PRECLINICAL AND CLINICAL IMAGING - Full Papers

Magnetic Resonance in Medicine 73:2262–2273 (2015)

Esophageal Carcinoma: Evaluation with q-Space Diffusion-Weighted MR Imaging Ex Vivo Ichiro Yamada,1* Keigo Hikishima,2,3 Naoyuki Miyasaka,4 Yutaka Tokairin,5 Eisaku Ito,6 Tatsuyuki Kawano,5 Daisuke Kobayashi,6 Yoshinobu Eishi,6 and Hideyuki Okano2 INTRODUCTION

Purpose: To determine the usefulness of q-space MR imaging as means of evaluating the depth of mural invasion, the histologic grades, and lymph node metastasis in esophageal carcinomas. Methods: Twenty esophageal specimens each containing a carcinoma were studied with a 7.0 Tesla MR imaging system. q-Space MR images were obtained with a 50–60 mm  25– 30 mm field of view, 256  128 matrix, 2 mm section thickness, 10 b values ranging from 0 to 7163 s/mm2, and a motion-probing gradient in the y-direction, and the MR images were compared with the histopathologic findings. Results: The mean displacement maps, probability for zero displacement maps, and kurtosis maps in all 20 carcinomas (100%) made it possible to identify the depth of tumor invasion of the esophageal wall. These q-space MR imaging parameters were significantly correlated with the histologic grades of the esophageal carcinomas (P < 0.01), and also significantly correlated with their nuclear-cytoplasmic ratios (P < 0.01 or P < 0.001) and tumor cellularity (cell density) (P < 0.01 or P < 0.001). The q-space MR imaging parameters were also capable of differentiating between the metastatic lymph nodes and nonmetastatic lymph nodes (P < 0.01). Conclusion: q-Space MR imaging ex vivo provides excellent diagnostic accuracy for evaluating mural invasion by esophageal carcinomas, the histologic grades of esophageal carcinomas, and lymph node metastasis by esophageal carcinomas. C 2014 Wiley PeriodMagn Reson Med 73:2262–2273, 2015. V icals, Inc. Key words: esophagus; esophageal carcinoma; q-space imaging; diffusion-weighted imaging; MR imaging

1 Department of Diagnostic Radiology and Oncology, Graduate School, Tokyo Medical and Dental University, Tokyo, Japan. 2 Department of Physiology, Keio University School of Medicine, Tokyo, Japan. 3 Central Institute for Experimental Animals, Kanagawa, Japan. 4 Department of Pediatrics, Perinatal and Maternal Medicine, Tokyo Medical and Dental University, Tokyo, Japan. 5 Department of Esophagogastric Surgery, Tokyo Medical and Dental University, Tokyo, Japan. 6 Department of Pathology, Tokyo Medical and Dental University, Tokyo, Japan. Grant sponsor: Grant-in-Aid for Scientific Research (C) of MEXT, Japan; Grant number: 23591753. *Correspondence to: Ichiro Yamada, M.D., Department of Diagnostic Radiology and Oncology, Graduate School, Tokyo Medical and Dental University, 1–5-45 Yushima, Bunkyo-ku, Tokyo 113–8519, Japan. E-mail: [email protected] Additional Supporting Information may be found in the online version of this article. Received 3 March 2014; revised 27 May 2014; accepted 3 June 2014 DOI 10.1002/mrm.25334 Published online 19 June 2014 in Wiley Online Library (wileyonlinelibrary. com). C 2014 Wiley Periodicals, Inc. V

Esophageal carcinoma is a common malignant neoplasm worldwide, and its incidence has been steadily increasing (1). The prognosis of esophageal carcinoma patients is strictly dependent on the depth of tumor invasion and whether lymph node metastasis is present and its extent (2,3), and accurate preoperative assessment of these prognostic factors has a definitive impact on the selection of appropriate treatment (4). Local staging of esophageal carcinoma is currently performed on the basis of computed tomography (CT) and endoscopic ultrasound (EUS) findings, but depth of tumor invasion and lymph node metastasis cannot be reliably assessed by these methods. CT does not allow evaluation of less extensive tumors, because the poor soft-tissue contrast makes it impossible to resolve the layers of the esophageal wall (5–7). EUS also entails many inherent problems, including technical failures in stenotic tumors, high operator dependency, artifactual interface echoes in the esophageal wall, and a limited sonographic range (7–9). Thus, the diagnostic methods that are available to evaluate tumor invasion and lymph node metastasis by esophageal carcinomas are very limited. MR imaging has been reported to be capable of visualizing mural invasion by esophageal carcinoma and offers an alternative to CT and EUS (10–12), but conventional MR imaging is also incapable of resolving the individual layers of the esophageal wall (13–18). q-Space diffusion-weighted MR imaging has recently been introduced to depict microstructures in the brain and spinal cord (19–21), and diffusional kurtosis has been reported to be useful in probing the histologic grade of tumors (22,23), which is also known to be one of the most important prognostic factors in esophageal carcinoma patients as well (24,25). The purpose of this study was to use q-space MR imaging to prospectively examine surgical specimens of the esophagus that contained an esophageal carcinoma and assess the usefulness of q-space MR imaging as means of evaluating the depth of mural invasion by esophageal carcinomas, evaluating the histologic grade of esophageal carcinomas, and determining whether lymph node metastasis was present. METHODS Study Population This study was approved by our Institutional Review Board, and written informed consent in regard to the purpose of this study and the use of clinical and histopathologic data was obtained from each patient. We

