JOURNAL OF MAGNETIC RESONANCE IMAGING 28:128 135 (2008)

Original Research

Detection of Lymph Node Metastasis in Cervical and Uterine Cancers by Diffusion-Weighted Magnetic Resonance Imaging at 3T
Gigin Lin, MD,1 Kung-Chu Ho, MD,2 Jiun-Jie Wang, PhD,1 Koon-Kwan Ng, MD,1 Yau-Yau Wai, MD,1 Yen-Ting Chen, MS,3 Chee-Jen Chang, PhD,3 Shu-Hang Ng, MD,1 Chyong-Huey Lai, MD,4 and Tzu-Chen Yen, MD, PhD2*
Purpose: To evaluate diffusion-weighted imaging (DWI) for detection of pelvic lymph node metastasis in patients with cervical and uterine cancers. Materials and Methods: Fifty patients scheduled for pelvic lymph node dissection were enrolled for 3T magnetic resonance imaging (MRI) using a single-shot echo-planar DWI technique, body-phased array coil, b 0, 1000 s/mm2. We measured short/long-axis diameters, mean apparent diffusion coefcient (ADC) values of all identiable nodes, relative ADC values between tumors and nodes, and utilized their cutoff values to validate the diagnostic accuracy internally. Histopathologic results served as the reference standard. Results: The relative ADC values between tumor and nodes were signicantly lower in metastatic than in benign nodes (0.06 vs. 0.21 10 3 mm2/s, P 0.001; cutoff value 3 2 0.10 10 mm /s). Compared to conventional MRI, the method combining size and relative ADC values resulted in better sensitivity (25% vs. 83%) and similar specicity (98% vs. 99%). The smallest metastatic lymph node detected by this method measured 5 mm on its short axis. Conclusion: The combination of size and relative ADC values was useful in detecting pelvic lymph node metastasis in patients with cervical and uterine cancers. Key Words: cervical cancer; diffusion-weighted imaging; lymph node metastasis; magnetic resonance imaging; 3T; uterine cancer J. Magn. Reson. Imaging 2008;28:128 135. 2008 Wiley-Liss, Inc.

1 Department of Radiology, Chang Gung Memorial Hospital and Chang Gung University College of Medicine, Taoyuan, Taiwan. 2 Department of Nuclear Medicine and Molecular Imaging Center, Chang Gung Memorial Hospital and Chang Gung University College of Medicine, Taoyuan, Taiwan. 3 Division of Biostatistics, Department of Obstetrics and Gynecology, Chang Gung Memorial Hospital and Chang Gung University College of Medicine, Taoyuan, Taiwan. 4 Division of Gynecologic Oncology, Chang Gung Memorial Hospital and Chang Gung University College of Medicine, Taoyuan, Taiwan. Contract grant sponsor: Chang Gung Memorial Hospital; Contract grant number: CMRPG-33111. Presented in part at the Joint Annual Meeting International Society for Magnetic Resonance in Medicine and the European Society for Magnetic Resonance in Medicine and Biology, Berlin, Germany, 19 25 May 2007. *Address reprint requests to: T.-C.Y., Department of Nuclear Medicine, Chang Gung Memorial Hospital, No. 5 Fu Hsing Street, Kwei Shan Hsiang, Taoyuan 333, Taiwan. E-mail: yen1110@adm.cgmh.org.tw Received August 27, 2007; Accepted March 7, 2008. DOI 10.1002/jmri.21412 Published online in Wiley InterScience (www.interscience.wiley.com).

THE PRESENCE OF lymph node metastasis is an important issue for patients with cervical and uterine cancers since it inuences the 5-year survival and affects treatment planning (1,2). A threshold diameter of 10 mm in the short axis is commonly applied in magnetic resonance (MR) imaging for distinguishing metastatic from benign nodes, with sensitivity ranging from 24% 73% (37). It follows that this cutoff cannot be considered completely satisfactory in the evaluation of nodal status in this patient group. Although a lymph nodespecic MR contrast has been demonstrated to achieve a higher diagnostic accuracy (6), it is currently not available in clinical practice. Diffusion-weighted imaging (DWI) is an MR technique that, at a microscopic level, can identify differences in the variation of molecular water mobility in extracellular spaces (8). The extent of water diffusion can be related to the cell organization and density, microstructure, and microcirculation. DWI can provide information about the biophysical properties of tissues in vivo. In oncology, DWI has proved to be valuable for tissue characterization and monitoring treatment response (9 11). Therefore, it may provide potential in detecting nodal metastasis. In this study we propose a method for detecting nodal metastases with the use of DWI. We also wanted to determine its accuracy in patients with cervical and uterine cancers. MATERIALS AND METHODS Our Institutional Review Board approved the protocol of this retrospective study. We added a DWI sequence for

