Quantitative Analysis of the Microvascular Architecture Observed on Magnification Endoscopy in Cancerous and Benign Gastric Lesions

• Introduction
• Patients and Methods
• Patients
• Endoscopic Procedure
• Evaluation of the Microvascular Architecture
• Histological Investigation
• Statistical Analysis
• Results
• Magnified Observation
• Extraction of Vascular Images
• Evaluation of the Microvascular Architecture
• Histological Investigation
• Discussion
• Acknowledgments
• References

Background and Study Aims: Gastric cancer remains a common malignant tumor in Japan. The aim of this study was to attempt a quantitative evaluation of the microvascular architecture observed by magnification endoscopy using image analysis, and to investigate whether this method is able to distinguish between gastric cancers and benign lesions.
Patients and Methods: A total of 132 patients were studied using magnification endoscopy, and image analysis was performed in 71 patients (32 patients with early gastric cancer, 39 patients with benign lesions). Analysis was not possible in the other 61 patients because the quality of the image was not good enough. A square region of interest was selected from the magnified images of the gastric mucosa. From this we extracted the vascular images corresponding to microvessels and calculated the mean caliber of vessels in the region of interest.
Results: Image analysis provided good-quality images of microvessels and enabled evaluation of the microvascular architecture. The mean caliber of vessels was 4.454 pixels in 17 differentiated adenocarcinomas, 4.319 pixels in 15 undifferentiated adenocarcinomas, and 4.034 pixels in the 39 benign lesions. This represented a significant difference between gastric cancers and benign lesions (P < 0.0001). Histopathological investigation of surgically resected tumors demonstrated the mean caliber of microvessels in cancerous lesions to be greater than that of microvessels in the surrounding mucosa.
Conclusions: Image analysis was useful for evaluating the microvascular architecture of the gastric mucosa, and calculation of the mean caliber of the vessels may prove helpful in the differential diagnosis of gastric cancers. However, analysis was not possible in 61 of the 132 patients studied because of inadequate image quality, and this represents a significant limitation of this diagnostic method.

Introduction

Gastric cancer remains a common malignant tumor in Japan and several methods, including radiological examination, chromoendoscopy, and endoscopic ultrasonography, are used in the diagnosis of this tumor [1] [2]. These methods can be used for both qualitative and quantitative evaluation of gastric cancers, but histological diagnosis is usually confirmed by endoscopic biopsy.

Electronic video endoscopy, which is currently widely used, demonstrates high resolution rates and allows observation of the mucosal surface structure and of subtle color changes. However, its resolution rates are not high enough to enable the endoscopist to visualize crypt structure or microvessels. Magnification endoscopy does provide a facility for the observation of crypt structure, and its findings correlate with histological findings [3] [4] [5]. Irregularities and destruction of minute mucosal structures are characteristic of gastric cancers, and these findings are useful for differentiation of these tumors from benign lesions [6] [7] [8] [9] [10]. Nevertheless, such observations are subjective, and quantitative evaluation is difficult. In addition, inflammation or ulceration often lead to modification of the mucosal structure and these changes might mask the true diagnosis.

Magnification endoscopy also enables the observation of microvessels in the mucosal surface. Irregular microvessels and disappearance of the normal capillary network are recognized findings in gastric cancer [11]. Recently, it has been reported that quantification of vascular density or color change, using a computer, is useful in the diagnosis of gastric cancer [12]. However, quantitative evaluation of architectural irregularities of microvessels has not been reported previously. The quantification of the mucosal microvascular architecture should potentially allow for more precise diagnosis, particularly when the process is automated by using a computer. The purpose of the present study was to make a quantitative evaluation of the gastric mucosal microvascular architecture observed by magnification endoscopy, and to investigate whether this type of image analysis is able to differentiate between gastric cancers and benign lesions.