2262

q-Space Imaging of Esophageal Carcinoma

studied 20 surgical specimens of the esophagus, each of which contained a tumor that had been histopathologically diagnosed as squamous cell carcinoma. The specimens were obtained from 20 consecutive esophageal carcinoma patients who were surgically treated at our institution. Eighteen of the 20 patients were male, and 2 were female. Their ages at the time of surgery ranged from 46 years to 83 years [mean age: 66 years 6 8 (standard deviation)]. Thirteen patients underwent no treatment before surgical resection, 5 patients underwent only chemotherapy, and 2 patients underwent both chemotherapy and radiation therapy. The specimens in these patients were analyzed and reported using the completely different techniques (diffusion-tensor MR imaging and tractography) in our previous works (26,27). All 20 specimens were imaged after fixation in 10% formalin. We did not examine the esophagus in vivo in this series. Imaging Technique q-Space MR imaging was performed by using a 7.0 Tesla (T) MR imaging unit (BioSpec 70/16; Bruker BioSpin, Ettlingen, Germany) equipped with actively shielded gradients that had a maximum strength of 700 mT/m. A four-channel phased-array surface coil was used to make all measurements. The orientation of MR images was set longitudinally along the long axis of the resected segment of the esophagus. q-Space MR imaging data sets were acquired by using a diffusion-weighted spin-echo pulse sequence. The imaging parameters were: repetition time, 3000 ms; echo time, 29 ms; field of view, 50–60  25–30 mm; matrix, 256  128; section thickness, 2 mm without intersection gaps; voxel size, 0.195–0.234  0.195–0.234  2 mm (0.076–0.110 mm3); and number of excitations, one. The diffusion-weighting gradients were applied in the y-direction of the MR images that was perpendicular to the long axis of the esophagus, and they had a duration time (d) of 4.5 ms, a separation time (D) of 18.5 ms, effective diffusion time (Deff ¼ D - d/3) of 17.0 ms, and 10 different gradient strengths (g): 0, 59, 118, 177, 236, 295, 354, 414, 472, and 507 mT/m. The resulting 10 b values were 0, 97, 387, 874, 1554, 2423, 3495, 4762, 6206, and 7163 s/ mm2, which corresponded to 10 q values of 0, 119, 238, 358, 478, 597, 717, 837, 955, and 1026 /cm, respectively. The acquisition time was 43 min. The q-space imaging was the only pulse sequence used in the present study, because our purpose was to compare the findings of qspace imaging with the histopathologic findings. Image Processing Based on the q-space imaging methodology (19), the raw q-space imaging data were analyzed using an in-house IDL program (28). Our exact algorithms were the same as the technique described by Cohen and Assaf (29). Briefly, we used the q-space approach to relate the signal intensity attenuation in diffusion measurements to the displacement probabilities through the reciprocal spatial vector q (defined as gdg/2p) according to the following equation:

2263

ED ðqÞ ¼

Z

PS ðR; DÞexpði2pqRÞdR

[1]

where ED ðqÞ is the measured signal intensity decay as a function of q for a specific diffusion time (D), PS ðR; DÞ is the averaged diffusion propagator (displacement probability), and R is the net displacement vector (R ¼ r – r0; r0 is the initial position of spins and r is the final position of spins after the time of D). Before a Fourier transform we performed the linear interpolation between elements, although no data extrapolation was performed. From the Eq. [1], by performing the Fourier transform of the signal intensity decay with respect to q, we obtained the displacement distribution profiles according to the following equation: Z PS ðR; DÞ ¼ ED ðqÞexpði2pqRÞdq [2] as illustrated in Figure 1. More details of the algorithms would be referred to the previous reports (19,29). After the Fourier transform of the signal intensity decay with respect to q, the following three q-space imaging parameters were calculated from the displacement distribution profiles for each voxel: mean displacement (in mm), probability for zero displacement [in arbitrary units (a.u.)], and kurtosis (in a.u.). Moments were calculated from the propagator and used to estimate the q-space imaging parameters. The mean displacement was measured as 0.425 times the full width at half maximum (FWHM) of the displacement distribution profiles. The probability for zero displacement was measured as the height of the profiles at zero displacement. The kurtosis was measured as the deviation from the Gaussian distribution according to the following equation: Kurtosis ¼

4 1  X Pj  P 1N pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  3 N j¼0 Variance

[3]

Due to the linear interpolation between elements, we used more elements for calculating the kurtosis than the actual number of sampling, so the kurtosis values were larger in the present study. The interpolation in q-space led to an extension of the tails of the displacement distribution profiles obtained by the Fourier transform. It is likely that the extension of the tails of the displacement distribution profiles led to further deviation from the Gaussian distribution and thus resulted in larger kurtosis values, because the kurtosis represents the deviation from the Gaussian distribution (30). Indeed, when they were calculated without the interpolation, the kurtosis values were considerably smaller, ranging from 0.256 (the submucosa layer) to 2.592 (the outer longitudinal muscle layer) (Refer to Fig. 1 and Table 1). Finally, we generated mean displacement maps, probability for zero displacement maps, and kurtosis maps based on these three parameters on a voxel-by-voxel basis. Using the maximal q-value (qmax), we detected the diffusion displacement resolution of approximately 9 mm in this study (Fig. 1). This diffusion displacement length was approximately equal to the reciprocal of qmax, i.e., the minimum encoding length (1/(2pqmax)) multiplied by 2p.

2264

Yamada et al.

FIG. 1. Graphs obtained from the layers of the normal esophageal wall. a: Normalized signal decay curves as a function of q value for the outer longitudinal muscle (OLM) layer and the submucosa (SM) layer of the normal esophageal wall. The outer longitudinal muscle layer shows higher signal intensities than the submucosa layer at all q values, suggestive of a higher degree of restricted diffusion in the outer longitudinal muscle layer than in the submucosa layer. b: Displacement distribution profiles obtained from the Fourier transform of the data shown in Figure 1a with respect to q value. The outer longitudinal muscle layer shows a higher and narrower peak than the submucosa layer, clearly demonstrating a higher degree of restricted diffusion in the outer longitudinal muscle layer than in the submucosa layer. a.u. ¼ arbitrary unit.

We did not correct for noise effects on the moment calculations of the propagators in this study. This is because the b0 images showed so high values of the signal-to-noise ratio (SNR) (ranging from 47.76 to 100.78) that the noise levels seemed sufficiently low compared with the signal levels. The previous report (29) has demonstrated that the q-space parameters hardly change when the SNR is more than 20, so we considered that the noise floor in our signal decay curves was sufficiently negligible and did not correct for the noise.

lymph nodes were compared with the histopathologic findings on a node-by-node basis. For the mean displacement maps, the probability for zero displacement maps, and the kurtosis maps, regions of interest were placed on the carcinoma, on each layer of the esophageal wall, and on the lymph nodes. The regions of interest were defined so that the size was approximately equal to the cross-section area of the primary tumor and lymph nodes. All quantitative measurements of the q-space MR images were made by analyzing the regions of interest with the ImageJ 1.47 software program (available at http://imagej.nih.gov/ij). The mean values of three or four regions of interest in the