2008 Wiley-Liss, Inc.

128

LN Metastasis in Cervix/Uterus CA: 3T DWI

129

patients who had an MR examination for primary local tumor staging of cervical or uterine cancers scheduled for surgical treatment and lymphadenectomy. Patients with the following conditions were excluded: 1) presence of any MR contraindication (cardiac pacemaker or debrillator, insulin pump, aneurysm clip, implanted neural stimulator, cochlear implant, metal shrapnel or bullet, etc.); 2) presence of pelvis or hip metal prostheses; 3) inability to give informed consent; 4) younger than 18 years of age. MR Images were acquired with a 3T MR scanner (Tim Trio, Siemens, Erlangen, Germany). We used the lower nine elements of the integrated spine coil and the lower six elements of the body-phased array coil to cover the entire pelvis. DWI was obtained in the axial plane using a single-shot spin-echo echo-planar with chemicalshift selective fat-suppression technique. The b-value was chosen to be 0 and 1000 s/mm2 to optimize the signal-to-noise ratio (SNR) in the DWI using the formula proposed by Xing et al (12). The diffusion-weighting gradients were applied in all three orthogonal directions, which were coincident with the slice-selective, phase-encoding, and readout directions of the gradient. The apparent diffusion coefcient (ADC) values were calculated as the average of the diffusion coefcients measured in each direction: ADC (D xx D yy D zz)/3 [1]

suspected lesions. Anatomical recording was based on the left and right common iliac, external iliac, and internal iliac (including obturator and parametrial) regions. The largest nodule with hyperintensities on DWI (b 1000 s/mm2) in each region was termed an identiable node. Short-axis and long-axis diameters of each identiable node were measured using electronic calipers on the workstation. The short-to-long axis ratio was calculated for each identiable node. The ADC values of each identiable node and primary tumor were measured by manually placed irregular regions of interest (ROIs) within main tumors and lymph nodes on the ADC maps. Two readers performed the measurement independently. Consensus was reached after discussion. The ADC values obtained from the two readers were averaged. Histopathology All patients underwent standard surgical procedures, the only exception being two subjects with adenocarcinoma operated on in early Stage IIB due to another ongoing study protocol. Surgeons with prior knowledge of MRI data carefully searched for any possible metastasis at the pelvic lymph node dissection. The resected nodes were anatomically labeled in left or right internal, external, and common iliac regions by the surgeons. A consensus read between surgeons, pathologists, and radiologists for providing the most accurate assignment between MR and histopathology was performed in weekly panel discussion of gynecologic oncology. The lymph nodes were then cut into parallel slices of thickness 23 mm. All nodal tissue was routinely processed and embedded in parafn, followed by staining with hematoxylin and eosin. The histopathology report included the number of total harvested lymph nodes and the identied metastatic ones in each region, respectively. An average of 21 nodes per patient were harvested from the pelvic sidewalls (range: 2 48, total 1038). Data Analysis Data were analyzed with SPSS v11.0 (Chicago, IL) and MedCalc for Windows, v9.2.0.0 (MedCalc Software, Mariakerke, Belgium). Receiver operating characteristic (ROC) curve analysis was used to determine the cutoff value where sensitivity and specicity reach the peak, in distinguishing benign and metastatic nodes in terms of the mean absolute ADC value (aADC), the relative ADC value between tumor and lymph node (rADC) by subtracting their mean values, the length of the short axis, the length of the long axis, as well as the short-to-long axis ratio. A conventional MR criterion and three proposed DWI criteria were used for evaluation of nodal metastasis. Their denitions were as follows: 1) The MRI criterion was based on general consideration of size, morphology, and margin of lymph nodes. Lymph nodes with a short-axis diameter larger than 10 mm, round-shape, lobulated, or spiculated margin indicated metastatic lymph nodes.7 Consensus of two board-certied radiologiststhe same who read the DWIwas recorded in a

