Ann Surg Oncol DOI 10.1245/s10434-015-4621-1
ORIGINAL ARTICLE – GASTROINTESTINAL ONCOLOGY
Usefulness of Two-Dimensional Values Measured Using Preoperative Multidetector Computed Tomography in Predicting Lymph Node Metastasis of Gastric Cancer Su Lim Lee, M.D.1, Han Hong Lee, M.D.2,3, Young Mi Ku, M.D.1, and Hae Myung Jeon, M.D.2 Department of Radiology, College of Medicine, The Catholic University of Korea, Seoul, Korea; 2Division of Gastrointestinal Surgery, Department of Surgery, College of Medicine, The Catholic University of Korea, Seoul, Korea; 3 Department of Surgery, Uijeongbu St. Mary’s Hospital, College of Medicine, The Catholic University of Korea, Uijeongbu-si, Gyenggi-Do, Korea 1
ABSTRACT Background. Multidetector computed tomography (MDCT) is essential for the prediction of lymph node (LN) metastasis in gastric cancer. However, the measurement method and size criteria for metastatic LNs using MDCT are unclear. Methods. MDCTs of gastric cancer patients who underwent surgery and had pathological staging were reviewed by radiologists. The two-dimensional cutoff values for LNs with suspected metastasis were calculated, and clinicopathological data were analyzed using those cutoff values. Results. The total number of enrolled patients was 327. The cutoff values of the maximal area with metastatic LNs were obtained significantly at stations 3, 4, and 6, and those values were 112.09, 33.79, and 85.88 mm2, respectively. The common cutoff value was 112.09 mm2, and the area under the curve was 0.617 (P = 0.002). The overall survival rate of the patients with LNs less than 112.09 mm2 was significantly better than those with LNs greater than 112.09 mm2 (P \ 0.001). In multivariate analysis, the maximal LN area was an independent prognostic factor (adjusted hazard ratio, 1.697 [95% confidence interval 1.116–2.582]).
Electronic supplementary material The online version of this article (doi:10.1245/s10434-015-4621-1) contains supplementary material, which is available to authorized users. Ó Society of Surgical Oncology 2015 First Received: 17 January 2015 H. H. Lee, M.D. e-mail: [email protected]
Conclusions. Using two-dimensional values for LNs measured by MDCT is a practical method of predicting metastatic LNs in gastric cancer. The maximal LN area value would be useful in both the preoperative staging and prognosis prediction of gastric cancer.
Continuing advancements in technology with improved resolution and the use of multiplanar reformatted images have made multidetector computed tomography (MDCT) a widely used diagnostic method for gastric cancer staging with high sensitivity.1,2 MDCT plays a crucial role in preoperative staging, which determines the treatment strategy for gastric cancer.3–6 The significance of MDCT is more emphasized in the N and M staging rather than in the T staging.7–11 Although there are various diagnostic criteria for lymph node (LN) metastasis using MDCT, such as enhancement, clustering, and shape, LN size is considered the most important.9,12,13 A short diameter greater than 10 mm is generally considered a metastatic LN with tumor infiltration.14–16 However, the determination of metastatic LNs according to the short diameter certainly has limitations in recognizing LN metastasis because no universal and appropriate cutoff value exists for nodal size to predict the pathological nodal status.16,17 Another major criticism of the uni-dimensional diameter measurement is that it cannot be exactly correlated with the entire shape and volume of the LNs. Several previous studies have reported the volume measurement of LN to be correlated with the accurate LN size.18 However, three-dimensional measurement of LNs shows no significant difference in the evaluation of the chemotherapy response.19 In addition, the latter method is more time-consuming and subjective. Therefore, modified
S. L. Lee et al.
and specialized criteria considering the advantages of the two measurements were created to address these problems. The maximal LN area as a two-dimensional value is defined by measuring the area of the largest LN visualized in the transverse or coronal reformatted images of MDCT. Thus, the purpose of the present study was to evaluate the maximal LN area that can predict LN metastasis on preoperative MDCT for gastric cancer; that area was assessed in each LN station. In addition, the correlation between the maximal LN area on MDCT and clinical outcome was investigated to demonstrate the usefulness of the two-dimensional value. METHODS
Solutions, Forchheim, Germany) after injection of intravenous contrast medium (Iopromide; Ultravist 300; Bayer Healthcare, Berlin, Germany). The images were acquired during the portal venous phase (70-s delay after contrast injection; 100 kVp; 220 mA; pitch, 1.25; gantry speed, 0.5 s per rotation) and included the region from the dome of the diaphragm to the symphysis pubis. Tube current modulation software (CareDose 4D; Siemens Medical Solutions) was used. The technicians used a computer software package (Syngo VB28B; Siemens Medical Solutions) to reconstruct the axial and coronal images with a section thickness and interval of 4 mm. The reconstructions were performed on a commercially available console system (Syngo; Wizard; Siemens Medical Solutions).
Patients
Image Analysis
Among the 407 patients diagnosed with gastric cancer and who underwent surgery at Uijeongbu St. Mary’s Hospital between 2005 and 2009, those who underwent preoperative staging as determined by MDCT and gastrectomy combined with LN dissection were included in the present study. Of those, after exclusion of patients who received neoadjuvant chemotherapy or had other organ malignancies, 327 patients with a pathological report, including LN metastasis at each station were enrolled in the present study. Pathological and clinical data, including operative details of all of the enrolled patients were collected retrospectively from the gastric cancer registry of Uijeongbu St. Mary’s Hospital. The pathological stage was classified according to the Seventh American Joint Cancer Committee (AJCC) TNM classification. Regular follow-up programs were conducted using tumor markers, abdominal computed tomography (CT), and endoscopic examination according to our standard protocol (every 3 and 6 months for advanced and early gastric cancer, respectively, for the first 3 years; every 12 months thereafter). The mean follow-up period was 47.5 ± 28.0 (range 0–108) months. The survival results were repeatedly identified using registration data from the Korea National Statistical Office and patients’ medical records. This study was approved by the Institutional Review Board of the Ethical Committee of the College of Medicine, Catholic University of Korea (UC14RISI0019), and patients’ records were anonymized and de-identified prior to analysis.
The presence and location of the largest LN were determined by consensus by two radiologists (S. L. L. and Y. M. K., with 6 and 10 years of clinical experience in abdominal radiology, respectively) according to the Japanese classification, using the picture archiving and communication system with dedicated software (M-view; Marotec Medical System, Seoul, Korea). The two observers determined the presence and location of the largest LN according to the Japanese classification by consensus using maximally magnified transverse and coronal reformatted images. Each observer—who was blinded to the surgical or histopathological results—separately measured an area of the largest LN by manually tracing the LN boundary using an electronic caliper (Fig. 1). Next, one of the two observers (S. L. L.) measured the area of the largest LN using the same methods as in the additional session. The measurement was performed three times for each LN, and the three measured values were averaged to minimize the measurement error. In addition, short-axis diameter of the largest LN was measured to conduct a comparative analysis between uni- and two-dimensional values.