Image Analysis An independent, blinded evaluation of q-space MR images in each surgical specimen was performed by two observers (I.Y., N.M.) who had no knowledge of the results of the histopathologic examination. When the observers could not fully agree on the findings, a consensus was achieved by discussion. The mean displacement maps, the probability for zero displacement maps, and the kurtosis maps were reviewed for the presence, signal intensity (SI), uniformity, and thickness of each layer of the esophageal wall. The contour and SI of the carcinoma were then analyzed, and the degree of carcinoma penetration into the esophageal wall was classified according to the deepest layer invaded: mucosa, submucosa, muscularis propria, or adventitia. The MR imaging criteria used by the observers to determine the depth of carcinoma invasion were: (a) discrete mass(es) in the layers, (b) abnormal SI within the thickened layers, (c) focal abnormal SI within the layers, and (d) mucosal ulceration with surrounding or underlying mass(es) (10). The MR images of the periesophageal lymph nodes adjacent to the primary tumor in the specimens were also analyzed for the nodal size, SI, and border contour, and the MR imaging findings in the

Table 1 Mean Displacement Values, Probability for Zero Displacement Values, and Kurtosis Values of the Esophageal Carcinomas and the Layers of the Esophageal Wall at q-Space Diffusion-Weighted MR Imaging

Tissue

Mean displacement values (mm)

Probability for zero displacement values (a.u.)

Kurtosis values (a.u.)

Carcinoma Epithelium LPM MM Submucosa ICM IMCT OLM Adventitia

6.01 6 0.50 5.97 6 0.42 8.24 6 1.04a 6.27 6 0.30 11.38 6 2.04a 5.83 6 0.38 8.96 6 1.14a 5.63 6 0.42 10.93 6 0.75a

49.8 6 3.8 48.0 6 5.0 34.7 6 4.5a 42.3 6 4.4 26.0 6 2.3a 51.3 6 7.3 33.4 6 4.8a 53.8 6 7.2 30.0 6 3.2a

52.3 6 5.7 51.2 6 1.2 33.0 6 5.3a 48.3 6 0.9 22.7 6 1.7a 57.2 6 10.1 34.2 6 7.3a 58.3 6 7.1 31.7 6 5.6a

a ¼ Significantly different from the corresponding value of the esophageal carcinomas. LPM ¼ lamina propria mucosae, MM ¼ muscularis mucosae, ICM ¼ inner circular muscle, IMCT ¼ intermuscular connective tissue, OLM ¼ outer longitudinal muscle. a.u. ¼ arbitrary unit.

q-Space Imaging of Esophageal Carcinoma

carcinoma, in each layer of the esophageal wall, and in the lymph nodes were then calculated. The q-space MR imaging findings in the 20 esophageal specimens were compared with the histopathologic findings, which served as the reference standard. The MR images were compared with specific histopathologic sections on a slice-by-slice level, and correlations were made by visual inspection. To match the MR images and the histologic sections, spatial correlations were made by identifying anatomic landmarks (eg, esophageal contour, blood vessels) that were depicted. Histologic Preparations and Examination After MR imaging, each surgical specimen was sectioned longitudinally so that the orientation of the sections corresponded to the orientation of the MR images. The sectioned specimens were embedded in paraffin and cut into 6-mm-thick slices with a microtome. The slices were then stained with hematoxylin and eosin stain, elasticavan Gieson (EVG) stain, and periodic acid-Schiff (PAS) stain. An experienced pathologist (E.I.) with no knowledge of the MR imaging findings identified carcinoma invasion into each layer of the esophageal wall, and the experienced pathologist classified the histologic grades of the esophageal carcinomas into well differentiated, moderately differentiated, or poorly differentiated based on the American Joint Committee on Cancer (AJCC) criteria (2,3). The nuclear-cytoplasmic ratio (N/C ratio) was calculated by using the following formula: N/C ratio ¼ nuclear area / (cellular area – nuclear area). The nuclearcytoplasmic ratio was determined in five 400 fields in each specimen, and the mean value of a total of 100 cells from each carcinoma was calculated. Tumor cellularity [cell density; TC (%)] was calculated by using the following formula: TC ¼ area of tumor cells / area of the histologic section  100. Tumor cellularity was determined in five 200 fields in each specimen, and the mean value of the five fields from each carcinoma was calculated. The ImageJ 1.47 software was used to calculate the nuclear area, cellular area, area of tumor cells, and area of the histologic section. For the tumor cellularity, cells were not counted and a constant cell area was not assumed. Statistical Analysis Means 6 standard deviations of the mean displacement values, probability for zero displacement values, and kurtosis values were calculated in the carcinoma, in each layer of the esophageal wall, and in the lymph nodes. All statistical analyses were performed with a commercial software package (IBM SPSS Statistics, version 20; IBM SPSS Japan, Tokyo, Japan). The differences in q-space imaging parameters between the carcinoma and the layers of the esophageal wall were analyzed by the Dunnett test. Correlations between the histologic grades and the q-space imaging parameters were assessed by the Spearman correlation coefficient by rank, and correlations between the nuclear-cytoplasmic ratios or tumor cellularity and the q-space imaging parameters

2265

were assessed by means of Pearson correlation coefficients. The q-space imaging parameters of the metastatic lymph nodes and nonmetastatic lymph nodes were compared by using the Mann-Whitney test. A P value less than 0.05 was considered evidence of a statistically significant difference. RESULTS Findings on Mean Displacement Maps, Probability for Zero Displacement Maps, and Kurtosis Maps of the Esophageal Carcinomas and the Layers of the Esophageal Wall The mean displacement maps, probability for zero displacement maps, and kurtosis maps of all 20 specimens (100%) clearly depicted the esophageal wall as consisting of the following eight layers: epithelium, lamina propria mucosae, muscularis mucosae, submucosa, inner circular muscle, intermuscular connective tissue, outer longitudinal muscle, and adventitia (Fig. 2; Supporting Table S1). These eight layers clearly corresponded to the layers of the esophageal wall observed in histologic sections. As shown in Supporting Table S1, the esophageal carcinomas were seen as low SI on the mean displacement maps, as high SI on the probability for the zero displacement maps, and as high SI on the kurtosis maps. The carcinomas on the mean displacement maps appeared as areas of lower SI than the lamina propria mucosae, submucosa, intermuscular connective tissue, and adventitia. On the probability for zero displacement maps, the carcinomas had higher SI than the lamina propria mucosae, submucosa, intermuscular connective tissue, and adventitia. The carcinomas also had higher SI on the kurtosis maps than the lamina propria mucosae, submucosa, intermuscular connective tissue, and adventitia. Mean Displacement Values, Probability for Zero Displacement Values, and Kurtosis Values of the Esophageal Carcinomas and the Layers of the Esophageal Wall As shown in Table 1, the mean displacement values of the esophageal carcinomas were statistically significantly lower than those of the lamina propria mucosae, submucosa, intermuscular connective tissue, and adventitia (P < 0.001). The probability for zero displacement values of the carcinomas, on the other hand, were statistically significantly higher than those of the lamina propria mucosae, submucosa, intermuscular connective tissue, and adventitia (P < 0.001). The kurtosis values of the carcinomas were also statistically significantly higher than those of the lamina propria mucosae, submucosa, intermuscular connective tissue, and adventitia (P < 0.001). Depth of Mural Invasion by the Esophageal Carcinomas at q-Space MR Imaging Histopathologic examination of the 20 esophageal carcinomas showed that 2 of the carcinomas were confined to the mucosa (T1a), 6 had invaded the submucosa (T1b), 3 had involved the muscularis propria (T2), and 9 had extended into the adventitia (T3/T4). The depth of

