Assessment for diagnosis of lymph node metastasis in esophageal cancer using endoscopic ultrasound elastography

As the first step (study 1), the lymph nodes of patients with esophageal cancer treated by surgical resection were examined by elastography and conventional B-mode imaging with a free hand linear array transducer before formalin fixation. The description form with elastography was classified, and the diagnostic criteria for metastatic lymph nodes were developed. The ultrasound platform which was used in all studies was an EUB-7500 (Hitachi Medical Systems, Tokyo, Japan). We used a convex EUS probe (EG-3870UTK; Hoya: Pentax division, Tokyo, Japan). The technical details of using these instruments are described elsewhere [8].

We examined the removed lymph node adhered by surrounding tissue (such as fiber or fat). We put each lymph node under the resected esophagus and scan the lymph node through the esophagus wall. Thereby, each lymph node could be visualized ultrasonographically and it was possible to obtain an image substantially similar to the image that has been examined in vivo using EUS elastography. Elastographic and B-mode images were displayed simultaneously side by side. The lymph node characteristics were carefully described for each lymph node: size (long axis), echogeneity (hypoechoic, hyperechoic), shape (round, oval, flat), border (regular, irregular), and elastographic pattern.

We compared the diagnostic performance of elastography with that of the conventional B-mode technique. Lymph nodes were considered malignant if three or more of the following B-mode criteria were present: lymph node >5 mm in width, round shape, hypoechoic pattern, and smooth border [3].

Lymph nodes were considered malignant if 50 % or more of the node appeared blue, or if the peripheral part of lesion was blue and the central part was red/yellow/green. Lymph nodes were considered negative for metastasis if they showed homogeneous green/yellow/red, or a mosaic pattern of green/yellow/red, or less than 50 % of the node was blue. The background of the creation of the diagnostic criteria will be discussed later.

Upon the development of the diagnostic criteria for lymph node metastasis by elastography, we referred to a scoring system based on the elastographic patterns reported in previous studies [9‐11].

As the second step (study 2), we preoperatively examined the lymph nodes of ESCC patients by EUS elastography and discriminated between benign and malignant nodes based on the newly-developed diagnostic criteria, and compared diagnostic performance of EUS elastography with that of the conventional B-mode EUS imaging. To　collate the lymph node examined by EUS elastography with the lymph node removed by the operation, we performed the following tasks: we recorded the position, size, and shape of each lymph node in EUS elastography, and at the time of operation, the position of the lymph node was checked, and the size and form of lymph node which was removed was again recorded with a free hand linear array transducer. We considered it the same lymph node whose position was the same, and size and form were most similar. A total of 115 lymph nodes were included. All lymph nodes visualized by EUS elastography were examined.

χ² test or Fisher’s exact test was used to compare diagnostic performance of elastography and B mode images. All calculations were done with SPSS 15.0 (SPSS Inc, Chicago, IL). P values of less than 0.05 were considered to be statistically significant.

We classified the elastographic images into 4 patterns, as follows:

Pattern 1: relatively homogeneous coloring or areas of two or three different colors with red, yellow, green (Fig. 1a)

.

Pattern 2: areas of red, yellow, green and blue were mixed (Fig. 1b).

Pattern 3: almost the entire area was blue (Fig. 1c).

Pattern 4: the margin of the tumor was depicted in blue, and the center in red, yellow or/and green (Fig. 1d).

Although there are minor differences, all of the previous studies indicated that malignancy is suggested by an increased proportion of blue. Pattern 4 seemed to be a characteristic finding that reflected the internal necrosis of cancer. When pattern 1 was assumed to be negative for metastasis, and the rest was assumed to be positive, the negative predictive value was excellent, at 94.0 %. When patterns 3 and 4 were assumed to be positive for metastasis, and 1 and 2 were assumed to be negative, the positive predictive value was excellent, at 95.7 %. Therefore, it seemed to be reasonable to regard that pattern 1 as negative and patterns 3 and 4 as positive findings. However, it was difficult to decide how to subdivide pattern 2 to increase the diagnostic performance.

