- Email: [email protected]
Detection of early gastric cancer by a real-time autofluorescence imaging system
Cancer Letters 165 (2001) 155±159
www.elsevier.com/locate/canlet
Detection of early gastric cancer by a real-time auto¯uorescence imaging system Masahiko Kobayashi a,*, Hisao Tajiri b, Eiji Seike a, Masanori Shitaya a, Shigetaka Tounou a, Masafumi Mine a, Kenichi Oba a a
Department of Internal Medicine, Japan Self Defense Forces Central Hospital, Ikejiri 1-2-24, Setagaya-ku, Tokyo 154-8532, Japan b Department of Clinical Research, National Shikoku Cancer Center, Horinouchi 13, Matsuyama, Ehime 790-0007, Japan Received 15 November 2000; received in revised form 12 January 2001; accepted 12 January 2001
Abstract A light-induced ¯uorescence endoscopy in the gastrointestinal tract system was used in 52 patients with 54 lesions (33 early gastric cancers, 21 benign lesions) to assess its ability to detect early gastric cancer. Comparing the images with the histological ®ndings, 21 of the 33 carcinomas appeared dark red, ten had a mixed pattern of dark red and white, and two could not be detected. Of the 21 benign lesions, 18 appeared light blue, as do normal mucosa, with this system. In 85% of the cancer lesions (28/33), cancer extension was correctly detected. The sensitivity and speci®city were 94 and 86%, respectively. Real-time auto¯uorescence endoscopy is a useful adjunct to conventional white-light endoscopy for detecting early gastric cancer. q 2001 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Auto¯uorescence endoscopy; Cancer extension; Light-induced ¯uorescence endoscopy in the gastrointestinal tract system
1. Introduction The use of ¯uorescence endoscopy to detect early carcinomas and to discriminate between normal tissue and neoplastic lesions has recently attracted considerable attention. This spectroscopic technique is classi®ed into two main types: one is photodynamic diagnosis using ¯uorescence drugs called photosensitizers, which react to various wavelengths of light [1,2], and the other method is auto¯uorescence diagnosis, which utilizes inherent tissue auto¯uorescence. Until recently, detecting the difference in auto¯uorescence as an endoscopic image has been very dif®cult because of its faintness, but the advent of powerful light sources and highly sensitive cameras has led to
the development of auto¯uorescence endoscopy systems, such as the light-induced ¯uorescence endoscopy in the gastrointestinal tract system (LIFE-GI; Xillix Technology Co., Canada and Olympus Optical Co., Japan) or the lung imaging or laser-induced ¯uorescence endoscopy system (LIFE system; Xillix Technology Co.), that can image green and red ¯uorescence in real time. These systems have been tested clinically and their usefulness has been reported [3,4]. The present study investigated the ability of the LIFE-GI system to differentiate early gastric cancer [5] from benign lesions. 2. Patients and methods 2.1. LIFE-GI system
* Corresponding author. Tel.: 181-3-34110151; fax: 181-334130093.
The LIFE-GI system comprises a mercury lamp and
0304-3835/00/$ - see front matter q 2001 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0304-383 5(01)00405-0
156
M. Kobayashi et al. / Cancer Letters 165 (2001) 155±159
mixed (dark red and white) and the healthy areas are light blue. 2.2. Patients
Fig. 1. Diagram of the LIFE-GI system.