Patients and Methods

Patients

Between January 2000 and March 2003, magnification endoscopy was performed in 132 patients with a suspected gastric neoplasm. Of these 132 patients, 84 patients had early gastric cancers and 48 patients had benign lesions that were difficult to distinguish from gastric cancer by conventional endoscopy. The early gastric cancers had been diagnosed on the basis of conventional endoscopy and endoscopic ultrasound prior to treatment. All gastric cancers were treated by endoscopic mucosal resection or surgery, and histological features were confirmed by examination of the resected specimens. We excluded patients with advanced gastric cancers from this study because these tumors were easy to diagnose with conventional endoscopy. The benign lesions were diagnosed histologically after endoscopic mucosal resection in four patients and by endoscopic biopsy in all the other patients. Informed consent was obtained in writing from all patients before the examinations.

Endoscopic Procedure

The magnifying endoscope we used was the Olympus GIF-Q240Z (Olympus Co., Tokyo, Japan), which is the same as the GIF-Q240 model, except for the lens and the control lever. It provides a magnification power up to × 80 on a 14-inch monitor. We used a clear attachment (D201-11 802) that projected 2 mm from the tip of endoscope in order to maintain a fixed distance from the mucosa.

After conventional endoscopy, we examined the patient at maximum magnification, looking principally at the minute surface structure and for microvessels. When microvessels were identified in this way they were recorded as still images, which were stored as digital images using an Olympus EVIP-230 endoscopic image processor.

Evaluation of the Microvascular Architecture

The digital still images were transmitted to a computer. We excluded images on which blood vessels were not shown clearly as unsuitable for the present study. The vascular images were extracted from suitable images and analyzed using Olympus Prostudy image analysis software (Figure [1]).

Regions of interest (ROI) were set by two gastroenterologists (Y.N., R.M.), who had no knowledge of the conventional endoscopic or histological findings in the study patients. They were requested to set ROIs to a location where the blood vessels appeared most clearly. One of the two ROIs was then selected randomly and analyzed. This was an uncontrolled, prospective study.

The mean caliber of vessels within the ROI was used as the index for evaluation of the vascular architecture. The mean caliber of vessels was calculated by dividing the total number of vessel pixel elements within the ROI by the vessel length. The vessel caliber was therefore expressed as a number of pixels.

Histological Investigation

Out of the 132 patients, 31 patients had early gastric cancers resected surgically. These were subsequently diagnosed histologically as differentiated adenocarcinoma in 14 patients and as undifferentiated adenocarcinoma in 17 patients. In these resected specimens, we measured the caliber of vessels in the cancerous lesion itself and also the caliber of vessels in the surrounding mucosa. The samples were stained with hematoxylin and eosin and cavities in the stained tissue which contained erythrocytes were considered to be blood vessels. Ten or more blood vessels from each lesion were measured at a magnification of × 200 under a light microscope, and the mean caliber of the vessels was calculated.

Statistical Analysis

Student’s unpaired t-test was used for the evaluation of vessel calibers calculated using image analysis and vessel calibers measured under a light microscope. A P value of less than 0.05 was considered significant. Statview 5.0 (SAS Inc., Cary, North Carolina, USA) software was used for the statistical analysis.

Results

Magnified Observation

Magnified endoscopic images were obtained in 132 patients (84 patients with early gastric cancer and 48 patients with benign lesions). In 61 patients, the images were unsuitable for analysis (images without any blood vessels in 19 patients, images with indistinct blood vessels in 34 patients, and images with hemorrhage in eight patients). Image analysis was performed in 71 patients, 32 patients with early gastric cancer and 39 patients with benign lesions. Of the 32 early gastric cancers, 17 were differentiated adenocarcinomas (three of the elevated type and 14 of the depressed type) and 15 were undifferentiated adenocarcinomas (all depressed-type). The histological diagnosis of benign lesions was confirmed by endoscopic mucosal resection or endoscopic biopsy. Of the 39 benign lesions, 33 were reported as “regenerated epithelium”, three showed inflammation with inflammatory cell infiltration, and three were adenomas.

Extraction of Vascular Images

The image analysis software allows the operator to choose between a free-shape-type and a box-type ROI. We used a square-shaped ROI to reduce the setting bias. The size was set at 80 × 80 pixels on a computer screen, which we considered to be the minimum size for adequate evaluation of the vascular architecture. The size of the ROI was same for all the images analyzed. Good-quality images of blood vessels were obtained from endoscopic images using this method and Figure [2] shows an example of the vascular images extracted by the image analysis software.