MDCT Technique The CT examinations were performed using a 128-detector-row CT scanner (definition AS?; Siemens Medical
Statistical Analysis Differences between groups were analyzed using Student’s t test for continuous variables and v2 test or Fisher’s exact test for proportions. The cutoff values were calculated using the receiver operating characteristic (ROC) curve. Survival analysis was performed using the Kaplan– Meier method with the log-rank test for univariate analyses; multivariate analysis for survival was performed using a Cox proportional hazards model. Statistical analyses were performed using SAS (Statistical Analysis System, SAS institute, Cary, NC) version 9.3, and P \ 0.05 was deemed to indicate statistical significance.
Two-Dimensional Values of Metastatic Lymph Nodes FIG. 1 CT measurements of the maximal area of the largest lymph node. a The maximal area (560.58 mm2) of the largest lymph node in station 3 on transverse CT and b maximal area (262.51 mm2) of the largest lymph node in station 3 on coronal reformatted images. The lesion boundary was drawn manually by a radiologist
RESULTS
Comparison Between Groups by Cutoff Value
The mean age of the enrolled patients, who comprised 216 males and 111 females, was 61.5 years. Greater than two-thirds of the patients received subtotal gastrectomy, and approximately 90% received LN dissection of D2 and over. The most common macroscopic type of early gastric cancer and advanced gastric cancer was IIc and Borrmann III, respectively, and the most common histological type was poorly differentiated tubular adenocarcinoma. The pathological stage of the enrolled patients comprised 182 stage I (55.7%), 47 stage II (14.4%), 85 stage III (26.0%), and 13 stage IV (4.0%) lesions based on the seventh AJCC classification system. One hundred thirty-four (41.0%) patients had LN metastasis, and 157 (48.0%) patients received adjuvant chemotherapy (Supplementary Table 1).
According to the common cutoff value, the enrolled patients were divided into two groups: the small area LN group and large area LN group. A total of 275 and 52 patients were included in the small and large area LN groups, respectively. There were no differences regarding the operative details between the two groups. However, the large area LN group comprised patients with larger and more advanced tumors; consequently, the proportion of patients who received noncurative surgery and adjuvant chemotherapy was higher in the large LN group (Supplementary Table 2). The patients with small area LN had a better overall survival rate than those with large area LN (P \ 0.001, Supplementary Fig. 1).
Cutoff Value of the Maximal LN Area to Predict Metastasis The LN area predicting pathological LN metastasis was calculated using an ROC curve for each LN station. Based on the value of each LN station, a common cutoff value for the maximal LN area that was considered a metastasis was also calculated. Among the first-, second-, third-tier and distant LN station, including stations 14 and 16, the cutoff values were obtained at stations 3, 4, 6, and 7. The LN values of stations 3, 4, and 6 were 112.09, 33.79, and 85.88 mm2, respectively, with statistical significance. However, the value of station 7 had no statistical significance (Table 1). The common cutoff value was 112.09 mm2, which was statistically significant (Table 1), and the area under the curve (AUC) was 0.617 (Fig. 2A). The AUCs of LN stations 3, 4, and 6 are described in Fig. 2A. The LN short-axis diameter predicting pathological LN metastasis was also calculated by same manner (Table 2). In the analysis of short-axis diameter, LN station 6 and common cutoff values had statistical significance (Fig. 2B).
Maximal LN Area as a Prognostic Factor Multivariate survival analysis using the Cox proportional hazard regression model was conducted, including clinical TNM stage based on the MDCT. It showed that 65 years and older, more than clinical subserosal invasion, clinical N3 and M1, and maximal LN area were significant prognostic factors. After adjustment for clinically relevant factors, the maximal LN area remained an independent predictor of gastric cancer (adjusted hazard ratio, 1.697 [95% confidence interval 1.116–2.582]; P = 0.013; Table 3). DISCUSSION A significant direct correlation has been observed between uni-dimensional measurement (long-axis or shortaxis diameter) of the largest LN visualized on MDCT and metastatic LN status.12 Although it is a simple modality that has high reproducibility for predicting metastatic LN status, the one-dimensional parameter on the axial plane of MDCT could be influenced by the orientation of the LN within the stations.12,18 A previous study demonstrated that 49.3% of the positive dissected LN were round, and 50.7%
S. L. Lee et al. TABLE 1 Two-dimensional cutoff values for predicting metastasis in each lymph node station and total lymph node stations Area
n
Cutoff value (mm2)
No. of metastases
Sensitivity (%)
95% CI
Specificity (%)
95% CI
P value
Total
327
C112.09
90
31.11
21.8–41.7
90.3
85.8–93.7
0.002
Station 0
80
18
1
4
0
2
1
3
105
[112.09
38
36.84
21.8–54.0
92.54
83.4–97.5
0.003
4
34
[33.79
13
84.62
54.6–98.1
71.43
47.8–88.7
0.004
5
3
6
27
[85.88
11
54.55
23.4–83.3
93.75
69.8–99.8
0.033
7
40
[72.11
8
62.50
24.5–91.5
71.87
53.3–86.3
0.356
8 9
21 1
1 0
11
3
0
12
5
0
14
2
0
16
1
0
0
1
CI confidence interval
were oval.14 If the metastatic LN had an oval shape, the short axis measurement on the axial plane may indicate under- or overestimation according to the direction of the LN. Recently, volumetric assessment of LNs has been considered a better parameter to reflect shape and size alterations regardless of the orientation of the LN.18,19 However, manual volumetric measurements more time consuming and has higher interobserver variability during the assessment. To compensate for the shortcomings of the uni-dimensional and volumetric measurement systems, the maximal LN area, defined as the measured area of the largest LN visualized in the transverse or coronal reformatted images of MDCT, was devised in the present study. It was supposed that the maximal LN area might be more precisely correlated with the real size of LNs than a unidimensional measurement and also be less time-consuming than volume measurement. The highlight of the present study was that it represents the first report of the two-dimensional cutoff value of LN metastasis at each LN station in gastric cancer. In fact, because MDCT is originally two-dimensional picture, the two-dimensional measurement is most practical and efficient in the use of MDCT without additional reconstruction of image. In addition, in order to increase the concordance rate between area value and LN metastasis, LN area in the MDCT was measured based on each LN station and analyzed with pathological report which LN metastasis was recorded by each station. Although the cutoff value could not be calculated at all LN stations, the present study suggested concrete values for LN stations, such as stations