a blue excitation ®lter, a white-light source, two highsensitivity charge-coupled device (CCD) cameras, a color CCD camera for white-light, an endoscope, and a monitor (Fig. 1). First, a 437-nm wavelength of blue light from the mercury lamp is ®ltered through the excitation ®lter and is used to illuminate the tissue through the endoscope. The green and red auto¯uorescence generated by the tissue (wavelength, 490±560 and .630, nm, respectively) is ampli®ed by the two built-in, high-sensitivity imaging elements in the CCD camera attached to the end of the endoscope. After the individual image signals are processed in real time, they are displayed on the monitor in pseudo-color; the cancerous areas are dark red or
The subjects were 52 patients with 54 lesions (33 early gastric cancers, ten gastric ulcers, six hyperplastic polyps and ®ve erosions). The clinicopathological parameters were assessed on the basis of the Japanese Classi®cation of Gastric Carcinoma [5]. Early gastric cancer was histologically de®ned as when tumor invasion was observed in the mucosa (including muscularis mucosae) or submucosa. The macroscopic classi®cation of early gastric cancer was based on the classi®cation of the Japan Gastroenterological Endoscopy Society. Histologically, early gastric cancers were divided into two groups, differentiated cancer (well and moderately differentiated adenocarcinoma and papillary adenocarcinoma) and undifferentiated cancer (poorly differentiated adenocarcinoma and signet-ring cell carcinoma). The details of 33 early gastric cancers are shown in Table 1. 2.3. Methods The subjects were initially examined with conventional white-light endoscopy (GIFQ40, Olympus Optical Co., Japan), followed by auto¯uorescence endoscopy. All procedures were documented with video and still images. After obtaining the LIFE images, biopsy samples were taken from the abnormal areas, as well as from adjacent normal mucosa. The
Table 1 Details of 33 early gastric cancers Macroscopic type a,b
Microscopic type
Depth of invasion
I 1 IIa: two cases
Differentiated: two cases
IIa: seven cases
Differentiated: seven cases
IIc: 18 cases
Differentiated: 13 cases Undifferentiated: ®ve cases Differentiated: two cases Undifferentiated: four cases
Mucosa: one case Submucosa: one case Mucosa: six cases Submucosa: one case Mucosa: 14 cases Submucosa: four cases Mucosa: three cases Submucosa: three cases
IIc 1 III: six cases
a Macroscopic type is indicated as follows: I, super®cial, ¯at tumor with minimal protrusion; IIa, super®cial, ¯at tumor with minimal super®cial elevation; IIc, super®cial, ¯at tumor with minimal depression; III, super®cial, ¯at tumor with minimal excavation. b ` 1 ' represents the combined type.
M. Kobayashi et al. / Cancer Letters 165 (2001) 155±159
157
LIFE images of gastric cancer were additionally compared with the histological ®ndings of the resected specimens. Informed consent was obtained from all patients who participated in this study by document. The study was approved by our institutional review board. 3. Results The relationship between the macroscopic type of cancer and auto¯uorescence was that all of the protruded and super®cial elevated types (I 1 IIa and IIa; nine cases) showed a dark red pattern; of the 18 super®cial depressed types (IIc), 11 appeared dark red, ®ve were mixed and two had a light blue pattern. In six cases of super®cial depressed and excavated types (IIc 1 III), one showed a dark red pattern and the remaining ®ve had mixed patterns (Table 2). Comparison of the auto¯uorescence images and the histological ®ndings revealed that of the 33 carcinomas, 21 appeared dark red (Figs. 2 and 3), ten had a mixed pattern of dark red and white ¯uorescence (Fig. 4), and the remaining two could not be detected. Furthermore, the cancer extension detected by this system corresponded well to that con®rmed by histology in 85% (28/33) of cases. With regard to the correlation of the color images obtained with the LIFE-GI system and the histological type, all of the cancerous lesions that appeared dark red were differentiated adenocarcinomas, and seven of the ten cases that had a mixed pattern were undifferentiated adenocarcinomas with ®brosis, with the remaining three being differentiated adenocarcinomas with submucosal invasion and/or ®brosis. Of the 21 benign lesions,
Fig. 2. White-light endoscopic image of super®cial depressed type early gastric cancer (upper left). Its auto¯uorescence shows a dark red pattern (upper right). Resected picture of the same site (lower left). Schema of sectioning (lower right), red bands show the tumor invasion. The histological type was differentiated cancer and the depth of invasion was the mucosa.
18 were light blue, similar to normal mucosa (Fig. 5), and the other three were a reddish color. The sensitivity and speci®city were 94 and 86%, respectively (Table 3). In total, 94% (31/33) of early gastric cancers could be detected only by the LIFE-GI system in this study.
Table 2 Macroscopic type and auto¯uorescence pattern Macroscopic type
Cases
Auto¯uorescence pattern a
Cases
I 1 IIa IIa IIc
2 7 18
Dark red Dark red Dark red Mixed Light blue Dark red Mixed
2 7 11 5 2 1 5
IIc 1 III a
6
Mixed, mixture of dark red and white patterns.
Fig. 3. White-light endoscopic image of super®cial protruded and super®cial elevated type early gastric cancer (upper left). Its auto¯uorescence shows a dark red pattern (upper right). Resected picture of the same site (lower left). Schema of sectioning (lower right), red bands show the tumor invasion. The histological type was differentiated cancer and the depth of invasion was the mucosa.