Evaluation of the Microvascular Architecture

Suitable images of blood vessels were obtained in 32 patients with early gastric cancer and in 39 patients with benign lesions. The mean caliber of vessels in the ROIs was calculated (Figure [3]): the mean caliber of vessels was 4.454 ± 0.293 pixels in differentiated adenocarcinomas and 4.319 ± 0.340 pixels in undifferentiated adenocarcinomas, compared with a mean caliber of 4.034 ± 0.254 pixels in the benign lesions. The mean vessel caliber in the adenocarcinomas was significantly larger than the mean vessel caliber in the benign lesions (P < 0.0001). The mean caliber of vessels in the differentiated adenocarcinomas was larger than that in the undifferentiated adenocarcinomas, but this difference was not statistically significant.

Histological Investigation

We regarded cavities containing erythrocytes in the stained tissue specimens as blood vessels (Figure [4]). Vessel calibers were measured under a light microscope (Figure [5]). For differentiated adenocarcinomas, the mean caliber of blood vessels within the cancerous lesion itself was 13.557 ± 2.689 µm and the mean caliber of vessels in the surrounding mucosa was 9.257 ± 0.881 µm, a significant difference (P < 0.0001). For undifferentiated adenocarcinomas, the mean caliber of vessels within the cancerous lesion was 12.688 ± 3.118 µm and the mean caliber of vessels in the surrounding mucosa was 9.871 ± 2.675 and this difference was also statistically significant (P < 0.01). The vessel caliber within differentiated adenocarcinomas tended to be larger than the vessel caliber in undifferentiated adenocarcinomas, but this was not a statistically significant difference.

Discussion

The results of our image analysis showed that the mean caliber of vessels within cancerous lesions was significantly greater than the mean vessel caliber in benign lesions. These results suggest that image analysis allows the differentiation of gastric cancers from benign lesions.

Magnified observation of the gastric mucosa was studied during the 1980s [13] [14] [15], but magnification endoscopy at that time was operatively poor, and it did not become popular. However, technological improvements in recent years have led to the re-emergence of magnification endoscopy in the study of gastric disease.

We have investigated the minute surface structure of gastric lesions with magnification endoscopy previously, and demonstrated the characteristic findings within cancerous lesions [6]. We also described the specific findings relating to microvessels observed in cancerous lesions, and reported that those findings were useful for the diagnosis of gastric cancer.

Yao et al. [11] reported that irregularities in vessel caliber and vessel tortuosity were characteristic of cancerous lesions. They also studied vascular density, concluding that low vascular density is characteristic of undifferentiated adenocarcinoma and that high vascular density is characteristic of differentiated adenocarcinoma. They suggested that such findings were useful in the delineation of gastric cancer boundaries. Tobita [7] investigated depressed lesions and reported that various forms of abnormal blood vessels could be identified in gastric lesions, and notably that reticular, dendriform, and coil-shaped vessels were specific to cancers. Tajiri et al. [8] reported that abnormal capillary vessels with screw-like shapes were characteristic of gastric cancer [8].

Yao et al. [12] evaluated microvessels within gastric lesions quantitatively and demonstrated that the measurement of mucosal hemoglobin content (calculated from endoscopic imaging data) enabled a distinction to be made between undifferentiated gastric carcinoma and differentiated gastric carcinoma. They reported that this method was useful for the evaluation of vascularity within the gastric mucosa and for the quantification of endoscopic color changes.

If the irregularity of the microvascular architecture could be evaluated quantitatively, it would be more useful for the diagnosis of gastric cancer. Shibuya et al. [16] evaluated microvascular architecture using image analysis, using a high-magnification bronchovideoscope. They reported that the image analysis was useful in differentiating between normal bronchial mucosa, bronchitis, and dysplasia. They used vascular area ratios within the ROI for evaluating vascular morphology. However, the microvascular distribution on the mucosal surface is inhomogeneous, and there is a large degree of variation due to the use of different locations of the ROI settings.