where Dxx, Dyy, and Dzz indicate the apparent diffusion coefcients measured in readout, phase-encoding, and slice selection directions, respectively. The sequence parameters were: repetition time /echo time 3300 msec / 79 msec; with four averages. The acceleration factor for the parallel imaging was 2. The matrix size was 128 128 with a eld of view (FOV) of 20 cm and slice thickness 4 mm to cover the entire pelvis. The total acquisition time was 5 minutes. Axial turbo spin-echo (TSE) T1 (600/9.4; matrix 256 320, FOV 20 cm, 2 averages) and axial and sagittal TSE T2-weighted images (5630/87, matrix 256x320, FOV 20 cm, 3 averages), with the slice thickness and sequence planning identical for DWI, were obtained for anatomical reference. The entire study was performed during normal respiration. Image Analysis The ADC value was then reconstructed for each pixel using b-factors of 0 and 1000 s/mm2 using the standard software on the console (Syngo, Siemens). DWI and T2-weighted images were fused with a dedicated image processing software (OsiriX v2.4, Los Angeles, CA). Readers were two board-certied radiologists with experience in gynecologic radiology of 19 and 3 years, respectively. They were familiar with brain DWI but new to female pelvis DWI. They reviewed MR images on an off-line Mac OS workstation (PowerBook G4, Apple, Cupertino, CA), blind to clinical data. T2-weighted images were used to determine the size and location of the any

130 Table 1 Patient Characteristics (n

Lin et al.

50) Uterine Carcinomas Cervical Carcinomas 24 42.3 8.9 2766 22.1 9.8 235 12.8 11.3 243 0 20 2 2 0 0 26.6 19.9 1094 10 10 3 1

No. of patients, n Age, years Mean SD Range Nodes sampled (per patient), n Mean SD Range Interval between MRI and surgery, days Mean SD Range FIGO stage (International Federation of Gynecology and Obstetrics) Ia Ib Iia Iib IIIa IIIc Long axis tumor size, mm Mean SD Range Histopathology Adenocarcinoma Squamous cell carcinoma Adenosquamous carcinoma Other histological types (carcinosarcoma, small cell carcinoma)

26 56.5 9.7 2575 19.5 8.9 548 5.8 3.5 114 4 15 1 1 1 4 35.0 19.6 880 22 0 1 3

weekly panel discussion of the gynecologic oncology group. For ROC curve analysis, a ve-point scale was developed to indicate the level of condence in the diagnosis: score 0, no lesion; score 1, denitely benign; score 2, probably benign; score 3, probably malignant; score 4, denitely malignant. 2) The aADC criterion was based on the cutoff for the absolute mean ADC value of lymph nodes, as follows: score 0, no identiable node; score 1, between 50% and 25% of the cutoff; score 2, between 25% and the cutoff value; score 3, between the cutoff and 25% of the cutoff; score 4, between 25% and 50% of the cutoff. 3) The rADC criterion, which was based on the cutoff value of the relative mean ADC value between tumors and lymph nodes; the scoring system was in keeping with the aADC criterion. 4) The cDWI criterion that was developed by combining the measurement of rADC and size, including the diameter of the short-axis, long-axis, and the short-tolong axis ratio. The diagnostic performances of criteria from the conventional MR, aADC, rADC, and cDWI were systematically assessed based on a nodal, region, or patient basis. The result from histopathology served as a standard reference. Scores 0, 1, and 2 were considered negative for nodal metastasis, whereas scores 3 and 4 were considered positive. True-positives (TP) and falsepositives (FP) were dened as malignant lesions as classied on imaging with a pathologic diagnosis of malignancy and benignity, respectively. False-negatives (FN) and true-negatives (TN) were dened as benign lesions as classied on imaging with a pathologic diagnosis of malignancy and benignity, respectively. Sensitivity was calculated as TP/(TP FN), specicity was calculated as TN/(TN FP), and the accuracy was (TP TN)/(TP

TN FP FN). The McNemar test was used to compare the sensitivity and specicity of the criterion in comparison. The exact binomial condence limits were calculated for the sensitivity, specicity, positive, and negative predictive values of metastatic lymph nodes for the region and patient analyses. The ROC curves including mean areas under the ROC curves were calculated for method comparison on a patient basis.