2266

Yamada et al.

FIG. 2. Images of the normal esophageal wall. a: Mean displacement map depicts the normal esophageal wall as consisting of the following eight layers: epithelium (Epi; low SI), lamina propria mucosae (LPM; high SI), muscularis mucosae (MM; low SI), submucosa (SM; high SI), inner circular muscle (ICM; low SI), intermuscular connective tissue (IMCT; high SI), outer longitudinal muscle (OLM; low SI), and adventitia (Adv; high SI). b: Probability for zero displacement map depicts the following eight layers: epithelium (high SI), lamina propria mucosae (low SI), muscularis mucosae (high SI), submucosa (low SI), inner circular muscle (high SI), intermuscular connective tissue (low SI), outer longitudinal muscle (high SI), and adventitia (low SI). a.u. ¼ arbitrary unit. c: Kurtosis map depicts the following eight layers: epithelium (high SI), lamina propria mucosae (low SI), muscularis mucosae (high SI), submucosa (low SI), inner circular muscle (high SI), intermuscular connective tissue (low SI), outer longitudinal muscle (high SI), and adventitia (low SI). d: Histologic section of the normal esophageal wall shows that it consists of alternating solid high-cellularity layers (epithelium, muscularis mucosae, inner circular muscle, and outer longitudinal muscle) and loose low-cellularity layers (lamina propria mucosae, submucosa, intermuscular connective tissue, and adventitia). Hematoxylin and eosin stain; original magnification, 20.

invasion of the esophageal wall by all 20 carcinomas (100%) was clearly demonstrated by the mean displacement maps, probability for zero displacement maps, and kurtosis maps. There were no false-positive or falsenegative cases in this study with respect to the assessments of invasion of the different esophageal layers. Carcinomas confined to the mucosa appeared as discrete thickenings of the mucosal layer that had low mean displacement values, high probability for zero displacement values, and high kurtosis values. Carcinomas that had invaded the submucosa appeared as irregular masses with low mean displacement values, high probability for zero displacement values, and high kurtosis values that had disrupted the muscularis mucosae layer (Fig. 3). Carcinomas that had involved the muscularis propria appeared as low-displacement, high-probability, and high-kurtosis masses that had partially replaced the muscularis propria layer. Carcinomas that had extended into the adventitia appeared as low-displacement, highprobability, and high-kurtosis masses that had com-

pletely disrupted the muscularis propria layer and invaded the adventitial layer (Fig. 4). Histologic Grades, Histologic Types, Nuclear-Cytoplasmic Ratios, and Tumor Cellularity of the Esophageal Carcinomas at q-Space MR Imaging The q-space MR imaging parameters were compared with the histologic grades, histologic types, nuclearcytoplasmic ratios, and tumor cellularity of the 20 esophageal carcinomas (Fig. 5). As shown in Figure 6 and Table 2, a statistically significant inverse correlation was found between the mean displacement values of the esophageal carcinomas and the histologic grades of the esophageal carcinomas (r ¼ 0.648; P < 0.01), and there was a statistically significant positive correlation between the probability for zero displacement values and the histologic grades of the esophageal carcinomas (r ¼ 0.669; P < 0.01). The kurtosis values were also statistically significantly positively correlated with the

q-Space Imaging of Esophageal Carcinoma

2267

FIG. 3. Esophageal carcinoma that has invaded the submucosa. a: Mean displacement map shows a mass lesion (arrows) that has low mean displacement values, which has disrupted the muscularis mucosae layer (arrowheads). b: Probability for zero displacement map shows that the mass lesion (arrows) has high probability for zero displacement values. a.u. ¼ arbitrary unit. c: Kurtosis map shows that the mass lesion (arrows) has high kurtosis values. d: Corresponding histopathologic section shows carcinoma that has invaded the submucosa, which has disrupted the muscularis mucosae layer (arrows). Elastica-van Gieson stain; original magnification, 10.

histologic grades of the esophageal carcinomas (r ¼ 0.643; P < 0.01). One tumor was diagnosed as a mucoepidermoid carcinoma and contained both squamous cell carcinoma and signet-ring cells showing mucus production. Another tumor was diagnosed as a carcinosarcoma and contained both squamous cell carcinoma and sarcomatous changes as well as marked edema and myxomatous matrix deposition. It is noteworthy that both of these tumors had statistically significantly higher mean displacement values, lower probability for zero displacement values, and lower kurtosis values than the respective values of pure squamous cell carcinoma (P < 0.05 for all) (Table 2). As shown in Figure 7, a statistically significant inverse correlation was found between the mean displacement values and the nuclear-cytoplasmic ratios of the esophageal carcinomas (r ¼ 0.712; P < 0.01), and there was a statistically significant positive correlation between the probability for zero displacement values and the nuclearcytoplasmic ratios of the esophageal carcinomas (r ¼ 0.793; P < 0.001). There was also a statistically significant positive correlation between the kurtosis values and the nuclear-cytoplasmic ratios of the esophageal carcinomas (r ¼ 0.728; P < 0.01). In addition, as shown in Figure 7, a statistically significant inverse correlation was found between the mean