158
M. Kobayashi et al. / Cancer Letters 165 (2001) 155±159
Fig. 4. White-light endoscopic image of super®cial depressed type early gastric cancer (upper left). Its auto¯uorescence shows a mixed (dark red and white) pattern (upper right). Resected picture of the same site (lower left). Schema of sectioning (lower right), red bands show the tumor invasion. The histological type was differentiated cancer and the depth of invasion was the submucosa.
4. Discussion There are a number of biological ¯uorophores, such as ¯avins (ribo¯avin, ¯avin mononucleotide and ¯avin dinucleotide), nicotinamide adenine dinucleotide, collagen and pyridoxal 5 0 -phosphate, that could be responsible for the observed differences in the auto¯uorescence spectra of normal and diseased tissues. Our previous study [6] showed that the main source of tissue ¯uorescence is submucosal collagen, and that the masking effect of the thickened mucosa overlaying the submucosal layer is an important factor in the reduced ¯uorescence of diseased tissues. In other words, when there is adenomatous or cancerous proliferation, in which the abnormal epithelium is thicker than normal, ¯uorescence-exciting lights,
such as lasers or specially ®ltered lamps, may be blocked or reabsorbed to a greater degree than in normal epithelium [7]. On the other hand, the increased blood volume in diseased tissues is a possible explanation of these differences in ¯uorescence because hemoglobin has an absorption effect in the green spectrum. DaCosta et al. reported that LIFE imaging in vivo was able to: (a), differentiate between hyperplastic and adenomatous polyps of the same gross size and shape; (b), detect ¯at adenomas that have no gross mucosal thickening and are topographically similar to surrounding mucosa; (c), delineate the boundary of Barrett's esophagus; and, in some cases, (d), detect dysplastic lesions in a Barrett's ®eld [8]. This suggests that there are important tissue changes other than, or in addition to, changes in gross tissue morphology, and these may include alterations in the local blood volume, tissue metabolic activity, and relative ¯uorophore concentrations. The ¯uorescence images obtained with the LIFEGI system and the histopathological ®ndings have been compared in a series of 61 gastric cancer lesions to clarify the colors seen in gastric cancer [9]. A total of 429 evaluation points in the specimens were selected according to the tumor size. The system detected 58 (95.1%) of the 61 gastric cancer lesions, and almost all appeared dark red. The detection rate increased as the depth of invasion increased, and increased signi®cantly as the mucosa that was invaded became thicker; that is, mucosal thickening as a result of cancer cell invasion had a great impact on the detection rate. In the present study, we used the LIFE-GI system to examine patients with early gastric cancer to investiTable 3 Sensitivity and speci®city of the LIFE-GI system a Total cases True positive True negative False negative Sensitivity (%) Speci®city (%)
Fig. 5. Fundic gland polyp, light blue pattern similar to normal mucosa.
54 31 18 2 94 86
a True positive, ¯uorescence image positive and a pathology of malignancy; true negative, ¯uorescence image negative and pathology benign; false positive, ¯uorescence image positive and pathology benign; false negative, ¯uorescence image negative and a pathology of malignancy.
M. Kobayashi et al. / Cancer Letters 165 (2001) 155±159
gate its clinical ef®cacy. Of the 33 early gastric cancers, comparison between the auto¯uorescence image and the histopathology of the resected specimen revealed that 21 (64%) appeared dark red, and ten (30%) had a mixed pattern of dark red and white ¯uorescence. In addition, the system correctly detected cancer extension in 85% (28/33) of cases. The color images obtained with the LIFE-GI system correlated well to the histological type and to ®brosis within the cancerous lesion. The LIFE-GI system has the possibility of detecting earlier and smaller malignant lesions which were not identi®ed by routine white-light endoscopy. In particular, it might be useful for differentiating between super®cial depressed or excavated types of early gastric cancer from benign lesions, such as erosions. An improved version of the LIFE-GI system, which uses a metal-halide lamp as a higher ¯uorescenceexciting light and is easier to operate, is now available for clinical use and we will be continuing our research to determine the exact relationship between the auto¯uorescence image and the histological ®ndings for the new system. 5. Conclusion Real-time auto¯uorescence endoscopy has promise as an adjunct to conventional white-light endoscopy for the early detection or assessment of the extension of early gastric cancer. Acknowledgements The authors would like to thank Olympus Optical
159
Co., Japan, for their technical support with the LIFEGI system. This work was supported in a Grant-in-Aid for Cancer Research from the Ministry of Health and Welfare of Japan.