RESULTS A cohort of 60 consecutive patients completed the DWI protocol between June 2006 and January 2007. We excluded from the nal analysis the rst three patients whose scans were used for the initial learning curve, as well as another patient with severe image distortion by artifacts. We also excluded six patients with primary a tumor measuring less than 4 mm on its short-axis diameter. A total of 50 patients were entered into the nal analysis. Table 1 shows the general characteristics of the study participants, as well as their disease distribution according to nal histopathologic diagnosis. The 50 exams used displayed satisfactory image quality without any remarkable distortion in the axial planes. All images were usable for ROI drawing and ADC measurement in both primary tumors and lymph nodes. Seven (14%) of the 50 patients had pathological evidence of lymph node metastases. Three patients had nodal involvement from cervical cancer (two adenocarcinoma and one adenosquamous carcinoma), two from endometrial cancer (one adenosquamous carcinoma, and one adenocarcinoma), and two from uterine carcinosarcomas. Metastatic lymph nodes were evident in

LN Metastasis in Cervix/Uterus CA: 3T DWI Table 2 Results of Identiable Nodes on DWI (n

131

83) Histologically Metastatic Histologically Benign 71 0.21 0.18 0.000.74 0.001 P-value

No. of identiable nodes Relative ADC difference, 10 Mean SD Range Cutoff value ADC value, 10-3 mm2/s Mean SD Range Cutoff value Short axis, mm Mean SD Range Cutoff value Long axis, mm Mean SD Range Cutoff value Short-to-long axis ratio Mean SD Range Cutoff value

12
3

mm2/s 0.06 0.03 0.010.10 0.10 0.83 0.15 0.601.06 0.78 6.4 2.6 313 4.5 7.7 2.7 315 11.5 0.72 0.15 0.51 0.61

0.75 0.19 0.461.50

0.639

4.0 1.9 111

0.171

9.2 3.7 415

0.113

0.55 0.20 0.11

0.131

12 (4%) of the 300 nodal regions (six in left and six in right external iliac regions). Different presentations of benign and malignant lymph nodes are summarized in Table 2. Metastatic nodes show a trend toward higher absolute ADC values than their benign counterparts, with a cutoff of 0.78 10 3 mm2/s ( 25% points: 0.60 10 3 mm2/s and 3 2 0.90 10 mm /s). Hence, the relative ADC values between tumor and nodes were closer for metastatic nodes, with a cutoff value of 0.10 10 3 mm2/s ( 3 2 25% points: 0.20 10 mm /s and 0.05 10 3 2 mm /s). The denition of the cDWI criterion is provided in the owchart (Fig. 1). All lymph nodes with a short-axis diameter larger than 4 mm were detected by DWI, regardless of their benign or metastatic nature. The 26 absolute ADC values for benign lymph nodes with a short-axis diameter 4 mm versus 4 mm were (mean SD) 0.77 0.21 10 3 mm2/s versus 0.69 0.14 10 3 mm2/s, a difference which failed to reach the statistical signicance (P 0.269). The diagnostic performances of different criteria for the detection of pelvic lymph node metastasis are summarized in Table 3. On a region basis, the sensitivities for the detection of malignancy from aADC, rADC, and cDWI were 75% (P 0.0313), 100% (P 0.0039), and 83% (P 0.0156), respectively, and signicantly higher than that of the conventional MR (25%). Both aADC and rADC showed a reduced specicity compared to the conventional MR criterion (both P 0.0001). In contrast, no statistically signicant difference in specicity was found between cDWI and conventional MR (P 0.6547). Moreover, on a patient basis the sensitivity by aADC, rADC, and cDWI were all signicantly improved when compared to conventional MR. Among aADC, rADC, and cDWI criteria, only cDWI maintained a level of specicity as high as conventional MR.

Figure 2 displays the ROC curve analysis for the patient-based analysis. The area under the ROC curve was signicantly higher for cDWI (0.965; 95% condence interval [CI]: 0.870, 0.995) compared to the rADC (0.832, 95% CI: 0.700, 0.923), aADC (0.679, 95% CI: 0.532, 0.804), or conventional MR (0.679, 95% CI:

Figure 1. Flow chart shows the process of cDWI criteria. SA, short-axis diameter; LA, long-axis diameter.

