ADC value which maximizes the sum of sensitivity and specificity. All statistical analysis was performed using SPSS v.20.0 (IBM Corp., Armonk, NY, USA). Two-tailed values of P < 0.05 were regarded as statistically significant. 3. Results The median size of seminomas and nonseminomas was 4.1 cm (range: 1.2–9.9 cm) and 4.7 cm (range: 1.5–11.5 cm), respectively. There was no size difference between the two groups (P = 0.73). The intraclass correlation coefficient between the two observers was high (ICC = 0.963, P < 0.001) assuring the reliability of the measurements. ICCs were 0.968 for round ROIs, 0.964 for square ROIs, 0.966 for freehand ROIs, and 0.951 for the multiple small, round ROIs. All ROI methods showed excellent correlation (P < 0.001). Similarly, ANOVA showed no difference (P = 0.658) between mean ADC of the first (mean ADC ± std error = 0.742 ± 0.027 × 10−3 mm2 /s) and the second observer (mean ADC ± std err = 0.747 ± 0.024 × 10−3 mm2 /s). Mean ADC ± std (×10−3 mm2 /s) was 0.749 ± 0.025 for round ROI, 0.746 ± 0.025 for square ROI, 0.765 ± 0.025 for freehand ROI and 0.718 ± 0.024 for multiple, small round ROIs (Table 2). Post hoc
comparisons between mean values of the measurement methods showed that multiple, small round ROIs provides lower ADC (Table 2) compared to the others. Also, the mean ADC ± std (×10−3 mm2 /s) of seminoma (0.608 ± 0.031) was lower (P < 0.001) compared to the mean ADC of NSGCN (0.88 ± 0.038). Similarly, separately for each ROI method, the mean ADC of seminoma was lower (P < 0.001 in all cases) compared to the mean ADC of NSGCN (Table 3, Fig. 2). Logistic regression analysis, both for the first and the second observer, showed that among the various ROI methods, only the round ROI is independent predictor of the pathology type (P = 0.008 in both cases). Similarly, ROC analysis showed that round ROI has the best discriminative accuracy with AUC = 0.910 (square ROI, AUC = 0.897; freehand ROI, AUC = 0.888; and, multiple small ROIs, AUC = 0.872) (Fig. 3). The optimal cut-off ADC using the round ROI was found 0.73 × 10−3 mm2 /s (J = 0.746). 4. Discussion Our results demonstrate that all ROI methods provide excellent interobserver agreement with excellent correlation when measuring ADC in testicular germ cell neoplasms. Using four distinct ROI methods, ADC of TGCNs was within the range reported in previous publications [3–6]. However, mean ADC of TGCNs was lower for multiple, small round ROIs than that measured with round, square
4
A.C. Tsili et al. / European Journal of Radiology 89 (2017) 1–6
Table 2 Mean ADC measurements for each method and their post hoc comparison. ROI methods
ADC (mean ± std. error 10−3 mm2 /s)
P(comparison to circular)
P(comparison to square)
P(comparison to freehand)
round square freehand multiple, small round ROIs
0.749 ± 0.025 0.746 ± 0.025 0.765 ± 0.025 0.718 ± 0.024
1.000 0.377 0,006
0.149 0,013
<0.001
Table 3 Mean ADC measurements for each method and histologic type. ROI methods
Histologic type
ADC (mean ± std. error10−6 mm2 /s)
P
round
seminoma NSGCN seminoma NSGCN seminoma NSGCN seminoma NSGCN
0.603 ± 0.010 0.894 ± 0.194 0.606 ± 0.099 0.885 ± 0.198 0.634 ± 0.091 0.896 ± 0.201 0.59 ± 0.082 0.846 ± 0.201
<0.001
square freehand multiple, small round ROIs
Fig. 2. Boxplots of ADC measured by both observers, for each histologic type and for each of the four ROI methods. Dark horizontal lines represent the mean values; the box represents the 25th and 75th percentiles and the whiskers the 5th and 95th percentiles.
Fig. 3. ROC curves for each ROI method.