displacement values and the tumor cellularity of the esophageal carcinomas (r ¼ 0.667; P < 0.01), and there was a statistically significant positive correlation between the probability for zero displacement values and the tumor cellularity of the esophageal carcinomas (r ¼ 0.695; P < 0.01). There was also a statistically significant positive correlation between the kurtosis values and the tumor cellularity of the esophageal carcinomas (r ¼ 0.738; P < 0.001). Lymph Node Metastasis by the Esophageal Carcinomas at q-Space MR Imaging As shown in Figure 5 and Table 3, the mean displacement values of the metastatic lymph nodes were found to be statistically significantly lower than those of the nonmetastatic lymph nodes (P < 0.01). The probability for zero displacement values of the metastatic lymph nodes were statistically significantly higher than those of the nonmetastatic lymph nodes (P < 0.01). The kurtosis values of the metastatic lymph nodes were also statistically significantly higher than those of the nonmetastatic lymph nodes (P < 0.01). Thus, it may be possible to differentiate between metastatic and nonmetastatic lymph nodes in esophageal carcinoma on the basis of these qspace MR imaging parameters.

2268

Yamada et al.

FIG. 4. Esophageal carcinomas that has extended into the adventitia. a: Mean displacement map clearly shows an irregular-shaped mass lesion (arrows) that has low mean displacement values, which has completely disrupted the muscularis propria layer and invaded the adventitial layer. b: Probability for zero displacement map shows that the mass lesion (arrows) has high probability for zero displacement values. a.u. ¼ arbitrary unit. c: Kurtosis map shows that the mass lesion (arrows) has high kurtosis values. d: Corresponding histopathologic section shows carcinoma that has extended into the adventitia (arrows). Hematoxylin and eosin stain; original magnification, 4.

DISCUSSION q-Space diffusion-weighted MR imaging is a nonGaussian diffusion-weighted imaging method that was first described by Assaf et al in 2000 in relation to the spinal cord (19), and it makes it possible to quantify the deviation of tissue diffusion from Gaussian behavior in restricted water diffusion. As a result, q-space MR imaging is able to more accurately reflect the microstructural complexity of tissue than standard diffusion-weighted imaging (19–21), and it may serve as a more effective method of assessing esophageal carcinomas in detail than standard diffusion-weighted imaging. The mean displacement maps, probability for zero displacement maps, and kurtosis maps generated from the q-space MR imaging data obtained in all 20 specimens (100%) in the present study clearly depicted the normal esophageal wall as consisting of eight layers, and the eight layers matched the eight tissue layers of the esophageal wall observed histologically, thereby clearly demonstrating that q-space MR imaging is capable of depicting the individual layers of the esophageal wall more accurately than CT and EUS can (5–9). Yamada et al (26) have proposed a new schematic illustration of the normal esophageal wall to explain the diffusiontensor MR imaging and tractography findings in the

esophageal wall, and their schematic illustration demonstrates that the alternating levels of cellularity (cell density) between the adjacent layers account for the clear depiction of the layers of the esophageal wall by q-space MR imaging in the present study. Our findings demonstrated that q-space MR imaging makes it possible to determine the extent of esophageal carcinomas by identifying areas in the esophageal wall that have low mean displacement values, high probability for zero displacement values, and high kurtosis values, and the depth of esophageal wall invasion by all 20 carcinomas (100%) was clearly demonstrated by using the mean displacement maps, probability for zero displacement maps, and kurtosis maps. In this connection, Yamada et al (27) have reported finding that diffusiontensor MR imaging and tractography, although based on a Gaussian model in three dimensions, are also useful for determining the depth of mural invasion by esophageal carcinomas. Our data revealed statistically significant correlations between the q-space MR imaging parameters and the histologic grades of the esophageal carcinomas. Previous studies have shown correlations between diffusional kurtosis obtained by diffusion kurtosis imaging (DKI) of cerebral gliomas and prostate cancer and the histologic

q-Space Imaging of Esophageal Carcinoma

2269

FIG. 5. Histopathologic sections of the esophageal carcinomas. a: Histopathologic section of a well-differentiated squamous cell carcinoma. Hematoxylin and eosin stain; original magnification, 200. b: Histopathologic section of a poorly-differentiated squamous cell carcinoma. Hematoxylin and eosin stain; original magnification, 200. c: Histopathologic section of tumor cells showing low nuclearcytoplasmic ratios. Elastica-van Gieson stain; original magnification, 400. d: Histopathologic section of tumor cells showing high nuclear-cytoplasmic ratios. Elastica-van Gieson stain; original magnification, 400. e: Histopathologic section of a nonmetastatic lymph node. Hematoxylin and eosin stain; original magnification, 40. f: Histopathologic section of a metastatic lymph node. Hematoxylin and eosin stain; original magnification, 20.

grades of the tumors (22,23), and the results of those studies suggested that diffusional kurtosis measured by DKI would increase with the aggressiveness of the tumor because of the greater microstructural complexity of higher grade tumors (23). Because q-space MR imaging is exquisitely sensitive to changes in tissue microstructures (19–21), q-space MR imaging may serve as an effective method of noninvasively assessing the histologic grades of esophageal carcinomas. Our data also demonstrated statistically significant correlations between the q-space MR imaging parameters and the nuclear-cytoplasmic ratios of the esophageal carcinomas. As is well known in histopathology, the

nuclear-cytoplasmic ratio is one of the most important histologic indicators of the degree of tumor malignancy and is a critical factor in determining the histologic grade of tumors. To our knowledge, the present study is the first to clearly demonstrate the usefulness of q-space MR imaging as a means of assessing the nuclearcytoplasmic ratio in all types of carcinoma. The q-space MR imaging parameters in the present study were also correlated with the tumor cellularity of the esophageal carcinomas. Previous studies have shown that while ADC values may be correlated with tumor cellularity, it is often difficult to differentiate histologic tumor grades on the basis of the ADC values alone,

2270

Yamada et al.