References [1] H. Tajiri, K. Yokoyama, N. Boku, A. Ohtsu, T. Fujii, S. Yoshida, T. Sato, K. Hakamata, K. Hayashi, I. Sakata, Fluorescent diagnosis of experimental gastric cancer using a tumorlocalizing photosensitizer, Cancer Lett. 111 (1997) 215± 220. [2] J. Haringsma, G.N.J. Tytgat, The value of ¯uorescence techniques in gastrointestinal endoscopy: better than the endoscopist's eye? I: The European experience, Endoscopy 30 (1998) 416±418. [3] K. Izuishi, H. Tajiri, M. Ryu, J. Furuse, Y. Maru, K. Inoue, M. Konishi, T. Kinoshita, Detection of bile duct cancer by auto¯uorescence cholangioscopy: a pilot study, Hepato-gastroenterology 46 (1999) 804±807. [4] H. Tajiri, M. Kobayashi, S. Yoshida, Fluorescence endoscopy in the gastrointestinal tract, Dig. Endosc. (2001) 12 in press. [5] Japanese Research Society for Gastric Cancer, Japanese Classi®cation of Gastric Carcinoma, Kanehara, Tokyo, 1995. [6] K. Izuishi, H. Tajiri, T. Fujii, N. Boku, A. Ohtsu, T. Ohnishi, M. Ryu, T. Kinoshita, S. Yoshida, The histological basis of detection of adenoma and cancer in the colon by auto¯uorescence endoscopic imaging, Endoscopy 31 (1999) 1±6. [7] H. Stepp, R. Sroka, R. Baumgartner, Fluorescence endoscopy of gastrointestinal diseases: basic principles, techniques, and clinical experience, Endoscopy 30 (1998) 379±386. [8] R.S. DaCosta, B.C. Wilson, N.E. Marcon, Recent advances in light-induced ¯uorescence endoscopy (LIFE) of the gastrointestinal tract, Dig. Endosc. 11 (1999) 108±118. [9] S. Abe, K. Izuishi, H. Tajiri, T. Kinoshita, T. Matsuoka, Correlation of in vitro ¯uorescence endoscopy images with histopathological ®ndings in stomach cancer, Endoscopy 32 (2000) 281±286.
www.elsevier.com/locate/canlet
Detection of early gastric cancer by a real-time auto¯uorescence imaging system Masahiko Kobayashi a,*, Hisao Tajiri b, Eiji Seike a, Masanori Shitaya a, Shigetaka Tounou a, Masafumi Mine a, Kenichi Oba a a
Department of Internal Medicine, Japan Self Defense Forces Central Hospital, Ikejiri 1-2-24, Setagaya-ku, Tokyo 154-8532, Japan b Department of Clinical Research, National Shikoku Cancer Center, Horinouchi 13, Matsuyama, Ehime 790-0007, Japan Received 15 November 2000; received in revised form 12 January 2001; accepted 12 January 2001
Abstract A light-induced ¯uorescence endoscopy in the gastrointestinal tract system was used in 52 patients with 54 lesions (33 early gastric cancers, 21 benign lesions) to assess its ability to detect early gastric cancer. Comparing the images with the histological ®ndings, 21 of the 33 carcinomas appeared dark red, ten had a mixed pattern of dark red and white, and two could not be detected. Of the 21 benign lesions, 18 appeared light blue, as do normal mucosa, with this system. In 85% of the cancer lesions (28/33), cancer extension was correctly detected. The sensitivity and speci®city were 94 and 86%, respectively. Real-time auto¯uorescence endoscopy is a useful adjunct to conventional white-light endoscopy for detecting early gastric cancer. q 2001 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Auto¯uorescence endoscopy; Cancer extension; Light-induced ¯uorescence endoscopy in the gastrointestinal tract system
1. Introduction The use of ¯uorescence endoscopy to detect early carcinomas and to discriminate between normal tissue and neoplastic lesions has recently attracted considerable attention. This spectroscopic technique is classi®ed into two main types: one is photodynamic diagnosis using ¯uorescence drugs called photosensitizers, which react to various wavelengths of light [1,2], and the other method is auto¯uorescence diagnosis, which utilizes inherent tissue auto¯uorescence. Until recently, detecting the difference in auto¯uorescence as an endoscopic image has been very dif®cult because of its faintness, but the advent of powerful light sources and highly sensitive cameras has led to
the development of auto¯uorescence endoscopy systems, such as the light-induced ¯uorescence endoscopy in the gastrointestinal tract system (LIFE-GI; Xillix Technology Co., Canada and Olympus Optical Co., Japan) or the lung imaging or laser-induced ¯uorescence endoscopy system (LIFE system; Xillix Technology Co.), that can image green and red ¯uorescence in real time. These systems have been tested clinically and their usefulness has been reported [3,4]. The present study investigated the ability of the LIFE-GI system to differentiate early gastric cancer [5] from benign lesions. 2. Patients and methods 2.1. LIFE-GI system
* Corresponding author. Tel.: 181-3-34110151; fax: 181-334130093.