3, 4, and 6, with a high potential for metastasis. The common two-dimensional cutoff value of LN metastasis, measured based on the analysis of each LN station, was identical to that of LN station 3 because of the high proportion of station 3. Furthermore, the maximal LN area, two-dimensional value, showed a higher predictability for LN metastasis than shot-axis diameter, uni-dimensional value. Both methods yielded their own common cutoff values. However, among the cutoff values by each LN station using the short-axis diameter, that of only station 6 had a statistical significance. To verify the clinical implication of the two-dimensional cutoff value, enrolled patients were divided into two groups according to the common cutoff value, 112.09 mm2. The patients with a LN area greater than the common value were associated with a more advanced tumor stage, and a larger LN area was demonstrated to be a prognostic predictor of gastric cancer patients in a multivariate survival analysis. Various size criteria for metastatic LNs according to location in gastric cancer have been reported. One study used [6 mm as a criterion for metastatic LN in the perigastric region and [8 mm in the left gastric region.20 Another study reported that benign regional LNs C6 mm are more frequently detected in gastric cancer patients than in healthy populations, particularly around the lesser curvature.17 In the present study, the cutoff value for metastatic LNs of station 3 is greater than that of other stations. The exact mechanism underlying the larger LN size in station 3 is unclear. Large lymphatic streams around the lesser curvature area wound give rise to these
Two-Dimensional Values of Metastatic Lymph Nodes
A a
100
b
100
80
Sensitivity
80
Sensitivity
FIG. 2 A Receiver operating characteristic curve for a the common (area under the curve (AUC) = 0.617, P = 0.002), b LN station 3 (AUC = 0.668, P = 0.003), c LN station 4 (AUC = 0.762, P = 0.004), and d LN station 6 (AUC = 0.733, P = 0.033) cutoff values of the maximal LN area. B Receiver operating characteristic curve for a the common (area under the curve (AUC) = 0.581, P = 0.028) and b LN station 6 (AUC = 0.795, P = 0.004) cutoff values of the short diameter
60
40
20
60
40
station=3 AUC=0.668, P value=0.003
20
AUC=0.617, P value=0.002
0
0 20
0
40
60
80
0
100
20
c
100
60
80
100
d
100
80
Sensitivity
80
Sensitivity
40
100-Specificity
100-Specificity
60
40
station=4 AUC=0.762, P value=0.004
20
60
40
station=6 AUC=0.733, P value=0.033
20
0
0
0
20
40
60
80
0
100
20
40
60
80
100
100-Specificity
100-Specificity
B 100
100
a
80
Sensitivity
Sensitivity
80
b
60
40
20
60
40
20
station=6 AUC=0.795, P value=0.0037
AUC=0.581, P value=0.028 0
0 0
20
40
60
100-Specificity
results.21,22 Many LN metastases in station 3 compared with other stations were observed in the present study and two-dimensional LN measurement using MDCT was thought to reflect such consequence. Another possible mechanism is through cancer cell-induced inflammation by activating inflammatory mediators.16,17 The high prevalence of tumors in the lesser curvature (53.5%) might be related to the presence of larger LNs with cancer cell-induced inflammation in station 3.
80
100
0
20
40
60
80
100
100-Specificity
Generally, the two-dimensional value would cause a wider variation than the uni-dimensional value. Therefore, the area value was measured several times by two observers to minimize the variability. Besides calculating the cutoff value, to verify the measurement reliability, value differences from one observer and between two observers were calculated and analyzed using Bland–Altman plots with 95% confidence intervals (Fig. 3). The Bland–Altman plots indicated that intraobserver (Fig. 3a) and interobserver
S. L. Lee et al. TABLE 2 Uni–dimensional cutoff values for predicting metastasis in each lymph node station and total lymph node stations Short diameter
n
Cutoff value (mm)
No. of metastases
Total
327
[7.39
90
Sensitivity (%)
95% CI
Specificity (%)
95% CI
P value
41.11
30.8–52.0
77.22
71.3–82.4
0.028
Station 0
80
18
1
4
0
2
1
3
105
[7.49
38
50.00
33.4–66.6
73.13
60.9–83.2
0.145
4
34
[6.4
13
30.77
10.4–61.1
38.10
19.0–61.3
0.084
5
3
6
27
[8.68
11
72.73
39.0–94.0
93.75
69.8–99.8
0.004
7
40
[6.38
8
75.00
34.9–96.8
59.38
40.6–76.3
0.428
8 9
21 1
1 0
11
3
0
12
5
0
14
2
0
16
1
0
0
1
CI confidence interval
TABLE 3 Univariate and multivariate Cox proportional hazard regression survival analysis Alive
Dead
P value
HR (95% CI)
P value
Adjust HR (95% CI)
1.913 (1.347–2.717)
\0.001
2.044 (1.433–2.916)
1.470 (0.993–2.177)
0.054
P value
Age (year) \65
118 (59.3)
55 (43.0)
C65
81 (40.7)
73 (57.0)
122 (61.3)
94 (73.4)
77 (38.7)
34 (26.6)
T1
137 (68.8)
32 (25.0)
T2 T3
30 (15.1) 14 (7.0)
11 (8.6) 19 (14.8)
T4a
15 (7.5)
47 (36.7)
3.617 (2.039–6.419)
\0.001
T4b
3 (1.5)
19 (14.8)
7.247 (3.043–17.255)
\0.001
N0
138 (69.4)
42 (32.8)
N1
32 (16.1)
18 (14.1)
0.981 (0.522–1.845)
N2
18 (9.1)
20 (15.6)
1.840 (0.969–3.497)
0.063
N3
11 (5.5)
48 (37.5)
2.663 (1.503–4.716)
\0.001
M0
197 (99.0)
114 (89.1)
M1
2 (1.0)
14 (10.9)
Small
180 (90.5)
95 (74.2)
Large
19 (9.5)
33 (25.8)
0.004
1.000
1.000 \0.001
Gender Male Female
0.024
1.000
Clinical T \0.001
1.000 1.135 (0.559–2.306) 3.400 (1.812–6.377)
0.726 0.001
Clinical N \0.001
1.000 0.953
Clinical M \0.001
1.000
1.000
6.298 (3.567–11.119)
\0.001
6.128 (3.321–11.102)
\0.001
2.147 (1.444–3.194)
\0.001
1.697 (1.116–2.582)
0.013
LN area
HR hazard ratio
\0.001
1.000
1.000
a
20 0
–20
+1.96 SD 15.2 Mean
b
–1.0 –1.96 SD
Inter-observer difference
FIG. 3 Bland–Altman plots showing a intraobserver and b interobserver agreement
Intra-observer difference
Two-Dimensional Values of Metastatic Lymph Nodes 30 20