FIG. 6. Box plots of the q-space MR imaging parameters in the different histologic grades of esophageal carcinomas. a: Comparison of the mean displacement values in the different histologic grades of esophageal carcinomas showing a statistically significant inverse correlation (r ¼ 0.648; P < 0.01). b: Comparison of the probability for zero displacement values in the different histologic grades of esophageal carcinomas showing a statistically significant positive correlation (r ¼ 0.669; P < 0.01). a.u. ¼ arbitrary unit. c: Comparison of the kurtosis values in the different histologic grades of esophageal carcinomas showing a statistically significant positive correlation (r ¼ 0.643; P < 0.01).

because there is considerable overlap between the ADC values of different histologic tumor grades (31–33). Based on the potential of q-space MR imaging to probe the nuclear-cytoplasmic ratio and tumor cellularity, it may become a more accurate method of differentiating between histologic tumor grades than standard diffusionweighted imaging. The three q-space imaging parameters show monotonic correlations with the histologic grades, nuclear-cytoplasmic ratios, and tumor cellularity, but they express restricted diffusion in different ways. In particular, the mean displacement provides the real range of restricted diffusion on the micrometer scale (19). Our findings also demonstrated that q-space MR imaging makes it possible to differentiate between metastatic lymph nodes and nonmetastatic lymph nodes in esophageal carcinoma. Previous reports have indicated that evaluating lymph nodes in esophageal carcinoma by all kinds of imaging techniques is challenging, because lymph node size alone is not a reliable diagnostic criterion for metastatic involvement (34). Recent studies have shown that although ADC values may be useful for evaluating lymph nodes for metastasis, it is often difficult to differentiate between metastatic lymph nodes and nonmetastatic lymph nodes on the basis of the ADC values alone because of the considerable overlap between the

ADC values of metastatic lymph nodes and nonmetastatic lymph nodes (35–38). Based on its ability to predict microstructural complexity q-space MR imaging may provide a tool for noninvasive assessment of lymph node metastasis in esophageal carcinoma. Conventional MR imaging has recently been used to preoperatively stage esophageal carcinomas. Riddell et al (15–18) performed high-spatial-resolution MR imaging with an external surface coil in esophageal carcinoma patients and demonstrated its potential as a feasible alternative technique for staging esophageal carcinoma, and Dave et al (14) have performed endoluminal MR imaging with a surface coil incorporated into the tip of an endoscope in esophageal carcinoma patients. We therefore think that using an external surface coil technique or endoluminal surface coil technique would make it possible to perform q-space MR imaging in esophageal carcinoma patients. There were several limitations in our study. First, our study was performed ex vivo, and the specimens were imaged after fixation in formalin. However, previous studies of other organs showed that although the ADC values of fixed tissues were lower than in vivo, the relative ADC difference between different tissue types in vivo was preserved in fixed tissues, and the difference in relative ADC values between fixed tissues and in vivo

Table 2 Mean Displacement Values, Probability for Zero Displacement Values, and Kurtosis Values of the Different Histologic Grades and Histologic Types of the Esophageal Carcinomas at q-Space Diffusion-Weighted MR Imaging Histologic grades and types WD (n ¼ 2) MD-WD (n ¼ 2) MD (n ¼ 11) PD-MD (n ¼ 2) PD (n ¼ 1) Mucoepidermoid carcinoma (n ¼ 1) Carcinosarcoma (n ¼ 1) a

Mean displacement values (mm)

Probability for zero displacement values (a.u.)

Kurtosis values (a.u.)

6.47 6 0.29a 6.03 6 0.14 5.81 6 0.11 5.77 6 0.34 5.42 6.61 7.66

47.2 6 2.9a 49.7 6 1.2 51.1 6 1.8 51.5 6 0.3 54.0 45.0 37.2

46.8 6 3.9a 50.7 6 3.3 54.5 6 1.8 55.4 6 4.7 60.3 45.6 35.2

¼ Significantly different for the different histologic grades of the esophageal carcinomas. WD ¼ well differentiated, MD ¼ moderately differentiated, PD ¼ poorly differentiated. a.u. ¼ arbitrary unit.

q-Space Imaging of Esophageal Carcinoma

2271

FIG. 7. Scatter plots of the nuclear-cytoplasmic ratios or tumor cellularity versus the q-space MR imaging parameters in esophageal carcinomas. a: Scatter plot of the nuclear-cytoplasmic ratios versus the mean displacement values in esophageal carcinomas showing a statistically significant inverse correlation (r ¼ 0.712; P < 0.01). b: Scatter plot of the nuclear-cytoplasmic ratios versus the probability for zero displacement values in esophageal carcinomas showing a statistically significant positive correlation (r ¼ 0.793; P < 0.001). a.u. ¼ arbitrary unit. c: Scatter plot of the nuclear-cytoplasmic ratios versus the kurtosis values in esophageal carcinomas showing a statistically significant positive correlation (r ¼ 0.728; P < 0.01). d: Scatter plot of the tumor cellularity versus the mean displacement values in esophageal carcinomas showing a statistically significant inverse correlation (r ¼ 0.667; P < 0.01). e: Scatter plot of the tumor cellularity versus the probability for zero displacement values in esophageal carcinomas showing a statistically significant positive correlation (r ¼ 0.695; P < 0.01). f: Scatter plot of the tumor cellularity versus the kurtosis values in esophageal carcinomas showing a statistically significant positive correlation (r ¼ 0.738; P < 0.001).

tissues was not statistically significant (39–41). We, therefore, think that the data obtained in the present study can be applied to q-space MR imaging of tissues in vivo as well as formalin-fixed tissues. The second limitation was that the imaging time in this study was very long (43 min), and shortening the scan time would be necessary to translate our data into in vivo q-space MR imaging. Clinical application of qspace MR imaging to patients with esophageal carcinoma in vivo may be technically difficult because of motion effects (peristalsis, respiratory movements, and patient motion) that would mean the possibility of artifacts, and

modifications of pulse sequences, development of faster MR imaging techniques, or the application of higher field strengths to clinical settings in the future may make it more technically feasible. In addition, higher gradient strengths also might be needed in translation of q-space imaging to clinical use, because the low gradient strengths in current clinical scanners may be insufficient to generate the high q values used in the q-space imaging. If data sampling were limited to that achievable in clinical scanners, the possibility of the histologic grade differentiation and nodal differentiation would be reduced considerably. In such a

Table 3 Mean Displacement Values, Probability for Zero Displacement Values, and Kurtosis Values of the Metastatic Lymph Nodes and Nonmetastatic Lymph Nodes of the Esophageal Carcinomas at q-Space Diffusion-Weighted MR Imaging Lymph nodes Nonmetastatic (n ¼ 6) Metastatic (n ¼ 8) a

Mean displacement values (mm)

Probability for zero displacement values (a.u.)