132 Table 3 Diagnostic Performance of Different Criteria for the Detection of Pelvic Lymph Node Metastasis Criteria Node basis, n 1038 MRI aADC rADC cDWI Region basis, n 300 MRI aADC rADC cDWI Patient basis, n 50 MRI aADC rADC cDWI FN TP TN FP Sn, % (95% CI) Sp, % (95% CI) Acc, % (95% CI) 95.8 (94.4, 87.6 (85.4, 89.5 (87.5, 98.5 (97.5, 95.7 (92.7, 88.3 (84.1, 90.7 (86.8, 98.3 (96.2, 86.0 (73.3, 62.0 (47.2, 64.0 (49.2, 94.0 (83.5, PPV, % (95% CI)

Lin et al.

NPV, % (95% CI)

22 3 0 4 9 3 0 2 5 2 0 1

4 23 26 22 3 9 12 10 2 5 7 6

990 22 15.4 (4.4,34.9) 97.8 (96.7,98.6) 886 126 88.5 (69.8, 97.6) 87.5 (85.4, 89.5) 903 109 100.0 89.2 (87.2, 91.1) 1000 12 84.6 (65.1, 95.6) 98.8 (97.9, 99.4) 284 256 260 285 41 26 25 41 4 25.0 (5.5, 57.2) 32 75.0 (42.8, 94.5) 28 100.0 3 83.3 (51.6, 97.9) 2 28.6 (3.7, 71.0) 17 71.4 (29.0, 96.3) 18 100.0 2 85.7 (42.1, 99.6) 98.6 (96.5, 88.9 (84.7, 90.3 (86.3, 99.0 (97.0, 95.3 (84.2, 60.5 (44.4, 58.1 (42.1, 95.3 (84.2, 99.6) 92.3) 93.4) 99.8) 99.4) 75.0) 73.0) 99.4)

96.9) 15.4 (4.4, 34.9) 97.8 (96.7, 98.6) 89.5) 15.4 (10.0, 22.3) 99.7 (99.0, 99.9) 91.3) 19.3 (13.0, 26.9) 100.0 99,1) 64.7 (46.5, 80.3) 99.6 (99.0, 99.9) 97.7) 42.9 (9.9, 81.6) 96.9 (94.2, 98.6) 91.7) 22.0 (10.6, 37.6) 98.8 (96.7, 99.8) 93.7) 30.0 (16.6, 46.5) 100.0 99.5) 76.9 (46.2, 95.0) 99.3 (97.5, 99.9) 94.2) 50.0 (6.8, 93.2) 89.1 (76.4, 96.4) 75.3) 22.7 (7.8, 45.4) 92.9 (76.5, 99.1) 77.1) 28.0 (12.1, 49.4) 100.0 98.7) 75.0 (34.9, 96.8) 97.6 (87.4, 99.9)

FN, false negative; TP, true positive; TN, true negative; FP, false positive; CI, condence interval; Sn, sensitivity; Sp, specicity; Acc, accuracy; PPV, positive predictive value; NPV, negative predictive value; MRI, conventional magnetic resonance imaging criteria; aADC, absolute apparent diffusion coefcient criteria; rADC, relative apparent diffusion coefcient difference between tumor and lymph node; cDWI, combined size and rADC criteria.

0.532, 0.804) criteria. It follows that cDWI yielded an improved diagnostic performance when compared to the other criteria (P 0.124 for rADC, P 0.017 for aADC, and P 0.015 for conventional MR). Using the cDWI criterion, the detection of metastatic lymph node was possible for short-axis diameters as small as 5 mm. False-negative lymph nodes with this method were retrospectively noted in two regions of cervical cancer, including one adenocarcinoma (rADC