Adachi et al. [17] injected a silicone rubber compound into blood vessels and investigated the microvascular architecture of gastric cancer and of the surrounding mucosa [17]. They reported that the caliber of blood vessels within the cancerous lesions was significantly larger than the caliber of vessels within the surrounding mucosa. They also stated that the vessel caliber in differentiated adenocarcinomas was slightly larger than the vessel caliber in undifferentiated adenocarcinomas. The caliber of vessels within cancerous lesions in the esophagus and large intestine was also shown to be larger than the caliber of vessels in the surrounding mucosa [18] [19]. For these reasons, we decided to use the caliber of vessels as the parameter for evaluating microvascular architecture in the present study.

Magnification endoscopy can provide precise images, but magnified observation is not so useful for lesions that are easy to diagnose using conventional observation methods because the magnified field of view is very small. However, magnified observation with image analysis might assist in the diagnosis of gastric lesions when there is difficulty in deciding whether they are gastric cancers or benign lesions.

Nevertheless, our study has shown that there are problems associated with using this method. First, the analysis itself could not be performed without first obtaining a clear image of the blood vessels by magnification endoscopy. In our investigation, images from only 71/132 patients were suitable for analysis. This was because we applied strict selection criteria for image quality. We think that this was a significant limitation of this study, and yet further technical improvements are required.

The second problem is bias concerning ROI settings. Endoscopists may arrive at their own diagnosis, based on conventional endoscopic findings, so there would be a strong bias in ROI setting. In order to reduce bias as much as possible in the present study, ROI settings were performed by two physicians who had no knowledge of the conventional endoscopic or histological findings. Further, prospective investigation of a greater number of patients is required.

In conclusion, the quantitative evaluation of gastric mucosal microvascular architecture using image analysis facilitates the differential diagnosis of gastric cancer, and magnification endoscopy with image analysis should progress to the development of automated diagnosis.

Acknowledgments

We wish to thank Hirokazu Nishimura and Hideki Tanaka (Olympus Co., Tokyo, Japan) for specialist technical support in preparing the image analysis software, and John Cole for editing the manuscript with regard to refinement of the use of English. This study was supported by a grant from the Japanese Foundation for Research and Promotion of Endoscopy.