–17.2
–40 –60 –80 –100
+1.96 SD 9.1 Mean
10 0
0.2 –1.96 SD
–10
–8.7
–20 –30 –40
–120
–50 0
500
1000
1500
2000
0
Average value (mm 2)
(Fig. 3b) differences were not significant, and the measurement results were acceptable for the analysis. The present study possessed several limitations. Although the superiority of the two-dimensional cutoff value compared with the uni-dimensional value was demonstrated in the current study, the comparison of our results with three-dimensional values was not conducted and the AUC values that represent predictability were somewhat low. In addition to LN size, there are other predictable factors for LN metastasis, such as enhancement, clustering, and shape in the use of MDCT.9,13 Because the present study evaluated the adequate criteria of LN size for prediction of metastasis, the low AUC value would be caused by the study design which other predictable factors were not included in the analysis. However, the present outcomes might be adequate to verify the practicality and usefulness of two-dimensional measurement to predict LN metastasis in gastric cancer. Second, comparative and survival analyses according to the cutoff value of each LN station could not be performed because of the small number of metastatic LNs at each station and the small number of patients with single LN metastases; both were insufficient for a subgroup analysis. Third, the imaging-surgical-pathologic correlations of LNs were evaluated using station-by-station analysis, not lesion-by-lesion analysis. To our knowledge, this is the first report of a two-dimensional value, based on each station, to predict LN metastasis using MDCT for gastric cancer. In summary, the maximal LN area as the two-dimensional value represents a practical and reproducible method of identifying metastatic LNs preoperatively. Using this two-dimensional value, an individualized cutoff value could be established at each LN station in gastric cancer. ACKNOWLEDGMENT This study was supported by a grant from the National Research Foundation of Korea (No. 2012R1A1A1043576) and Fukuoka Foundation for Sound Health Cancer Research Fund. The statistical consultation was supported by the Catholic Research
500
1000
1500
2000
Average value (mm 2)
Coordinating Center of the Korea Health 21 R&D Project (A070001), Ministry of Health & Welfare, Republic of Korea. DISCLOSURE
The authors declare no conflict of interest.
REFERENCES 1. Stabile Ianora AA, Pedote P, Scardapane A, Memeo M, Rotondo A, Angelelli G. Preoperative staging of gastric carcinoma with multidetector spiral CT. Radiol Med. 2003;106:467–80. 2. Kwee RM, Kwee TC. Imaging in local staging of gastric cancer: a systematic review. J Clin Oncol. 2007;25:2107–16. 3. Kim AY, Kim HJ, Ha HK. Gastric cancer by multidetector row CT: preoperative staging. Abdom Imaging. 2005;30:465–72. 4. Chen CY, Wu DC, Kang WY, Hsu JS. Staging of gastric cancer with 16-channel MDCT. Abdom Imaging. 2006;31:514–20. 5. Matsuki M, Kani H, Tatsugami F, et al. Preoperative assessment of vascular anatomy around the stomach by 3D imaging using MDCT before laparoscopy-assisted gastrectomy. AJR Am J Roentgenol. 2004;183:145–51. 6. Pan Z, Zhang H, Yan C, et al. Determining gastric cancer resectability by dynamic MDCT. Eur Radiol. 2010;20:613–20. 7. Yan C, Zhu ZG, Yan M, et al. Value of multidetector-row computed tomography in the preoperative T and N staging of gastric carcinoma: a large-scale Chinese study. J Surg Oncol. 2009;100:205–14. 8. Hur J, Park MS, Lee JH, et al. Diagnostic accuracy of multidetector row computed tomography in T- and N staging of gastric cancer with histopathologic correlation. J Comput Assist Tomogr. 2006;30:372–7. 9. Kwee RM, Kwee TC. Imaging in assessing lymph node status in gastric cancer. Gastric Cancer. 2009;12:6–22. 10. Anzidei M, Napoli A, Zaccagna F, et al. Diagnostic performance of 64-MDCT and 1.5-T MRI with high-resolution sequences in the T staging of gastric cancer: a comparative analysis with histopathology. Radiol Med. 2009;114:1065–79. 11. Kim SH, Kim JJ, Lee JS, et al. Preoperative N staging of gastric cancer by stomach protocol computed tomography. J Gastric Cancer. 2013;13:149–56. 12. Yan C, Zhu ZG, Yan M, et al. Size of the largest lymph node visualized on multi-detector-row computed tomography (MDCT) is useful in predicting metastatic lymph node status of gastric cancer. J Int Med Res. 2010;38:22–33. 13. Lee MH, Choi D, Park MJ, Lee MW. Gastric cancer: imaging and staging with MDCT based on the 7th AJCC guidelines. Abdom Imaging. 2012;37:531–40.
S. L. Lee et al. 14. Morgagni P, Petrella E, Basile B, et al. Preoperative multidetector-row computed tomography scan staging for lymphatic gastric cancer spread. World J Surg Oncol. 2012;10:197. 15. Tokunaga M, Sugisawa N, Tanizawa Y, Bando E, Kawamura T, Terashima M. The impact of preoperative lymph node size on long-term outcome following curative gastrectomy for gastric cancer. Ann Surg Oncol. 2013;20:1598–603. 16. Zhang XP, Wang ZL, Tang L, Sun YS, Cao K, Gao Y. Support vector machine model for diagnosis of lymph node metastasis in gastric cancer with multidetector computed tomography: a preliminary study. BMC Cancer. 2011;11:10. 17. Park HS, Kim YJ, Ko SY, et al. Benign regional lymph nodes in gastric cancer on multidetector row CT. Acta Radiol. 2012;53:501–7. 18. Buerke B, Puesken M, Muter S, et al. Measurement accuracy and reproducibility of semiautomated metric and volumetric lymph node analysis in MDCT. AJR Am J Roentgenol. 2010;195:979–85.
19. Lee SM, Kim SH, Lee JM, et al. Usefulness of CT volumetry for primary gastric lesions in predicting pathologic response to neoadjuvant chemotherapy in advanced gastric cancer. Abdom Imaging. 2009;34:430–40. 20. Kim YN, Choi D, Kim SH, et al. Gastric cancer staging at isotropic MDCT including coronal and sagittal MPR images: endoscopically diagnosed early vs. advanced gastric cancer. Abdom Imaging. 2009;34:26–34. 21. Lee HH, Yoo HM, Song KY, Jeon HM, Park CH. Risk of limited lymph node dissection in patients with clinically early gastric cancer: indications of extended lymph node dissection for early gastric cancer. Ann Surg Oncol. 2013;20:3534–40. 22. Kinami S, Fujimura T, Ojima E, et al. PTD classification: proposal for a new classification of gastric cancer location based on physiological lymphatic flow. Int J Clin Oncol. 2008;13:320–9.