The LIFE-GI system comprises a mercury lamp and
0304-3835/00/$ - see front matter q 2001 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0304-383 5(01)00405-0
156
M. Kobayashi et al. / Cancer Letters 165 (2001) 155±159
mixed (dark red and white) and the healthy areas are light blue. 2.2. Patients
Fig. 1. Diagram of the LIFE-GI system.
a blue excitation ®lter, a white-light source, two highsensitivity charge-coupled device (CCD) cameras, a color CCD camera for white-light, an endoscope, and a monitor (Fig. 1). First, a 437-nm wavelength of blue light from the mercury lamp is ®ltered through the excitation ®lter and is used to illuminate the tissue through the endoscope. The green and red auto¯uorescence generated by the tissue (wavelength, 490±560 and .630, nm, respectively) is ampli®ed by the two built-in, high-sensitivity imaging elements in the CCD camera attached to the end of the endoscope. After the individual image signals are processed in real time, they are displayed on the monitor in pseudo-color; the cancerous areas are dark red or
The subjects were 52 patients with 54 lesions (33 early gastric cancers, ten gastric ulcers, six hyperplastic polyps and ®ve erosions). The clinicopathological parameters were assessed on the basis of the Japanese Classi®cation of Gastric Carcinoma [5]. Early gastric cancer was histologically de®ned as when tumor invasion was observed in the mucosa (including muscularis mucosae) or submucosa. The macroscopic classi®cation of early gastric cancer was based on the classi®cation of the Japan Gastroenterological Endoscopy Society. Histologically, early gastric cancers were divided into two groups, differentiated cancer (well and moderately differentiated adenocarcinoma and papillary adenocarcinoma) and undifferentiated cancer (poorly differentiated adenocarcinoma and signet-ring cell carcinoma). The details of 33 early gastric cancers are shown in Table 1. 2.3. Methods The subjects were initially examined with conventional white-light endoscopy (GIFQ40, Olympus Optical Co., Japan), followed by auto¯uorescence endoscopy. All procedures were documented with video and still images. After obtaining the LIFE images, biopsy samples were taken from the abnormal areas, as well as from adjacent normal mucosa. The
Table 1 Details of 33 early gastric cancers Macroscopic type a,b
Microscopic type
Depth of invasion
I 1 IIa: two cases
Differentiated: two cases
IIa: seven cases
Differentiated: seven cases
IIc: 18 cases
Differentiated: 13 cases Undifferentiated: ®ve cases Differentiated: two cases Undifferentiated: four cases
Mucosa: one case Submucosa: one case Mucosa: six cases Submucosa: one case Mucosa: 14 cases Submucosa: four cases Mucosa: three cases Submucosa: three cases
IIc 1 III: six cases
a Macroscopic type is indicated as follows: I, super®cial, ¯at tumor with minimal protrusion; IIa, super®cial, ¯at tumor with minimal super®cial elevation; IIc, super®cial, ¯at tumor with minimal depression; III, super®cial, ¯at tumor with minimal excavation. b ` 1 ' represents the combined type.