0.02 10 3 mm2/s, short-axis 4 mm, long-axis 8 mm, short-to-long axis ratio 0.5) and one adenosquamous carcinoma (rADC 0.04 10 3 mm2/s, short-axis 3 mm, long-axis 4 mm, short-to-long ratio 0.75). Falsepositive results of cDWI criterion occurred in two regions of cervical cancer and in another region of endometrial cancer. Histopathology disclosed adenocarcinoma in all these cases. Interestingly, metastatic lymph nodes in all three regions of carcinosarcoma were correctly identied by the cDWI criterion. Regions containing metastatic lymph nodes were identied by DWI (b 1000), but the single lymph nodes could not be visualized in all occasions. Twelve of the 26 malignant lymph nodes were identied in 12 nodal regions. The remaining 14 malignant lymph nodes could not be identied in retrospect on both DWI and MRI images. Overall, only two regions were overlooked on a per-region basis. On a pathologic basis, the nodes that remained unidentied by MRI were less than 2 mm in short-axis diameter, with microscopic metastatic foci from adenocarcinoma and adenosquamous carcinoma of the cervix, uterine carcinosarcoma, and endometrial adenocarcinoma, respectively. DISCUSSION Our results demonstrate the clinical usefulness of DWI for detecting nodal metastasis in patients with gynecologic cancer. A progressive renement of diagnostic accuracy was observed from conventional MR to aADC, rADC, and cDWI criteria. The use of cDWI criteria resulted in a signicant increase in sensitivity, without loss of specicity compared to conventional MR imaging. To our knowledge, only one study has investigated the potential usefulness of DWI for the noninvasive detection of malignant lymph nodes, for patients with head and neck cancer (13). In that report, the mean ADC of metastatic lymph nodes was signicantly lower

Figure 2. Graph shows that the ROC curve for cDWI is superior to rADC, aADC, and MRI criteria for detecting nodal metastases on a patient basis. MRI, conventional MR image; aADC, absolute apparent diffusion coefcient (ADC) value; rADC, relative ADC value between primary tumor and lymph node; cDWI, combination of size and rADC criteria; AUC, area under the curve.

LN Metastasis in Cervix/Uterus CA: 3T DWI

133

Figure 3. MR images of a 52-year-old woman with uterine carcinosarcoma. From left to right, a d. a: Axial T2-weighted image (T2WI). b: Axial diffusion-weighted image (DWI) in b 1000. c: Fusion of T2WI and DWI. A lymph node was noted in the left external iliac region (arrow), 5-mm in short axis, 7-mm in long axis. The absolute apparent diffusion coefcient (ADC) value of lymph node was 0.79 10 3 mm2/s, and ADC value of uterine tumor was 0.86 10 3 mm2/s, therefore the relative ADC value was 0.07 10 3 mm2/s. By using DWI, lymph node metastasis was predicted. d: Histopathologic examination (hematoxylin and eosin, magnication, 400) showed a tumor deposit within the node.

than that of benign cervical lymph nodes (cutoff value of ADC 1.38 10 3 mm2/s, P 0.04). Notably, this method was successful for detection of lymph nodes as small as 9 mm in size, with an accuracy of 96%, sensitivity of 98%, specicity of 88%, positive predictive value (PPV) of 98.5%, and negative predictive value (NPV) of 83.7%. However, in our study the ADC value of the metastatic lymph nodes was higher than that of benign nodes, albeit not signicantly. In the aADC criteria, we used the ADC value as the sole determinant factor. Although the sensitivity was higher than conventional MR, its specicity was decreased. A low PPV might raise issues in clinical judgment. We thus propose the use of the rADC criterion. For the development of the rADC criterion, we assumed that regional lymph nodes invaded by tumor cells would display similar cellularity and/or microarchitecture, in a way similar to the primary tumor. If this had been the case, the ADC value in the malignant lymph nodes would be similar to that of the primary tumor. The rADC criterion eventually resulted in better sensitivity. However, our results showed that apparently elongated lymph nodes could be overestimated by

the rADC criterion. This may lead to a loss of specicity when compared with conventional measures using size only. Hence, the cDWI criterion was proposed by using a combination of four measurable DWI parameters: ADC of primary tumor, ADC of lymph node, and short- and long-axis diameters of lymph nodes. With the use of the combined criterion, detection of lymph node metastasis could be as small as 5 mm in the short axis. We also showed that this criterion might overcome the reduced specicity of aADC or rADC criteria. Rockall et al (6) used a new lymph node-specic agent, ferumoxtran-10, to improve the sensitivity of conventional MR for preoperative prediction of nodal metastasis. Their results are in keeping with our ndings. Notably, their study population was similar to that described in the present study with regard to the population setting, sample size, number of sampled lymph nodes, as well as histological characteristics. In their study the sensitivity, specicity, PPV, and NPV were very high, namely 93%, 97%, 59%, and 100% on a node basis and 100%, 94%, 82%, and 100% on a patient basis. A similar diagnostic accuracy was achieved by