ORIGINAL ARTICLE – GASTROINTESTINAL ONCOLOGY
Usefulness of Two-Dimensional Values Measured Using Preoperative Multidetector Computed Tomography in Predicting Lymph Node Metastasis of Gastric Cancer Su Lim Lee, M.D.1, Han Hong Lee, M.D.2,3, Young Mi Ku, M.D.1, and Hae Myung Jeon, M.D.2 Department of Radiology, College of Medicine, The Catholic University of Korea, Seoul, Korea; 2Division of Gastrointestinal Surgery, Department of Surgery, College of Medicine, The Catholic University of Korea, Seoul, Korea; 3 Department of Surgery, Uijeongbu St. Mary’s Hospital, College of Medicine, The Catholic University of Korea, Uijeongbu-si, Gyenggi-Do, Korea 1
ABSTRACT Background. Multidetector computed tomography (MDCT) is essential for the prediction of lymph node (LN) metastasis in gastric cancer. However, the measurement method and size criteria for metastatic LNs using MDCT are unclear. Methods. MDCTs of gastric cancer patients who underwent surgery and had pathological staging were reviewed by radiologists. The two-dimensional cutoff values for LNs with suspected metastasis were calculated, and clinicopathological data were analyzed using those cutoff values. Results. The total number of enrolled patients was 327. The cutoff values of the maximal area with metastatic LNs were obtained significantly at stations 3, 4, and 6, and those values were 112.09, 33.79, and 85.88 mm2, respectively. The common cutoff value was 112.09 mm2, and the area under the curve was 0.617 (P = 0.002). The overall survival rate of the patients with LNs less than 112.09 mm2 was significantly better than those with LNs greater than 112.09 mm2 (P \ 0.001). In multivariate analysis, the maximal LN area was an independent prognostic factor (adjusted hazard ratio, 1.697 [95% confidence interval 1.116–2.582]).
Electronic supplementary material The online version of this article (doi:10.1245/s10434-015-4621-1) contains supplementary material, which is available to authorized users. Ó Society of Surgical Oncology 2015 First Received: 17 January 2015 H. H. Lee, M.D. e-mail: [email protected]
Conclusions. Using two-dimensional values for LNs measured by MDCT is a practical method of predicting metastatic LNs in gastric cancer. The maximal LN area value would be useful in both the preoperative staging and prognosis prediction of gastric cancer.
Continuing advancements in technology with improved resolution and the use of multiplanar reformatted images have made multidetector computed tomography (MDCT) a widely used diagnostic method for gastric cancer staging with high sensitivity.1,2 MDCT plays a crucial role in preoperative staging, which determines the treatment strategy for gastric cancer.3–6 The significance of MDCT is more emphasized in the N and M staging rather than in the T staging.7–11 Although there are various diagnostic criteria for lymph node (LN) metastasis using MDCT, such as enhancement, clustering, and shape, LN size is considered the most important.9,12,13 A short diameter greater than 10 mm is generally considered a metastatic LN with tumor infiltration.14–16 However, the determination of metastatic LNs according to the short diameter certainly has limitations in recognizing LN metastasis because no universal and appropriate cutoff value exists for nodal size to predict the pathological nodal status.16,17 Another major criticism of the uni-dimensional diameter measurement is that it cannot be exactly correlated with the entire shape and volume of the LNs. Several previous studies have reported the volume measurement of LN to be correlated with the accurate LN size.18 However, three-dimensional measurement of LNs shows no significant difference in the evaluation of the chemotherapy response.19 In addition, the latter method is more time-consuming and subjective. Therefore, modified
S. L. Lee et al.
and specialized criteria considering the advantages of the two measurements were created to address these problems. The maximal LN area as a two-dimensional value is defined by measuring the area of the largest LN visualized in the transverse or coronal reformatted images of MDCT. Thus, the purpose of the present study was to evaluate the maximal LN area that can predict LN metastasis on preoperative MDCT for gastric cancer; that area was assessed in each LN station. In addition, the correlation between the maximal LN area on MDCT and clinical outcome was investigated to demonstrate the usefulness of the two-dimensional value. METHODS
Solutions, Forchheim, Germany) after injection of intravenous contrast medium (Iopromide; Ultravist 300; Bayer Healthcare, Berlin, Germany). The images were acquired during the portal venous phase (70-s delay after contrast injection; 100 kVp; 220 mA; pitch, 1.25; gantry speed, 0.5 s per rotation) and included the region from the dome of the diaphragm to the symphysis pubis. Tube current modulation software (CareDose 4D; Siemens Medical Solutions) was used. The technicians used a computer software package (Syngo VB28B; Siemens Medical Solutions) to reconstruct the axial and coronal images with a section thickness and interval of 4 mm. The reconstructions were performed on a commercially available console system (Syngo; Wizard; Siemens Medical Solutions).
Patients
Image Analysis
Among the 407 patients diagnosed with gastric cancer and who underwent surgery at Uijeongbu St. Mary’s Hospital between 2005 and 2009, those who underwent preoperative staging as determined by MDCT and gastrectomy combined with LN dissection were included in the present study. Of those, after exclusion of patients who received neoadjuvant chemotherapy or had other organ malignancies, 327 patients with a pathological report, including LN metastasis at each station were enrolled in the present study. Pathological and clinical data, including operative details of all of the enrolled patients were collected retrospectively from the gastric cancer registry of Uijeongbu St. Mary’s Hospital. The pathological stage was classified according to the Seventh American Joint Cancer Committee (AJCC) TNM classification. Regular follow-up programs were conducted using tumor markers, abdominal computed tomography (CT), and endoscopic examination according to our standard protocol (every 3 and 6 months for advanced and early gastric cancer, respectively, for the first 3 years; every 12 months thereafter). The mean follow-up period was 47.5 ± 28.0 (range 0–108) months. The survival results were repeatedly identified using registration data from the Korea National Statistical Office and patients’ medical records. This study was approved by the Institutional Review Board of the Ethical Committee of the College of Medicine, Catholic University of Korea (UC14RISI0019), and patients’ records were anonymized and de-identified prior to analysis.
The presence and location of the largest LN were determined by consensus by two radiologists (S. L. L. and Y. M. K., with 6 and 10 years of clinical experience in abdominal radiology, respectively) according to the Japanese classification, using the picture archiving and communication system with dedicated software (M-view; Marotec Medical System, Seoul, Korea). The two observers determined the presence and location of the largest LN according to the Japanese classification by consensus using maximally magnified transverse and coronal reformatted images. Each observer—who was blinded to the surgical or histopathological results—separately measured an area of the largest LN by manually tracing the LN boundary using an electronic caliper (Fig. 1). Next, one of the two observers (S. L. L.) measured the area of the largest LN using the same methods as in the additional session. The measurement was performed three times for each LN, and the three measured values were averaged to minimize the measurement error. In addition, short-axis diameter of the largest LN was measured to conduct a comparative analysis between uni- and two-dimensional values.