<0.001 <0.001 <0.001
and freehand ROI. The multiple, small round ROIs include only the most viable, solid part of the neoplasm, seen as areas of highest restriction on the ADC maps, resulting in lower ADC values. In these measurements, areas of hemorrhage and/or necrosis along with the edges of the tumor are more likely to be excluded, than with the other three methods. In cases of TGCNs that require immediate chemotherapy instead of radical orchiectomy, the characterization of the histologic type of testicular tumor is of great importance [19]. Conventional MRI findings have been found closely to correlate with the histopathologic characteristics of TGCNs [17,18]. Recently, measurements of ADC provided an additional tool in differentiating seminomas from NSGCNs [7]. In a previous report of 26 histologically proven TGCNs, we found ADC useful in the preoperative discrimination of seminomas from nonseminomatous tumors. Specifically, the mean ADC ± std (×10−3 mm2 /s) of seminomas (0.59 ± 0.009) was significantly lower compared to that of NSGCTs (0.90 ± 0.33) in this report [7]. This is in accordance with present results. Histopathologic characteristics of seminomatous tumors, including presence of large, uniform cells with large nuclei, prominent nucleoli and abundant cytoplasm, arranged in nests, outlined by lymphocytebearing fibrovascular septa probably account for the significant restriction in the extracellular movement of water protons in cases of seminomas when compared to NSGCNs [15,16]. Mean ADCs from all ROI methods showed significant differences between seminomas and NSGCNs, with round ROI providing the most accurate technique for the characterization of the histologic subtype of TGCNs, with an optimal cut-off value of 0.73 × 10−3 mm2 /s. This is in discrepancy with the first part of our study, where multiple, small round ROIs proved the most useful technique to characterize TGCNs. The frequent presence of areas of necrosis and/or hemorrhage in nonseminomas and the coexistence of fibrovascular septa in seminomas might probably create difficulties in the delineation of small, round ROIs when measuring ADC values, and the therefore in the histologic characterization of TGCNs. Various publications on DWI have established its role in the detection and characterization of malignancies [1,2]. In most studies, ADC measurements are performed by one observer and a variety of methods for ROI shape, including round, square, polygonal and small round ROIs. Methods for ROI placement also vary, including whole tumor volume, a single-slice or small tumor samples [20,21]. There are prior reports that have assessed the interobserver reliability and the effect of different ROI methods in ADC measurements in various malignancies [22–28]. In a retrospective study of 69 patients with endometrial carcinoma, Inoue
A.C. Tsili et al. / European Journal of Radiology 89 (2017) 1–6
et al. investigated the influence of four manual ROI methods, similar to ours, that is round, square, freehand and multiple, small ROIs on both the minimum and mean ADC tumor measurements and also assessed interobserver variability [22]. The authors found that all four ROI methods had excellent interobserver agreement, with excellent correlation, as proved also in our study. Each minimum ADC was significantly different, except between square and round ROI. Mean ADCs showed significant differences only between freehand ROI and the other methods. This study concluded that mean ADC provides more stable results, with round or square ROI representing a more suitable method for ADC measurements [22]. In a recent study of 54 women with ovarian tumors with a solid component, the effect of similar ROI methods on ADC values and their diagnostic ability in differentiating benign from malignant tumors were assessed [23]. The interobserver reliability in ADC measurement was good or excellent, with the exception of the square ROI. Freehand ROI showed the lowest and round ROI showed the highest minimum ADC. Freehand ROI showed the highest mean ADC, and round and square ROIs showed the lowest mean ADC. The authors concluded that ROI shape influences the ADC measurements and the optimal cutoff ADC used to differentiate benign from malignant ovarian tumors [23]. In another study, the influence of three ROI protocols, including ‘whole-volume’, ‘single-slice’, using a single freehand ROI and ‘small solid samples’, using three round/oval-shaped ROIs placed within the most solid tumoral part of three independent tumourcontaining slices was assessed, in 46 patients with locally advanced rectal carcinoma [25]. The authors concluded that variations in ROI size and positioning had a significant effect on tumour ADC and interobserver variability. ADC of the whole tumour volume provided the most reproducible results in the same report [25]. Ahlawat et al. assessed the interobserver reliability of three different manual ROI protocols for determining the mean and minimum ADC in 73 musculoskeletal soft tissue masses