Kurtosis values (a.u.)

8.82 6 0.56a 6.57 6 0.48

34.2 6 1.5a 45.3 6 5.0

30.6 6 1.4a 45.6 6 5.7

¼ Significantly different between the metastatic lymph nodes and nonmetastatic lymph nodes of the esophageal carcinomas. a.u. ¼ arbitrary unit.

2272

situation, however, the DKI method may be helpful for this purpose because it is an alternative non-Gaussian method for the limited data sampling (22,23). The third limitation was that the influences of microperfusion were not included in the present ex vivo specimens. There might be some influences on the quantitative data if microperfusion would be added. For this purpose, however, the combined use of q-space imaging and intravoxel incoherent motion (IVIM) method might be helpful (42), because the IVIM method is capable of characterizing microperfusion and restricted diffusion simultaneously (43). The fourth limitation was the rather limited number of specimens examined in this study, and this leaves the question open on the reproducibility; we should therefore continue to collect more esophageal q-space MR imaging data. Also, although 13 patients underwent no neo-adjuvant treatment, seven other patients underwent neo-adjuvant treatment (five patients, chemotherapy alone; two patients, chemotherapy and radiation therapy); thus the neo-adjuvant treatment may have caused necrosis and fibrosis in the seven patients. However, our purpose was to compare the q-space imaging findings and histopathologic findings at the same point as surgical resection, so our comparison seems compatible with this purpose. CONCLUSIONS It should be noted that the present study was done in limited and probably selected patient group and that the results cannot be extended to tumors treated by chemoradiation. However, the results of the present study have demonstrated that q-space MR imaging is capable of depicting the individual tissue layers of the esophageal wall ex vivo. q-Space MR imaging ex vivo provides excellent diagnostic accuracy for evaluating mural invasion by esophageal carcinomas, the histologic grades of esophageal carcinomas, and lymph node metastasis by esophageal carcinomas. REFERENCES 1. Esophageal Cancer. Center for Cancer Control and Information Services. National Cancer Center of Japan Web site. http://ganjoho.jp/public/cancer/esophagus/index.html. Published November 25, 1996. Updated January 6, 2013. Accessed August 20, 2013. 2. Rice TW, Rusch VW, Ishwaran H, Blackstone EH; Worldwide Esophageal Cancer Collaboration. Cancer of the esophagus and esophagogastric junction: data-driven staging for the seventh edition of the American Joint Committee on Cancer/International Union Against Cancer Cancer Staging Manuals. Cancer 2010;116:3763–3773. 3. Edge SB, Byrd DR, Compton CC, Fritz AG, Greene FL, Trotti A, editors. AJCC cancer staging manual. 7th ed. New York: Springer; 2010. p 103–115. 4. Treatment Option. Esophageal Cancer. Center for Cancer Control and Information Services. National Cancer Center of Japan Web site. http://ganjoho.jp/public/cancer/esophagus/treatment_option.html. Published November 25, 1996. Updated December 21, 2012. Accessed August 20, 2013. 5. Panebianco V, Grazhdani H, Iafrate F, Petroni M, Anzidei M, Laghi A, Passariello R. 3D CT protocol in the assessment of the esophageal neoplastic lesions: can it improve TNM staging? Eur Radiol 2006;16: 414–421. 6. Onbas O, Eroglu A, Kantarci M, Polat P, Alper F, Karaoglanoglu N, Okur A. Preoperative staging of esophageal carcinoma with multidetector CT and virtual endoscopy. Eur J Radiol 2006;57:90–95.

Yamada et al. 7. Wakelin SJ, Deans C, Crofts TJ, Allan PL, Plevris JN, Paterson-Brown S. A comparison of computerised tomography, laparoscopic ultrasound and endoscopic ultrasound in the preoperative staging of oesophago-gastric carcinoma. Eur J Radiol 2002;41:161–167. 8. Preston SR, Clark GW, Martin IG, Ling HM, Harris KM. Effect of endoscopic ultrasonography on the management of 100 consecutive patients with oesophageal and junctional carcinoma. Br J Surg 2003; 90:1220–1224. 9. Schlick T, Heintz A, Junginger T. The examiner’s learning effect and its influence on the quality of endoscopic ultrasonography in carcinoma of the esophagus and gastric cardia. Surg Endosc 1999;13:894– 898. 10. Yamada I, Murata Y, Izumi Y, Kawano T, Endo M, Kuroiwa T, Shibuya H. Staging of esophageal carcinoma in vitro with 4.7-T MR imaging. Radiology 1997;204:521–526. 11. Yamada I, Izumi Y, Kawano T, Yoshino N, Tetsumura A, Ohashi K, Shibuya H. Superficial esophageal carcinoma: an in vitro study of high-resolution MR imaging at 1.5T. J Magn Reson Imaging 2001;13: 225–231. 12. Yamada I, Izumi Y, Kawano T, Yoshino N, Tetsumura A, Kumagai J, Shibuya H. Esophageal carcinoma: evaluation with high-resolution three-dimensional constructive interference in steady state MR imaging in vitro. J Magn Reson Imaging 2006;24:1326–1332. 13. Feldman DR, Kulling DP, Hawes RH, Kay CL, Muckenfuss VR, Cotton PB, Bohning DE, Young JW. MR endoscopy: preliminary experience in human trials. Radiology 1997;202:868–870. 14. Dave UR, Williams AD, Wilson JA, Amin Z, Gilderdale DJ, Larkman DJ, Thursz MR, Taylor-Robinson SD, deSouza NM. Esophageal cancer staging with endoscopic MR imaging: pilot study. Radiology 2004; 230:281–286. 15. Riddell AM, Richardson C, Scurr E, Brown G. The development and optimization of high spatial resolution MRI for imaging the oesophagus using an external surface coil. Br J Radiol 2006;79: 873–879. 16. Riddell AM, Hillier J, Brown G, King DM, Wotherspoon AC, Thompson JN, Cunningham D, Allum WH. Potential of surface-coil MRI for staging of esophageal cancer. AJR Am J Roentgenol 2006;187: 1280–1287. 17. Riddell AM, Allum WH, Thompson JN, Wotherspoon AC, Richardson C, Brown G. The appearances of oesophageal carcinoma demonstrated on high-resolution, T2-weighted MRI, with histopathological correlation. Eur Radiol 2007;17:391–399. 18. Riddell AM, Davies DC, Allum WH, Wotherspoon AC, Richardson C, Brown G. High-resolution MRI in evaluation of the surgical anatomy of the esophagus and posterior mediastinum. AJR Am J Roentgenol 2007;188:W37–W43. 19. Assaf Y, Mayk A, Cohen Y. Displacement imaging of spinal cord using q-space diffusion-weighted MRI. Magn Reson Med 2000;44: 713–722. 20. Bar-Shir A, Duncan ID, Cohen Y. QSI and DTI of excised brains of the myelin-deficient rat. Neuroimage 2009;48:109–116. 21. Anaby D, Duncan ID, Smith CM, Cohen Y. White matter maturation in the brains of Long Evans shaker myelin mutant rats by ex-vivo QSI and DTI. Magn Reson Imaging 2013;31:1097–1104. 22. Raab P, Hattingen E, Franz K, Zanella FE, Lanfermann H. Cerebral gliomas: diffusional kurtosis imaging analysis of microstructural differences. Radiology 2010;254:876–881. 23. Rosenkrantz AB, Sigmund EE, Johnson G, Babb JS, Mussi TC, Melamed J, Taneja SS, Lee VS, Jensen JH. Prostate cancer: feasibility and preliminary experience of a diffusional kurtosis model for detection and assessment of aggressiveness of peripheral zone cancer. Radiology 2012;264:126–135. 24. Yuequan J, Shifeng C, Bing Z. Prognostic factors and family history for survival of esophageal squamous cell carcinoma patients after surgery. Ann Thorac Surg 2010;90:908–913. 25. Liu J, Xie X, Zhou C, Peng S, Rao D, Fu J. Which factors are associated with actual 5-year survival of oesophageal squamous cell carcinoma? Eur J Cardiothorac Surg 2012;41:e7–e11. 26. Yamada I, Hikishima K, Miyasaka N, et al. Diffusion-tensor MRI and tractography of the esophageal wall ex vivo. J Magn Reson Imaging 2014;40:567–576. 27. Yamada I, Hikishima K, Miyasaka N, Kawano T, Tokairin Y, Ito E, Kobayashi D, Eishi Y, Okano H. Esophageal carcinoma: ex vivo evaluation with diffusion-tensor MR imaging and tractography at 7 T. Radiology 2014. doi:10.1148/radiol.14132170.