References

• 1 Moreto M. Diagnosis of esophagogastric tumors.  Endoscopy. 2003;  35 36-42
  Thieme ConnectPubMedSearch in Google Scholar
• 2 Kida M, Kobayashi K, Saigenji K. Routine chromoendoscopy for gastrointestinal diseases: indications revised.  Endoscopy. 2003;  35 590-596
  Thieme ConnectPubMedSearch in Google Scholar
• 3 Takemoto T, Sakaki N. Magnification endoscopy. In: Sivak MV Gastroenterological endoscopy. Philadelphia; Saunders 2000: 165-175
  Search in Google Scholar
• 4 Cales P, Oberti F, Delmotte J S. et al . Gastric mucosal surface in cirrhosis evaluated by magnifying endoscopy and scanning electronic microscopy.  Endoscopy. 2000;  32 614-623
  Thieme ConnectPubMedSearch in Google Scholar
• 5 Yagi K, Nakamura A, Sekine A. Comparison between magnifying endoscopy and histological, culture and urease test findings from the gastric mucosa of the corpus.  Endoscopy. 2002;  34 376-381
  Thieme ConnectPubMedSearch in Google Scholar
• 6 Otsuka Y, Niwa Y, Ohmiya N. et al . Usefulness of magnifying endoscopy in the diagnosis of early gastric cancer.  Endoscopy. 2004;  36 165-169
  Thieme ConnectPubMedSearch in Google Scholar
• 7 Tobita K. Study on minute surface structures of the depressed-type early gastric cancer with magnifying endoscopy.  Dig Endosc. 2001;  13 121-126
  PubMedSearch in Google Scholar
• 8 Tajiri H, Doi T, Endo H. et al . Routine endoscopy using magnifying endoscope for gastric cancer diagnosis.  Endoscopy. 2002;  34 772-777
  Thieme ConnectPubMedSearch in Google Scholar
• 9 Kiesslish R, Jung M. Magnification endoscopy: does it improve mucosal surface analysis for the diagnosis of gastrointestinal neoplasias?.  Endoscopy. 2002;  34 819-822
  Thieme ConnectPubMedSearch in Google Scholar
• 10 Dinis-Ribeiro M, Costa-Pereira A, Lopes C. et al . Magnification chromoendoscopy for the diagnosis of gastric intestinal metaplasia and dysplasia.  Gastrointest Endosc. 2003;  57 498-504
  CrossrefPubMedSearch in Google Scholar
• 11 Yao K, Oishi T, Matsui T. et al . Novel magnified endoscopic findings of microvascular architecture in intramucosal gastric cancer.  Gastrointest Endosc. 2002;  56 279-284
  CrossrefPubMedSearch in Google Scholar
• 12 Yao K, Yao T, Matsui T. et al . Hemoglobin content in intramucosal gastric carcinoma as a marker of histologic differentiation: a clinical application of quantitative electronic endoscopy.  Gastrointest Endosc. 2000;  52 241-245
  CrossrefPubMedSearch in Google Scholar
• 13 Sakaki N, Iida Y, Okazaki Y. et al . Magnifying endoscopic observation of gastric mucosa, particularly in patients with atrophic gastritis.  Endoscopy. 1978;  10 269-274
  PubMedSearch in Google Scholar
• 14 Okada T, Nishizawa M. Magnified observation of elevated lesions of the stomach based on magnifying fiberoptic endoscopy and dissecting microscopy.  Endoscopy. 1981;  13 192-196
  PubMedSearch in Google Scholar
• 15 Nishizawa M, Okada T. Magnified observation of elevated borderline lesions (adenoma) of the stomach based on dissecting microscopy and magnifying fiberoptic endoscopy.  Endoscopy. 1981;  13 234-237
  PubMedSearch in Google Scholar
• 16 Shibuya K, Hoshino H, Chiyo M. et al . Subepithelial vascular patterns in bronchial dysplasias using a high magnification bronchovideoscope.  Thorax. 2002;  57 902-907
  CrossrefPubMedSearch in Google Scholar
• 17 Adachi Y, Mori M, Enjoji M. Microvascular architecture of early gastric carcinoma.  Cancer. 1993;  72 32-36
  CrossrefPubMedSearch in Google Scholar
• 18 Kumagai Y, Inoue Y, Nagai K. et al . Magnifying endoscopy, stereoscopic microscopy, and the microvascular architecture of superficial esophageal carcinoma.  Endoscopy. 2002;  34 364-375
  PubMedSearch in Google Scholar
• 19 Konerding M A, Fait E, Gaumann A. 3D microvascular architecture of pre-cancerous lesions and invasive carcinomas of the colon.  Br J Cancer. 2001;  84 1354-1362
  CrossrefPubMedSearch in Google Scholar

Y. Niwa, M. D.

Division of Therapeutic Medicine · Department of Internal Medicine · Nagoya · University Graduate School of Medicine