MDCT Technique The CT examinations were performed using a 128-detector-row CT scanner (definition AS?; Siemens Medical
Statistical Analysis Differences between groups were analyzed using Student’s t test for continuous variables and v2 test or Fisher’s exact test for proportions. The cutoff values were calculated using the receiver operating characteristic (ROC) curve. Survival analysis was performed using the Kaplan– Meier method with the log-rank test for univariate analyses; multivariate analysis for survival was performed using a Cox proportional hazards model. Statistical analyses were performed using SAS (Statistical Analysis System, SAS institute, Cary, NC) version 9.3, and P \ 0.05 was deemed to indicate statistical significance.
Two-Dimensional Values of Metastatic Lymph Nodes FIG. 1 CT measurements of the maximal area of the largest lymph node. a The maximal area (560.58 mm2) of the largest lymph node in station 3 on transverse CT and b maximal area (262.51 mm2) of the largest lymph node in station 3 on coronal reformatted images. The lesion boundary was drawn manually by a radiologist
RESULTS
Comparison Between Groups by Cutoff Value
The mean age of the enrolled patients, who comprised 216 males and 111 females, was 61.5 years. Greater than two-thirds of the patients received subtotal gastrectomy, and approximately 90% received LN dissection of D2 and over. The most common macroscopic type of early gastric cancer and advanced gastric cancer was IIc and Borrmann III, respectively, and the most common histological type was poorly differentiated tubular adenocarcinoma. The pathological stage of the enrolled patients comprised 182 stage I (55.7%), 47 stage II (14.4%), 85 stage III (26.0%), and 13 stage IV (4.0%) lesions based on the seventh AJCC classification system. One hundred thirty-four (41.0%) patients had LN metastasis, and 157 (48.0%) patients received adjuvant chemotherapy (Supplementary Table 1).
According to the common cutoff value, the enrolled patients were divided into two groups: the small area LN group and large area LN group. A total of 275 and 52 patients were included in the small and large area LN groups, respectively. There were no differences regarding the operative details between the two groups. However, the large area LN group comprised patients with larger and more advanced tumors; consequently, the proportion of patients who received noncurative surgery and adjuvant chemotherapy was higher in the large LN group (Supplementary Table 2). The patients with small area LN had a better overall survival rate than those with large area LN (P \ 0.001, Supplementary Fig. 1).
Cutoff Value of the Maximal LN Area to Predict Metastasis The LN area predicting pathological LN metastasis was calculated using an ROC curve for each LN station. Based on the value of each LN station, a common cutoff value for the maximal LN area that was considered a metastasis was also calculated. Among the first-, second-, third-tier and distant LN station, including stations 14 and 16, the cutoff values were obtained at stations 3, 4, 6, and 7. The LN values of stations 3, 4, and 6 were 112.09, 33.79, and 85.88 mm2, respectively, with statistical significance. However, the value of station 7 had no statistical significance (Table 1). The common cutoff value was 112.09 mm2, which was statistically significant (Table 1), and the area under the curve (AUC) was 0.617 (Fig. 2A). The AUCs of LN stations 3, 4, and 6 are described in Fig. 2A. The LN short-axis diameter predicting pathological LN metastasis was also calculated by same manner (Table 2). In the analysis of short-axis diameter, LN station 6 and common cutoff values had statistical significance (Fig. 2B).
Maximal LN Area as a Prognostic Factor Multivariate survival analysis using the Cox proportional hazard regression model was conducted, including clinical TNM stage based on the MDCT. It showed that 65 years and older, more than clinical subserosal invasion, clinical N3 and M1, and maximal LN area were significant prognostic factors. After adjustment for clinically relevant factors, the maximal LN area remained an independent predictor of gastric cancer (adjusted hazard ratio, 1.697 [95% confidence interval 1.116–2.582]; P = 0.013; Table 3). DISCUSSION A significant direct correlation has been observed between uni-dimensional measurement (long-axis or shortaxis diameter) of the largest LN visualized on MDCT and metastatic LN status.12 Although it is a simple modality that has high reproducibility for predicting metastatic LN status, the one-dimensional parameter on the axial plane of MDCT could be influenced by the orientation of the LN within the stations.12,18 A previous study demonstrated that 49.3% of the positive dissected LN were round, and 50.7%
S. L. Lee et al. TABLE 1 Two-dimensional cutoff values for predicting metastasis in each lymph node station and total lymph node stations Area
n
Cutoff value (mm2)
No. of metastases
Sensitivity (%)
95% CI
Specificity (%)
95% CI
P value
Total
327
C112.09
90
31.11
21.8–41.7
90.3
85.8–93.7
0.002
Station 0
80
18
1
4
0
2
1
3
105
[112.09
38
36.84
21.8–54.0
92.54
83.4–97.5
0.003
4
34
[33.79
13
84.62
54.6–98.1
71.43
47.8–88.7
0.004
5
3
6
27
[85.88
11
54.55
23.4–83.3
93.75
69.8–99.8
0.033
7
40
[72.11
8
62.50
24.5–91.5
71.87
53.3–86.3
0.356
8 9
21 1
1 0
11
3
0
12
5
0
14
2
0
16
1
0
0
1
CI confidence interval
were oval.14 If the metastatic LN had an oval shape, the short axis measurement on the axial plane may indicate under- or overestimation according to the direction of the LN. Recently, volumetric assessment of LNs has been considered a better parameter to reflect shape and size alterations regardless of the orientation of the LN.18,19 However, manual volumetric measurements more time consuming and has higher interobserver variability during the assessment. To compensate for the shortcomings of the uni-dimensional and volumetric measurement systems, the maximal LN area, defined as the measured area of the largest LN visualized in the transverse or coronal reformatted images of MDCT, was devised in the present study. It was supposed that the maximal LN area might be more precisely correlated with the real size of LNs than a unidimensional measurement and also be less time-consuming than volume measurement. The highlight of the present study was that it represents the first report of the two-dimensional cutoff value of LN metastasis at each LN station in gastric cancer. In fact, because MDCT is originally two-dimensional picture, the two-dimensional measurement is most practical and efficient in the use of MDCT without additional reconstruction of image. In addition, in order to increase the concordance rate between area value and LN metastasis, LN area in the MDCT was measured based on each LN station and analyzed with pathological report which LN metastasis was recorded by each station. Although the cutoff value could not be calculated at all LN stations, the present study suggested concrete values for LN stations, such as stations