compared with whole tumor volume ADC [27]. The authors showed that a fast selective observer-based method provided similar results to whole tumor volume analysis with high interobserver agreement and a significant decrease in measurement time [27]. Another report by Nogueira et al. compared two different methods of ROIs on ADC measurement in 39 breast lesions, using a small area of highest restriction and a large ROI, including the whole lesion. Although, interobserver variability was low for both methods, small ROIs showed less overlap in ADC and higher reproducibility, suggesting that this method might improve lesion characterization [28]. A considerable influence of the ROI method on mean ADC and the interobserver variability in pancreatic adenocarcinoma has been reported by Liu et al. [26]. The authors in a prospective study of 21 patients with surgically-proven pancreatic adenocarcinoma compared the influence of three techniques of ROI placement for ADC measurements, including whole-volume, single-slice, and small solid samples of tumor [26]. The study concluded that ADC based on the small solid samples of tumor provided the highest diagnostic performance in assessing pancreatic adenocarcinoma, and this was in accordance with our results [26]. DWI is usually performed before the intravenous administration of a gadolinium chelate contrast agent, as it was in the present study. Up to now, there are limited data regarding the effectiveness of DWI after gadolinium administration, with studies reporting no significant change of ADC values after contrast material administration [29,30]. This is important in cases when DWI fails or data interpretation is difficult due to the presence of artifacts, allowing DW sequences to be repeated after contrast administration [29,30]. This study has some limitations to be acknowledged. First, it is a retrospective review of relatively small patient numbers. No subgroup analysis of the different histologic subtypes of NSGCNS was performed due to the small number of cases. Further studies
5
are required to validate the present results in a larger study cohort. Nonseminomas comprise a heterogeneous group of neoplasms and histologic diversity of these tumors could result in ADC differences. Second, we did not use the whole tumor volume ROI in our report. Although, the whole-volume ROI is considered a reliable method, this technique is time-consuming and not easily applicable in routine clinical practice. Another limitation is that the measurement of ADC by the four ROI methods was performed at the same time, and therefore, there might be some bias. Furthermore, each reader independently had drawn the largest cross-section of the tumor. Therefore, ROIs might be placed in different areas by each reader. Finally, intraobserver agreement was not assessed. 5. Conclusion In concluding, ROI shape has no marked effect on interobserver variability in ADC measurements in testicular germ cell neoplasms. Multiple, small round ROIs and round ROI have been proved more useful methods for ADC measurement in characterizing TGCNs and in differentiating seminomas from nonseminomatous tumors, respectively. Although histology remains the gold standard for the characterization of TGCNs, larger prospective studies could probably justify the potential role of MRI, including DWI as a non-invasive tool, helpful in the histologic differentiation of these neoplasms. Conflict of interest I declare no financial or commercial interests related to this paper. References [1] T.C. Kwee, T. Takahara, R. Ochiai, et al., Whole-body diffusion-weighted magnetic resonance imaging, Eur. J. Radiol. 70 (2009) 409–417. [2] R. Bammer, Basic principles of diffusion-weighted imaging, Eur. J. Radiol. 45 (2003) 169–184. [3] A.C. Tsili, M.I. Argyropoulou, D. Giannakis, S. Tsampalas, N. Sofikitis, K. Tsampoulas, Diffusion-weighted MR imaging of normal and abnormal scrotum: preliminary results, Asian J. Androl. 14 (2012) 649–654. [4] A.M. Algebally, H.I. Tantawy, R.R. Yousef, W. Szmigielski, A. Darweesh, Advantage of adding diffusion weighted imaging to routine MRI examinations in the diagnostics of scrotal lesions, Pol. J. Radiol. 80 (2015) 442–449. [5] A.C. Tsili, D. Giannakis, A. Sylakos, et al., Apparent diffusion coefficient values of normal testis and variations with age, Asian J. Androl. 16 (2014) 493–497. [6] A.C. Tsili, A. Ntorkou, D. Baltogiannis, et al., The role of apparent diffusion coefficient values in detecting testicular intraepithelial neoplasia: preliminary results, Eur. J. Radiol. 84 (2015) 828–833. [7] A.C. Tsili, A. Sylakos, A. Ntorkou, et al., Apparent diffusion coefficient values and dynamic contrast enhancement patterns in differentiating seminomas from nonseminomatous testicular neoplasms, Eur. J. Radiol. 