M. Kobayashi et al. / Cancer Letters 165 (2001) 155±159
157
LIFE images of gastric cancer were additionally compared with the histological ®ndings of the resected specimens. Informed consent was obtained from all patients who participated in this study by document. The study was approved by our institutional review board. 3. Results The relationship between the macroscopic type of cancer and auto¯uorescence was that all of the protruded and super®cial elevated types (I 1 IIa and IIa; nine cases) showed a dark red pattern; of the 18 super®cial depressed types (IIc), 11 appeared dark red, ®ve were mixed and two had a light blue pattern. In six cases of super®cial depressed and excavated types (IIc 1 III), one showed a dark red pattern and the remaining ®ve had mixed patterns (Table 2). Comparison of the auto¯uorescence images and the histological ®ndings revealed that of the 33 carcinomas, 21 appeared dark red (Figs. 2 and 3), ten had a mixed pattern of dark red and white ¯uorescence (Fig. 4), and the remaining two could not be detected. Furthermore, the cancer extension detected by this system corresponded well to that con®rmed by histology in 85% (28/33) of cases. With regard to the correlation of the color images obtained with the LIFE-GI system and the histological type, all of the cancerous lesions that appeared dark red were differentiated adenocarcinomas, and seven of the ten cases that had a mixed pattern were undifferentiated adenocarcinomas with ®brosis, with the remaining three being differentiated adenocarcinomas with submucosal invasion and/or ®brosis. Of the 21 benign lesions,
Fig. 2. White-light endoscopic image of super®cial depressed type early gastric cancer (upper left). Its auto¯uorescence shows a dark red pattern (upper right). Resected picture of the same site (lower left). Schema of sectioning (lower right), red bands show the tumor invasion. The histological type was differentiated cancer and the depth of invasion was the mucosa.
18 were light blue, similar to normal mucosa (Fig. 5), and the other three were a reddish color. The sensitivity and speci®city were 94 and 86%, respectively (Table 3). In total, 94% (31/33) of early gastric cancers could be detected only by the LIFE-GI system in this study.
Table 2 Macroscopic type and auto¯uorescence pattern Macroscopic type
Cases
Auto¯uorescence pattern a
Cases
I 1 IIa IIa IIc
2 7 18
Dark red Dark red Dark red Mixed Light blue Dark red Mixed
2 7 11 5 2 1 5
IIc 1 III a
6
Mixed, mixture of dark red and white patterns.
Fig. 3. White-light endoscopic image of super®cial protruded and super®cial elevated type early gastric cancer (upper left). Its auto¯uorescence shows a dark red pattern (upper right). Resected picture of the same site (lower left). Schema of sectioning (lower right), red bands show the tumor invasion. The histological type was differentiated cancer and the depth of invasion was the mucosa.
158
M. Kobayashi et al. / Cancer Letters 165 (2001) 155±159
Fig. 4. White-light endoscopic image of super®cial depressed type early gastric cancer (upper left). Its auto¯uorescence shows a mixed (dark red and white) pattern (upper right). Resected picture of the same site (lower left). Schema of sectioning (lower right), red bands show the tumor invasion. The histological type was differentiated cancer and the depth of invasion was the submucosa.
4. Discussion There are a number of biological ¯uorophores, such as ¯avins (ribo¯avin, ¯avin mononucleotide and ¯avin dinucleotide), nicotinamide adenine dinucleotide, collagen and pyridoxal 5 0 -phosphate, that could be responsible for the observed differences in the auto¯uorescence spectra of normal and diseased tissues. Our previous study [6] showed that the main source of tissue ¯uorescence is submucosal collagen, and that the masking effect of the thickened mucosa overlaying the submucosal layer is an important factor in the reduced ¯uorescence of diseased tissues. In other words, when there is adenomatous or cancerous proliferation, in which the abnormal epithelium is thicker than normal, ¯uorescence-exciting lights,
such as lasers or specially ®ltered lamps, may be blocked or reabsorbed to a greater degree than in normal epithelium [7]. On the other hand, the increased blood volume in diseased tissues is a possible explanation of these differences in ¯uorescence because hemoglobin has an absorption effect in the green spectrum. DaCosta et al. reported that LIFE imaging in vivo was able to: (a), differentiate between hyperplastic and adenomatous polyps of the same gross size and shape; (b), detect ¯at adenomas that have no gross mucosal thickening and are topographically similar to surrounding mucosa; (c), delineate the boundary of Barrett's esophagus; and, in some cases, (d), detect dysplastic lesions in a Barrett's ®eld [8]. This suggests that there are important tissue changes other than, or in addition to, changes in gross tissue morphology, and these may include alterations in the local blood volume, tissue metabolic activity, and relative ¯uorophore concentrations. The ¯uorescence images obtained with the LIFEGI system and the histopathological ®ndings have been compared in a series of 61 gastric cancer lesions to clarify the colors seen in gastric cancer [9]. A total of 429 evaluation points in the specimens were selected according to the tumor size. The system detected 58 (95.1%) of the 61 gastric cancer lesions, and almost all appeared dark red. The detection rate increased as the depth of invasion increased, and increased signi®cantly as the mucosa that was invaded became thicker; that is, mucosal thickening as a result of cancer cell invasion had a great impact on the detection rate. In the present study, we used the LIFE-GI system to examine patients with early gastric cancer to investiTable 3 Sensitivity and speci®city of the LIFE-GI system a Total cases True positive True negative False negative Sensitivity (%) Speci®city (%)
Fig. 5. Fundic gland polyp, light blue pattern similar to normal mucosa.