3, 4, and 6, with a high potential for metastasis. The common two-dimensional cutoff value of LN metastasis, measured based on the analysis of each LN station, was identical to that of LN station 3 because of the high proportion of station 3. Furthermore, the maximal LN area, two-dimensional value, showed a higher predictability for LN metastasis than shot-axis diameter, uni-dimensional value. Both methods yielded their own common cutoff values. However, among the cutoff values by each LN station using the short-axis diameter, that of only station 6 had a statistical significance. To verify the clinical implication of the two-dimensional cutoff value, enrolled patients were divided into two groups according to the common cutoff value, 112.09 mm2. The patients with a LN area greater than the common value were associated with a more advanced tumor stage, and a larger LN area was demonstrated to be a prognostic predictor of gastric cancer patients in a multivariate survival analysis. Various size criteria for metastatic LNs according to location in gastric cancer have been reported. One study used [6 mm as a criterion for metastatic LN in the perigastric region and [8 mm in the left gastric region.20 Another study reported that benign regional LNs C6 mm are more frequently detected in gastric cancer patients than in healthy populations, particularly around the lesser curvature.17 In the present study, the cutoff value for metastatic LNs of station 3 is greater than that of other stations. The exact mechanism underlying the larger LN size in station 3 is unclear. Large lymphatic streams around the lesser curvature area wound give rise to these
Two-Dimensional Values of Metastatic Lymph Nodes
A a
100
b
100
80
Sensitivity
80
Sensitivity
FIG. 2 A Receiver operating characteristic curve for a the common (area under the curve (AUC) = 0.617, P = 0.002), b LN station 3 (AUC = 0.668, P = 0.003), c LN station 4 (AUC = 0.762, P = 0.004), and d LN station 6 (AUC = 0.733, P = 0.033) cutoff values of the maximal LN area. B Receiver operating characteristic curve for a the common (area under the curve (AUC) = 0.581, P = 0.028) and b LN station 6 (AUC = 0.795, P = 0.004) cutoff values of the short diameter
60
40
20
60
40
station=3 AUC=0.668, P value=0.003
20
AUC=0.617, P value=0.002
0
0 20
0
40
60
80
0
100
20
c
100
60
80
100
d
100
80
Sensitivity
80
Sensitivity
40
100-Specificity
100-Specificity
60
40
station=4 AUC=0.762, P value=0.004
20
60
40
station=6 AUC=0.733, P value=0.033
20
0
0
0
20
40
60
80
0
100
20
40
60
80
100
100-Specificity
100-Specificity
B 100
100
a
80
Sensitivity
Sensitivity
80
b
60
40
20
60
40
20
station=6 AUC=0.795, P value=0.0037
AUC=0.581, P value=0.028 0
0 0
20
40
60
100-Specificity
results.21,22 Many LN metastases in station 3 compared with other stations were observed in the present study and two-dimensional LN measurement using MDCT was thought to reflect such consequence. Another possible mechanism is through cancer cell-induced inflammation by activating inflammatory mediators.16,17 The high prevalence of tumors in the lesser curvature (53.5%) might be related to the presence of larger LNs with cancer cell-induced inflammation in station 3.
80
100
0
20
40
60
80
100
100-Specificity
Generally, the two-dimensional value would cause a wider variation than the uni-dimensional value. Therefore, the area value was measured several times by two observers to minimize the variability. Besides calculating the cutoff value, to verify the measurement reliability, value differences from one observer and between two observers were calculated and analyzed using Bland–Altman plots with 95% confidence intervals (Fig. 3). The Bland–Altman plots indicated that intraobserver (Fig. 3a) and interobserver
S. L. Lee et al. TABLE 2 Uni–dimensional cutoff values for predicting metastasis in each lymph node station and total lymph node stations Short diameter
n
Cutoff value (mm)
No. of metastases
Total
327
[7.39
90
Sensitivity (%)
95% CI
Specificity (%)
95% CI
P value
41.11
30.8–52.0
77.22
71.3–82.4
0.028
Station 0
80
18
1
4
0
2
1
3
105
[7.49
38
50.00
33.4–66.6
73.13
60.9–83.2
0.145
4
34
[6.4
13
30.77
10.4–61.1
38.10
19.0–61.3
0.084
5
3
6
27
[8.68
11
72.73
39.0–94.0
93.75
69.8–99.8
0.004
7
40
[6.38
8
75.00
34.9–96.8
59.38
40.6–76.3
0.428
8 9
21 1
1 0
11
3
0
12
5
0
14
2
0
16
1
0
0
1
CI confidence interval
TABLE 3 Univariate and multivariate Cox proportional hazard regression survival analysis Alive
Dead
P value
HR (95% CI)
P value
Adjust HR (95% CI)
1.913 (1.347–2.717)
\0.001
2.044 (1.433–2.916)
1.470 (0.993–2.177)
0.054
P value
Age (year) \65
118 (59.3)
55 (43.0)
C65
81 (40.7)
73 (57.0)
122 (61.3)
94 (73.4)
77 (38.7)
34 (26.6)
T1
137 (68.8)
32 (25.0)
T2 T3
30 (15.1) 14 (7.0)
11 (8.6) 19 (14.8)
T4a
15 (7.5)
47 (36.7)
3.617 (2.039–6.419)
\0.001
T4b
3 (1.5)
19 (14.8)
7.247 (3.043–17.255)
\0.001
N0
138 (69.4)
42 (32.8)
N1
32 (16.1)
18 (14.1)
0.981 (0.522–1.845)
N2
18 (9.1)
20 (15.6)
1.840 (0.969–3.497)
0.063
N3
11 (5.5)
48 (37.5)
2.663 (1.503–4.716)
\0.001
M0
197 (99.0)
114 (89.1)
M1
2 (1.0)
14 (10.9)
Small
180 (90.5)
95 (74.2)
Large
19 (9.5)
33 (25.8)
0.004
1.000
1.000 \0.001
Gender Male Female
0.024
1.000
Clinical T \0.001
1.000 1.135 (0.559–2.306) 3.400 (1.812–6.377)
0.726 0.001
Clinical N \0.001
1.000 0.953
Clinical M \0.001
1.000
1.000
6.298 (3.567–11.119)
\0.001
6.128 (3.321–11.102)
\0.001
2.147 (1.444–3.194)
\0.001
1.697 (1.116–2.582)
0.013
LN area
HR hazard ratio
\0.001
1.000
1.000
a
20 0
–20
+1.96 SD 15.2 Mean
b
–1.0 –1.96 SD
Inter-observer difference
FIG. 3 Bland–Altman plots showing a intraobserver and b interobserver agreement
Intra-observer difference
Two-Dimensional Values of Metastatic Lymph Nodes 30 20