65 Tsuruma-cho · Showa-ku · Nagoya, 466-8550 · Japan ·

Fax: +81-52-744-2180

Email: yniwa@med.nagoya-u.ac.jp

References

• 1 Moreto M. Diagnosis of esophagogastric tumors.  Endoscopy. 2003;  35 36-42
  Thieme ConnectPubMedSearch in Google Scholar
• 2 Kida M, Kobayashi K, Saigenji K. Routine chromoendoscopy for gastrointestinal diseases: indications revised.  Endoscopy. 2003;  35 590-596
  Thieme ConnectPubMedSearch in Google Scholar
• 3 Takemoto T, Sakaki N. Magnification endoscopy. In: Sivak MV Gastroenterological endoscopy. Philadelphia; Saunders 2000: 165-175
  Search in Google Scholar
• 4 Cales P, Oberti F, Delmotte J S. et al . Gastric mucosal surface in cirrhosis evaluated by magnifying endoscopy and scanning electronic microscopy.  Endoscopy. 2000;  32 614-623
  Thieme ConnectPubMedSearch in Google Scholar
• 5 Yagi K, Nakamura A, Sekine A. Comparison between magnifying endoscopy and histological, culture and urease test findings from the gastric mucosa of the corpus.  Endoscopy. 2002;  34 376-381
  Thieme ConnectPubMedSearch in Google Scholar
• 6 Otsuka Y, Niwa Y, Ohmiya N. et al . Usefulness of magnifying endoscopy in the diagnosis of early gastric cancer.  Endoscopy. 2004;  36 165-169
  Thieme ConnectPubMedSearch in Google Scholar
• 7 Tobita K. Study on minute surface structures of the depressed-type early gastric cancer with magnifying endoscopy.  Dig Endosc. 2001;  13 121-126
  PubMedSearch in Google Scholar
• 8 Tajiri H, Doi T, Endo H. et al . Routine endoscopy using magnifying endoscope for gastric cancer diagnosis.  Endoscopy. 2002;  34 772-777
  Thieme ConnectPubMedSearch in Google Scholar
• 9 Kiesslish R, Jung M. Magnification endoscopy: does it improve mucosal surface analysis for the diagnosis of gastrointestinal neoplasias?.  Endoscopy. 2002;  34 819-822
  Thieme ConnectPubMedSearch in Google Scholar
• 10 Dinis-Ribeiro M, Costa-Pereira A, Lopes C. et al . Magnification chromoendoscopy for the diagnosis of gastric intestinal metaplasia and dysplasia.  Gastrointest Endosc. 2003;  57 498-504
  CrossrefPubMedSearch in Google Scholar
• 11 Yao K, Oishi T, Matsui T. et al . Novel magnified endoscopic findings of microvascular architecture in intramucosal gastric cancer.  Gastrointest Endosc. 2002;  56 279-284
  CrossrefPubMedSearch in Google Scholar
• 12 Yao K, Yao T, Matsui T. et al . Hemoglobin content in intramucosal gastric carcinoma as a marker of histologic differentiation: a clinical application of quantitative electronic endoscopy.  Gastrointest Endosc. 2000;  52 241-245
  CrossrefPubMedSearch in Google Scholar
• 13 Sakaki N, Iida Y, Okazaki Y. et al . Magnifying endoscopic observation of gastric mucosa, particularly in patients with atrophic gastritis.  Endoscopy. 1978;  10 269-274
  PubMedSearch in Google Scholar
• 14 Okada T, Nishizawa M. Magnified observation of elevated lesions of the stomach based on magnifying fiberoptic endoscopy and dissecting microscopy.  Endoscopy. 1981;  13 192-196
  PubMedSearch in Google Scholar
• 15 Nishizawa M, Okada T. Magnified observation of elevated borderline lesions (adenoma) of the stomach based on dissecting microscopy and magnifying fiberoptic endoscopy.  Endoscopy. 1981;  13 234-237
  PubMedSearch in Google Scholar
• 16 Shibuya K, Hoshino H, Chiyo M. et al . Subepithelial vascular patterns in bronchial dysplasias using a high magnification bronchovideoscope.  Thorax. 2002;  57 902-907
  CrossrefPubMedSearch in Google Scholar
• 17 Adachi Y, Mori M, Enjoji M. Microvascular architecture of early gastric carcinoma.  Cancer. 1993;  72 32-36
  CrossrefPubMedSearch in Google Scholar
• 18 Kumagai Y, Inoue Y, Nagai K. et al . Magnifying endoscopy, stereoscopic microscopy, and the microvascular architecture of superficial esophageal carcinoma.  Endoscopy. 2002;  34 364-375
  PubMedSearch in Google Scholar
• 19 Konerding M A, Fait E, Gaumann A. 3D microvascular architecture of pre-cancerous lesions and invasive carcinomas of the colon.  Br J Cancer. 2001;  84 1354-1362
  CrossrefPubMedSearch in Google Scholar

Y. Niwa, M. D.

Division of Therapeutic Medicine · Department of Internal Medicine · Nagoya · University Graduate School of Medicine

65 Tsuruma-cho · Showa-ku · Nagoya, 466-8550 · Japan ·

Fax: +81-52-744-2180

Email: yniwa@med.nagoya-u.ac.jp