To solve this problem, we measured the area of lymph node and the blue area inside as a circle or ellipse for every lymph node shown to be pattern 2, evaluated the percentage of blue, and a receiver operator characteristic curve was generated to determine the cut off value (Fig. 2). If the cut-off value was assumed to be 50 %, the sensitivity was 85 % and the specificity was 89 %.

Therefore, the lymph nodes that were observed to be 50 % or more blue were assumed to be metastasis positive in this study. In other words, the criteria for diagnosing negative lymph nodes using elastography were the image like pattern 1 and 2, and the image which the blue part of a lymph node was less than 50 % (Fig. 3a), and the criteria for diagnosing positive lymph nodes were the presence of blue accountings for 50 % or more of the lesion (including pattern 3), and lesions characterized by pattern 4 (Fig. 3b).

Typical images are shown in Fig. 4. The elastographic images are shown in upper stage, lower is pathological each. (a) A pattern 1 sample which was shown to be green by elastography and proved to be negative for metastasis. (b) A pattern 2 sample which blue dots were seen but the percentage compared with the whole area of the lymph node was small, and it was pathologically negative. (c) Also pattern 3 sample which the blue ratio among the lymph node is more than 50 % and proved to be positive for metastasis. Localization of blue indicated by elastography and localization of metastatic focus in pathological finding appears to be similar. (d) Also pattern 3 sample which was shown to be blue and proved to be metastatic positive. The occupation rate was 100 %. (e) A pattern 4 sample whose border of the lymph node was blue, and the inside was depicted with green, red, and yellow. It was pathologically positive for metastasis accompanied by necrosis inside.

Histological examination of the surgical specimens revealed benign tissue in 253 lymph nodes, and malignant tissue in 69 lymph nodes. The elastographic images were interpreted as benign in 247 lymph nodes, and malignant in 55 lymph nodes. The calculated sensitivity, specificity, positive and negative predictive values of the elastography were 79.7, 97.6, 90.2 and 94.6 %, respectively, with an accuracy of 93.8 %. The calculated sensitivity, specificity, positive and negative predictive values for the conventional B-mode images were 62.3, 60.6, 30.3 and 85.6 %, respectively, with an accuracy of 61.1 %. The sensitivity and specificity of elastography were significantly better than those of the B-mode technique (Table 1). A detailed investigation revealed that the smaller metastatic lymph node can be detected by elastography compared with B-mode. A representative case is shown in Fig. 5. The lymph node was 4 × 3.1 mm, round shape, smooth border, an internal echo was homogenous in B mode (Fig. 5a), and it was diagnosed as negative. On the other hand, the entire lymph node was shown to be blue by elastography, and the lymph node was classified as having Pattern 3, and was therefore diagnosed as positive (Fig. 5b). The pathological findings revealed that the whole lymph node was almost completely occupied by the carcinoma cells (Fig. 5c).

In study 2, the histological examination of the surgical specimens revealed benign tissue in 91 lymph nodes, and malignant tissue in 24 lymph nodes. The images of EUS elastography were interpreted as benign in 86 lymph nodes, and malignant in 22 lymph nodes. The calculated sensitivity, specificity, positive and negative predictive values of the EUS elastography were 91.2, 94.5, 81.5 and 97.8 %, respectively, with an accuracy of 93.9 %. The calculated sensitivity, specificity, positive and negative predictive values of the conventional B-mode EUS were 62.5, 73.5, 38.5 and 88.2 %, respectively, with an accuracy of 71.3 %. The sensitivity and specificity of EUS elastography were significantly better than those of the B-mode technique (Table 2).

In study 2, with since the observation of anal side of the tumor which is making stenosis is difficult, and the lymph node at the position away from gastrointestinal tract cannot be targeted for inspection, it is thought that there were fewer evaluation lymph nodes in one case in Study2 than Study1. The compression during EUS examination is naturally obtained by arterial pulsations and respiratory movements [12].