q-Space Imaging of Esophageal Carcinoma 28. Hikishima K, Yagi K, Numano T, Homma K, Nitta N, Nakatani T, Hyodo K. Volumetric q-space imaging by 3D diffusion-weighted MRI. Magn Reson Imaging 2008;26:437–445. 29. Cohen Y, Assaf Y. High b-value q-space analyzed diffusion-weighted MRS and MRI in neuronal tissues - a technical review. NMR Biomed 2002;15:516–542. 30. Ong HH, Wright AC, Wehrli SL, Souza A, Schwartz ED, Hwang SN, Wehrli FW. Indirect measurement of regional axon diameter in excised mouse spinal cord with q-space imaging: simulation and experimental studies. Neuroimage 2008;40:1619–1632. 31. Sugahara T, Korogi Y, Kochi M, et al. Usefulness of diffusionweighted MRI with echo-planar technique in the evaluation of cellularity in gliomas. J Magn Reson Imaging 1999;9:53–60. 32. Kono K, Inoue Y, Nakayama K, Shakudo M, Morino M, Ohata K, Wakasa K, Yamada R. The role of diffusion-weighted imaging in patients with brain tumors. AJNR Am J Neuroradiol 2001;22:1081–1088. 33. Koral K, Mathis D, Gimi B, Gargan L, Weprin B, Bowers DC, Margraf L. Common pediatric cerebellar tumors: correlation between cell densities and apparent diffusion coefficient metrics. Radiology 2013;268: 532–537. 34. Li R, Chen TW, Hu J, Guo DD, Zhang XM, Deng D, Li H, Chen XL, Tang HJ. Tumor volume of resectable adenocarcinoma of the esophagogastric junction at multidetector CT: association with regional lymph node metastasis and N stage. Radiology 2013;269:130–138. 35. Sumi M, Sakihama N, Sumi T, et al. Discrimination of metastatic cervical lymph nodes with diffusion-weighted MR imaging in patients with head and neck cancer. AJNR Am J Neuroradiol 2003;24:1627– 1634.

2273 36. King AD, Ahuja AT, Yeung DK, Fong DK, Lee YY, Lei KI, Tse GM. Malignant cervical lymphadenopathy: diagnostic accuracy of diffusion-weighted MR imaging. Radiology 2007;245:806–813. 37. Vandecaveye V, De Keyzer F, Vander Poorten V, Dirix P, Verbeken E, Nuyts S, Hermans R. Head and neck squamous cell carcinoma: value of diffusion-weighted MR imaging for nodal staging. Radiology 2009; 251:134–146. 38. Wu LM, Xu JR, Hua J, Gu HY, Zhu J, Hu J. Value of diffusionweighted MR imaging performed with quantitative apparent diffusion coefficient values for cervical lymphadenopathy. J Magn Reson Imaging 2013;38:663–670. 39. Xu J, Humphrey PA, Kibel AS, Snyder AZ, Narra VR, Ackerman JJ, Song SK. Magnetic resonance diffusion characteristics of histologically defined prostate cancer in humans. Magn Reson Med 2009;61: 842–850. 40. Bourne R, Kurniawan N, Cowin G, Sved P, Watson G. 16 T diffusion microimaging of fixed prostate tissue: preliminary findings. Magn Reson Med 2011;66:244–247. 41. Williams MC, Does MD, Price RR. ADC mapping of whole excised human prostate. Proc Intl Soc Magn Reson Med 2004;11:2064. 42. Lu Y, Jansen JF, Mazaheri Y, Stambuk HE, Koutcher JA, Shukla-Dave A. Extension of the intravoxel incoherent motion model to nongaussian diffusion in head and neck cancer. J Magn Reson Imaging 2012;36:1088–1096. 43. Yamada I, Aung W, Himeno Y, Nakagawa T, Shibuya H. Diffusion coefficients in abdominal organs and hepatic lesions: evaluation with intravoxel incoherent motion echo-planar MR imaging. Radiology 1999;210:617–623.