84 (2015) 1219–1226. [8] D. Maki, Y. Watanabe, M. Nagayama, et al., Diffusion-weighted magnetic resonance imaging in the detection of testicular torsion: feasibility study, J. Magn. Reson. Imaging 34 (2011) 1137–1142. [9] Y. Watanabe, M. Nagayama, A. Okumura, et al., MR imaging of testicular torsion: features of testicular hemorrhagic necrosis and clinical outcomes, J. Magn. Reson. Imaging 26 (2007) 100–108. [10] M. Kantarci, S. Doganay, A. Yalcin, Y. Aksoy, B. Yilmaz-Cankaya, B. Salman, Diagnostic performance of diffusion-weighted MRI in the detection of nonpalpable undescended testes: comparison with conventional MRI and surgical findings, AJR Am. J. Roentgenol. 195 (2010) W268–73. [11] S. Emad-Eldin, N. Abo-Elnagaa, S.A. Hanna, A.H. Abdel-Satar, The diagnostic utility of combined diffusion-weighted imaging and conventional magnetic resonance imaging for detection and localization of non palpable undescended testes, J. Med. Imaging Radiat. Oncol. 60 (2016) 344–351. [12] E. Karakas, O. Karakas, N. Cullu, et al., Diffusion-weighted MRI of the testes in patients with varicocele: a preliminary study, AJR Am. J. Roentgenol. 202 (2014) 324–328. [13] P. Albers, W. Albrecht, F. Algaba, et al., Guidelines on Testicular Cancer, European Association of Urology, 2015, Available at: http://uroweb.org/wpcontent/uploads/11-Testicular-Cancer LR1.pdf. [14] American Cancer Society. Cancer facts and figures 2016. Available at: http:// www.cancer.org/acs/groups/content/@research/documents/document/acspc047079.pdf. [15] T.M. Ulbright, D.M. Berney, et al., Testicular and paratesticular tumors, in: S.E. Mills, D. Carter, J.K. Greenson (Eds.), Sternberg’s Diagnostic Surgical
6
[16]
[17]
[18]
[19]
[20]
[21]
[22]
A.C. Tsili et al. / European Journal of Radiology 89 (2017) 1–6 Pathology, 5th edition, Lippincott Williams & Wilkins, Philadelphia, 2010, pp. 1944–2004. WHO histological classification of testis tumours, in: J.N. Eble, G. Sauter, J.I. Epstein, I.A. Sesterhenn (Eds.), Pathology & Genetics. Tumours of the Urinary System and Male Genital Organs, IARC Press, Lyon, 2004, pp. 212–274. J.O. Johnson, R.F. Mattrey, J. Phillipson, Differentiation of seminomatous from nonseminomatous testicular tumors with MR imaging, AJR Am. J. Roentgenol. 154 (1990) 539–543. A.C. Tsili, C. Tsampoulas, X. Giannakopoulos, et al., MRI in the histologic characterization of testicular neoplasms, AJR Am. J. Roentgenol. 189 (2007) W331–7. J. Oldenburg, S.D. Fosså, J. Nuver, et al., Testicular seminoma and non-seminoma: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up, Ann. Oncol. 24 (Suppl. 6) (2013) vi125–vi132. E. Giannotti, S. Waugh, L. Priba, Z. Davis, E. Crowe, S. Vinnicombe, Assessment and quantification of sources of variability in breast apparent diffusion coefficient (ADC) measurements at diffusion weighted imaging, Eur. J. Radiol. 84 (2015) 1729–1736. P. Clauser, M. Marcon, M. Maieron, C. Zuiani, M. Bazzocchi, P.A. Baltzer, Is there a systematic bias of apparent diffusion coefficient (ADC) measurements of the breast if measured on different workstations? An inter- and intra-reader agreement study, Eur. Radiol. 26 (2016) 2291–2296. C. Inoue, S. Fujii, S. Kaneda, et al., Apparent diffusion coefficient (ADC) measurement in endometrial carcinoma: effect of region of interest methods on ADC values, J. Magn. Reson. Imaging 40 (2014) 157–161.
[23] N. Mukuda, S. Fujii, C. Inoue, et al., Apparent diffusion coefficient (ADC) measurement in ovarian tumor: effect of region-of-interest methods on ADC values and diagnostic ability, J. Magn. Reson. Imaging 43 (2016) 720–725. [24] I.M. Blazic, G.B. Lilic, M.M. Gajic, Quantitative assessment of rectal cancer response to neoadjuvant combined chemotherapy and radiation therapy: comparison of three methods of positioning region of interest for ADC measurements at diffusion-weighted MR imaging, Radiology 2 (2016) 151908 [Epub ahead of print]. [25] D.M.J. Lambregts, G.L. Beets, M. Maas, et al., Tumour ADC measurements in rectal cancer: effect of ROI methods on ADC values and interobserver variability, Eur. Radiol. 21 (2011) 2567–2574. [26] L. Liu, C. Ma, J. Li, et al., Comparison of the diagnostic performances of three techniques of ROI placement for ADC measurements in pancreatic adenocarcinoma, Acad. Radiol. 22 (2015) 1385–1392. [27] S. Ahlawat, P. Khandheria, F. Del Grande, et al., Interobserver variability of selective region-of-interest measurement protocols for quantitative diffusion weighted imaging in soft tissue masses: comparison with whole tumor volume measurements, J. Magn. Reson. Imaging 43 (2016) 446–454. [28] L. Nogueira, S. Brandão, E. Matos, et al., Region of interest demarcation for quantification of the apparent diffusion coefficient in breast lesions and its interobserver variability, Diagn. Interv. Radiol. 21 (2015) 123–127. [29] C.L. Wang, Y.W. Chea, D.T. Boll, et al., Effect of gadolinium chelate contrast agents on diffusion weighted MR imaging of the liver, spleen, pancreas and kidney at 3 T, Eur. J. Radiol. 80 (2011) e1–7. [30] S. Kinner, L. Umutlu, S. Blex, et al., Diffusion weighted MR imaging in patients with HCC and liver cirrhosis after administration of different gadolinium contrast agents: is it still reliable, Eur. J. Radiol. 81 (2012) e625–8.