Figure 4. MR images of a 47-year-old woman with cervical adenocarcinoma. From left to right, a d. a: Axial T2-weighted image (T2WI). b: Axial diffusion-weighted image (DWI) in b 1000. c: Fusion of T2WI and DWI. A lymph node was noted in the left external iliac region (arrow), 10-mm in short axis, 11-mm in long axis. The absolute apparent diffusion coefcient (ADC) value of lymph node was 0.63 10 3 mm2/s, and ADC value of cervical tumor was 1.03 10 3 mm2/s, therefore the relative ADC value was 0.40 10 3 mm2/s. By using DWI, reactive lymph node was more favored than nodal metastasis. d: Histopathologic examination (hematoxylin and eosin, magnication, 400) showed only lymphocytes inltrate but without evidence of tumor metastasis.

134

Lin et al.

cDWI criteria in our study, using a combined evaluation of relative ADC values and size criteria with the use of DWI. The venous plexus makes it difcult to identify pelvic lymph nodes in conventional MR. DWI may improve their detection by homogenous background suppression. In our study the majority of lesions detected using a high b-value (b 1000 s/mm2) were primary tumors and lymph nodes, thereby facilitating lesion identication. However, the absence of anatomical reference can hamper standardization of DWI interpretation in the clinical setting. In the current study this shortcoming was overcome by image fusion. The DWI was fused with the T2-weighted image, thereby providing a good estimation of the anatomy. Interpreters can detect pelvic lymph nodes much more effortlessly on the fused images, as can be seen in Figures 3 and 4. By using the combined criterion, there were two falsenegative-diagnosed regions, one with microscopic deposits of adenocarcinoma and the other with adenosquamous carcinoma. Three false-positive-diagnosed regions all occurred in patients with adenocarcinoma. Certain overlapping in cellular density among cancer cells and normal lymphatic structures may have resulted in misclassication. Due to the limitations in the spatial resolution of MR systems, we argue that microscopic metastases would still remain a challenge, particularly for patients with adenocarcinoma or adenosquamous carcinoma. In carcinosarcoma patients a higher degree of condence could be reached inasmuch as all its three metastatic regions were correctly identied by using combined criteria. Our study showed that DWI in gynecological study is feasible in 3T MRI. Respiratory gating is not required and the complete acquisition can be as short as within 5 minutes. Although increased susceptibility artifact to local eld inhomogeneity in near air areas has limited the utility of DWI in 3T MRI (14), our study explored its value in cervical and uterine cancers since deep locations of pelvic lymph nodes are not affected by this artifact. Hence, detection of nodal metastases in gynecologic cancer using DWI can be a straightforward diagnostic approach. Accordingly, tumor and lymph nodes are highlighted, and background signal is homogeneously suppressed. Since gadolinium-based contrast medium is not necessary in DWI imaging, this technique may be useful in patients with moderate to end-stage renal disease in order to prevent the development of nephrogenic systemic brosis (15,16). It is thus posited that DWI may improve the surgical and nonsurgical management of patients with cervical and uterine cancers in several ways. The high sensitivity of rADC can serve as a map for presurgical planning, thereby avoiding unnecessary extensive lymphadenectomy, especially in patients at high surgical or anesthesiological risk. The eld used in intensity-modulated radiotherapy could be condently simulated according to the extent of nodal involvement as dened by the combined criterion. Some limitations of the present report merit consideration. The major limitation of our report relies in the