–17.2
–40 –60 –80 –100
+1.96 SD 9.1 Mean
10 0
0.2 –1.96 SD
–10
–8.7
–20 –30 –40
–120
–50 0
500
1000
1500
2000
0
Average value (mm 2)
(Fig. 3b) differences were not significant, and the measurement results were acceptable for the analysis. The present study possessed several limitations. Although the superiority of the two-dimensional cutoff value compared with the uni-dimensional value was demonstrated in the current study, the comparison of our results with three-dimensional values was not conducted and the AUC values that represent predictability were somewhat low. In addition to LN size, there are other predictable factors for LN metastasis, such as enhancement, clustering, and shape in the use of MDCT.9,13 Because the present study evaluated the adequate criteria of LN size for prediction of metastasis, the low AUC value would be caused by the study design which other predictable factors were not included in the analysis. However, the present outcomes might be adequate to verify the practicality and usefulness of two-dimensional measurement to predict LN metastasis in gastric cancer. Second, comparative and survival analyses according to the cutoff value of each LN station could not be performed because of the small number of metastatic LNs at each station and the small number of patients with single LN metastases; both were insufficient for a subgroup analysis. Third, the imaging-surgical-pathologic correlations of LNs were evaluated using station-by-station analysis, not lesion-by-lesion analysis. To our knowledge, this is the first report of a two-dimensional value, based on each station, to predict LN metastasis using MDCT for gastric cancer. In summary, the maximal LN area as the two-dimensional value represents a practical and reproducible method of identifying metastatic LNs preoperatively. Using this two-dimensional value, an individualized cutoff value could be established at each LN station in gastric cancer. ACKNOWLEDGMENT This study was supported by a grant from the National Research Foundation of Korea (No. 2012R1A1A1043576) and Fukuoka Foundation for Sound Health Cancer Research Fund. The statistical consultation was supported by the Catholic Research
500
1000
1500
2000
Average value (mm 2)
Coordinating Center of the Korea Health 21 R&D Project (A070001), Ministry of Health & Welfare, Republic of Korea. DISCLOSURE
The authors declare no conflict of interest.
REFERENCES 1. Stabile Ianora AA, Pedote P, Scardapane A, Memeo M, Rotondo A, Angelelli G. Preoperative staging of gastric carcinoma with multidetector spiral CT. Radiol Med. 2003;106:467–80. 2. Kwee RM, Kwee TC. Imaging in local staging of gastric cancer: a systematic review. J Clin Oncol. 2007;25:2107–16. 3. Kim AY, Kim HJ, Ha HK. Gastric cancer by multidetector row CT: preoperative staging. Abdom Imaging. 2005;30:465–72. 4. Chen CY, Wu DC, Kang WY, Hsu JS. Staging of gastric cancer with 16-channel MDCT. Abdom Imaging. 2006;31:514–20. 5. Matsuki M, Kani H, Tatsugami F, et al. Preoperative assessment of vascular anatomy around the stomach by 3D imaging using MDCT before laparoscopy-assisted gastrectomy. AJR Am J Roentgenol. 2004;183:145–51. 6. Pan Z, Zhang H, Yan C, et al. Determining gastric cancer resectability by dynamic MDCT. Eur Radiol. 2010;20:613–20. 7. Yan C, Zhu ZG, Yan M, et al. Value of multidetector-row computed tomography in the preoperative T and N staging of gastric carcinoma: a large-scale Chinese study. J Surg Oncol. 2009;100:205–14. 8. Hur J, Park MS, Lee JH, et al. Diagnostic accuracy of multidetector row computed tomography in T- and N staging of gastric cancer with histopathologic correlation. J Comput Assist Tomogr. 2006;30:372–7. 9. Kwee RM, Kwee TC. Imaging in assessing lymph node status in gastric cancer. Gastric Cancer. 2009;12:6–22. 10. Anzidei M, Napoli A, Zaccagna F, et al. Diagnostic performance of 64-MDCT and 1.5-T MRI with high-resolution sequences in the T staging of gastric cancer: a comparative analysis with histopathology. Radiol Med. 2009;114:1065–79. 11. Kim SH, Kim JJ, Lee JS, et al. Preoperative N staging of gastric cancer by stomach protocol computed tomography. J Gastric Cancer. 2013;13:149–56. 12. Yan C, Zhu ZG, Yan M, et al. Size of the largest lymph node visualized on multi-detector-row computed tomography (MDCT) is useful in predicting metastatic lymph node status of gastric cancer. J Int Med Res. 2010;38:22–33. 13. Lee MH, Choi D, Park MJ, Lee MW. Gastric cancer: imaging and staging with MDCT based on the 7th AJCC guidelines. Abdom Imaging. 2012;37:531–40.
S. L. Lee et al. 14. Morgagni P, Petrella E, Basile B, et al. Preoperative multidetector-row computed tomography scan staging for lymphatic gastric cancer spread. World J Surg Oncol. 2012;10:197. 15. Tokunaga M, Sugisawa N, Tanizawa Y, Bando E, Kawamura T, Terashima M. The impact of preoperative lymph node size on long-term outcome following curative gastrectomy for gastric cancer. Ann Surg Oncol. 2013;20:1598–603. 16. Zhang XP, Wang ZL, Tang L, Sun YS, Cao K, Gao Y. Support vector machine model for diagnosis of lymph node metastasis in gastric cancer with multidetector computed tomography: a preliminary study. BMC Cancer. 2011;11:10. 17. Park HS, Kim YJ, Ko SY, et al. Benign regional lymph nodes in gastric cancer on multidetector row CT. Acta Radiol. 2012;53:501–7. 18. Buerke B, Puesken M, Muter S, et al. Measurement accuracy and reproducibility of semiautomated metric and volumetric lymph node analysis in MDCT. AJR Am J Roentgenol. 2010;195:979–85.
19. Lee SM, Kim SH, Lee JM, et al. Usefulness of CT volumetry for primary gastric lesions in predicting pathologic response to neoadjuvant chemotherapy in advanced gastric cancer. Abdom Imaging. 2009;34:430–40. 20. Kim YN, Choi D, Kim SH, et al. Gastric cancer staging at isotropic MDCT including coronal and sagittal MPR images: endoscopically diagnosed early vs. advanced gastric cancer. Abdom Imaging. 2009;34:26–34. 21. Lee HH, Yoo HM, Song KY, Jeon HM, Park CH. Risk of limited lymph node dissection in patients with clinically early gastric cancer: indications of extended lymph node dissection for early gastric cancer. Ann Surg Oncol. 2013;20:3534–40. 22. Kinami S, Fujimura T, Ojima E, et al. PTD classification: proposal for a new classification of gastric cancer location based on physiological lymphatic flow. Int J Clin Oncol. 2008;13:320–9.