54 31 18 2 94 86
a True positive, ¯uorescence image positive and a pathology of malignancy; true negative, ¯uorescence image negative and pathology benign; false positive, ¯uorescence image positive and pathology benign; false negative, ¯uorescence image negative and a pathology of malignancy.
M. Kobayashi et al. / Cancer Letters 165 (2001) 155±159
gate its clinical ef®cacy. Of the 33 early gastric cancers, comparison between the auto¯uorescence image and the histopathology of the resected specimen revealed that 21 (64%) appeared dark red, and ten (30%) had a mixed pattern of dark red and white ¯uorescence. In addition, the system correctly detected cancer extension in 85% (28/33) of cases. The color images obtained with the LIFE-GI system correlated well to the histological type and to ®brosis within the cancerous lesion. The LIFE-GI system has the possibility of detecting earlier and smaller malignant lesions which were not identi®ed by routine white-light endoscopy. In particular, it might be useful for differentiating between super®cial depressed or excavated types of early gastric cancer from benign lesions, such as erosions. An improved version of the LIFE-GI system, which uses a metal-halide lamp as a higher ¯uorescenceexciting light and is easier to operate, is now available for clinical use and we will be continuing our research to determine the exact relationship between the auto¯uorescence image and the histological ®ndings for the new system. 5. Conclusion Real-time auto¯uorescence endoscopy has promise as an adjunct to conventional white-light endoscopy for the early detection or assessment of the extension of early gastric cancer. Acknowledgements The authors would like to thank Olympus Optical
159
Co., Japan, for their technical support with the LIFEGI system. This work was supported in a Grant-in-Aid for Cancer Research from the Ministry of Health and Welfare of Japan.
References [1] H. Tajiri, K. Yokoyama, N. Boku, A. Ohtsu, T. Fujii, S. Yoshida, T. Sato, K. Hakamata, K. Hayashi, I. Sakata, Fluorescent diagnosis of experimental gastric cancer using a tumorlocalizing photosensitizer, Cancer Lett. 111 (1997) 215± 220. [2] J. Haringsma, G.N.J. Tytgat, The value of ¯uorescence techniques in gastrointestinal endoscopy: better than the endoscopist's eye? I: The European experience, Endoscopy 30 (1998) 416±418. [3] K. Izuishi, H. Tajiri, M. Ryu, J. Furuse, Y. Maru, K. Inoue, M. Konishi, T. Kinoshita, Detection of bile duct cancer by auto¯uorescence cholangioscopy: a pilot study, Hepato-gastroenterology 46 (1999) 804±807. [4] H. Tajiri, M. Kobayashi, S. Yoshida, Fluorescence endoscopy in the gastrointestinal tract, Dig. Endosc. (2001) 12 in press. [5] Japanese Research Society for Gastric Cancer, Japanese Classi®cation of Gastric Carcinoma, Kanehara, Tokyo, 1995. [6] K. Izuishi, H. Tajiri, T. Fujii, N. Boku, A. Ohtsu, T. Ohnishi, M. Ryu, T. Kinoshita, S. Yoshida, The histological basis of detection of adenoma and cancer in the colon by auto¯uorescence endoscopic imaging, Endoscopy 31 (1999) 1±6. [7] H. Stepp, R. Sroka, R. Baumgartner, Fluorescence endoscopy of gastrointestinal diseases: basic principles, techniques, and clinical experience, Endoscopy 30 (1998) 379±386. [8] R.S. DaCosta, B.C. Wilson, N.E. Marcon, Recent advances in light-induced ¯uorescence endoscopy (LIFE) of the gastrointestinal tract, Dig. Endosc. 11 (1999) 108±118. [9] S. Abe, K. Izuishi, H. Tajiri, T. Kinoshita, T. Matsuoka, Correlation of in vitro ¯uorescence endoscopy images with histopathological ®ndings in stomach cancer, Endoscopy 32 (2000) 281±286.