Published online Jul 14, 2013. doi: 10.3748/wjg.v19.i26.4221
Revised: March 28, 2013
Accepted: June 1, 2013
Published online: July 14, 2013
AIM: To evaluate the feasibility of a new computed virtual chromoendoscopy (CVC) device (M i-scan) in the diagnosis of gastric neoplasia.
METHODS: Patients with superficial lesions no larger than 1.0 cm found during high definition endoscopy were included. Those with advanced or obviously protruded or depressed lesions, lesions larger than 1.0 cm and/or lesions which were not amenable to observation by zoom function were excluded. The endoscopist was required to give the real-time descriptions of surface pit patterns of the lesions, based on surface pattern classification of enhanced magnification endoscopy. According to previous reports, types I-III represent non-neoplastic lesions, and types IV-V represent neoplastic lesions. Diagnosis with M i-scan and biopsy was performed before histopathological diagnosis. Magnified images of gastric lesions with and without enhancement were collected for further analysis. The diagnostic yield of real-time M i-scan and effects on magnification image quality by tone enhancement (TE), surface enhancement (SE) and color enhancement (CE) were calculated. The selected images were sent to another endoscopist. The endoscopist rated the image quality of each lesion at 3 levels. Ratings of image quality were based on visualization of pit pattern, vessel and demarcation line.
RESULTS: One hundred and eighty-three patients were recruited. Five patients were excluded for advanced gastric lesions, 1 patient was excluded for poor preparation and 2 patients were excluded for superficial lesions larger than 1.0 cm; 132 patients were excluded for no lesions found by high definition endoscopy. In the end, 43 patients with 43 lesions were included. Histopathology revealed 10 inflammation, 14 atrophy, 10 metaplasia, 1 low grade dysplasia (LGD), 5 high grade dysplasia (HGD) and 3 cancers. For 7 lesions classified into type I, histopathology revealed 6 atrophy and 1 metaplasia; for 10 lesions classified into type II, histopathology revealed 2 inflammation, 7 atrophy and 1 metaplasia; for 10 lesions classified into type III, histopathology revealed 1 inflammation, 8 metaplasia and 1 LGD; for 9 lesions classified into type IV, histopathology revealed 4 inflammation, 1 atrophy and 4 HGD; for 7 lesions classified into type V, histopathology revealed 3 inflammation, 1 HGD and 3 cancers. A total of 172 still images, including 43 images by white light (MWL) and 129 images by M i-scan (43 with TE, 43 with SE and 43 with CE), were selected and sent to the endoscopist who did the analysis. General image quality of M i-scan with TE and SE was significantly better than that of MWL (TE, 4.55 ± 1.07; SE, 4.30 ± 1.02; MWL, 3.25 ± 0.99; P < 0.001). Visualization of pit pattern was significantly improved by M i-scan with SE (1.93 ± 0.25 vs 1.50 ± 0.50, P < 0.001). Microvessel visualization was significantly improved by M i-scan with TE (1.23 ± 0.78 vs 0.76 ± 0.73, P < 0.001). Demarcation line visualization was improved by M i-scan with both TE and SE (TE, 1.75 ± 0.52; SE, 1.56 ± 0.59; MWL, 0.98 ± 0.44; P < 0.001). M i-scan with CE did not show any significant improvements of image quality in general or in the 3 key parameters. Although M i-scan with TE and SE slightly increased the diagnostic yield of MWL, there was no significant difference (P > 0.1).
CONCLUSION: Although digital enhancement improves the image quality of magnification endoscopy, its value in improving the diagnostic yield seems to be limited.
Core tip: In this study, the authors applied a new endoscopic device combining magnification endoscopy and virtual chromoendoscopy, equipped with surface enhancement, tone enhancement and color enhancement (M i-scan), in the diagnosis of 43 patients with small superficial gastric lesions. The results showed that real-time diagnosis of the gastric cancerous lesions by using M i-scan corresponded well with their histopathology. In comparisons between different enhancement capabilities using offline images, images with surface enhancement and tone enhancement were found to be slightly superior to those with color enhancement.