denition of the cutoff criteria for the ADC. Accordingly, they were determined by applying a post-hoc analysis of our data. This might have lead to an inated accuracy of DWI. Second, patients with advanced disease were not included in the present study; therefore, only 7 out of 50 patients had nodal metastases. The high specicity obtained in our study should be interpreted cautiously given the relatively small number of metastatic nodes examined. Hence, given the small sample size no rm conclusion could be reached on the sensitivity and specicity of DWI. Third, the in-plane resolution in our study was 1.6 1.6 4 mm, suggesting that spatial resolution remains a limitation of MR. Fourth, no patient with squamous cell carcinomas was found to have pathologic evidence of lymph node metastases. Thus, no conclusion can be directly inferred for patients with this specic type of malignancy. Finally, in our study the analysis was conned within the pelvis. Further investigations will focus on the assessment of DWI for detection of para-aortic node metastasis or other extrapelvic lesions. In conclusion, a combined evaluation of relative ADC values and size criteria using DWI improves the preoperative characterization of lymph node metastases compared with conventional MR. DWI may provide important information in planning the optimal surgical or radiotherapy treatment. ACKNOWLEDGMENT We thank Lee Chiu, MD, for his inspiration of DWI in oncology. REFERENCES
1. Lutman CV, Havrilesky LJ, Cragun JM, et al. Pelvic lymph node count is an important prognostic variable for FIGO stage I and II endometrial carcinoma with high-risk histology. Gynecol Oncol 2006;102:9297. 2. Kodama J, Seki N, Nakamura K, Hongo A, Hiramatsu Y. Prognostic factors in pathologic parametrium-positive patients with stage IBIIB cervical cancer treated by radical surgery and adjuvant therapy. Gynecol Oncol 2007;105:757761. 3. Yu KK, Hricak H, Subak LL, Zaloudek CJ, Powell CB. Preoperative staging of cervical carcinoma: phased array coil fast spin-echo versus body coil spin-echo T2-weighted MR imaging. AJR Am J Roentgenol 1998;171:707711. 4. Reinhardt MJ, Ehritt-Braun C, Vogelgesang D, et al. Metastatic lymph nodes in patients with cervical cancer: detection with MR imaging and FDG PET. Radiology 2001;218:776 782. 5. Bipat S, Glas AS, van der Velden J, Zwinderman AH, Bossuyt PM, Stoker J. Computed tomography and magnetic resonance imaging in staging of uterine cervical carcinoma: a systematic review. Gynecol Oncol 2003;91:59 66. 6. Rockall AG, Sohaib SA, Harisinghani MG, et al. Diagnostic performance of nanoparticle-enhanced magnetic resonance imaging in the diagnosis of lymph node metastases in patients with endometrial and cervical cancer. J Clin Oncol 2005;23:28132821. 7. Choi HJ, Kim SH, Seo SS, et al. MRI for pretreatment lymph node staging in uterine cervical cancer. AJR Am J Roentgenol 2006;187: W538 543. 8. Rowley HA, Grant PE, Roberts TP. Diffusion MR imaging. Theory and applications. Neuroimaging Clin N Am 1999;9:343361. 9. Ross BD, Moffat BA, Lawrence TS, et al. Evaluation of cancer therapy using diffusion magnetic resonance imaging. Mol Cancer Ther 2003;2:581587. 10. Charles-Edwards EM, deSouza NM. Diffusion-weighted magnetic resonance imaging and its application to cancer. Cancer Imaging 2006;6:135143.

LN Metastasis in Cervix/Uterus CA: 3T DWI

11. Colagrande S, Carbone SF, Carusi LM, Cova M, Villari N. Magnetic resonance diffusion-weighted imaging: extraneurological applications. Radiol Med (Torino) 2006;111:392 419. 12. Xing D, Papadakis NG, Huang CL, Lee VM, Carpenter TA, Hall LD. Optimised diffusion-weighting for measurement of apparent diffusion coefcient (ADC) in human brain. Magn Reson Imaging 1997; 15:771784. 13. Abdel Razek AA, Soliman NY, Elkhamary S, Alsharaway MK, Tawk A. Role of diffusion-weighted MR imaging in cervical lymphadenopathy. Eur Radiol 2006;16:1468 1477.

135
14. Naganawa S, Kawai H, Fukatsu H, et al. Diffusion-weighted imaging of the liver: technical challenges and prospects for the future. Magn Reson Med Sci 2005;4:175186. 15. Pedersen M. Safety update on the possible causal relationship between gadolinium-containing MRI agents and nephrogenic systemic brosis. J Magn Reson Imaging 2007;25:881 883. 16. Khurana A, Runge VM, Narayanan M, Greene JF Jr, Nickel AE. Nephrogenic systemic brosis: a review of 6 cases temporally related to gadodiamide injection (omniscan). Invest Radiol 2007;42: 139 145.