DISCUSSION
Image enhancement systems increase the contrast of the blood vessels for improved visibility of the submucosal vasculatures. Using the NBI enhancement system, the ELS classified the submucosal vascular patterns in laryngeal lesions as longitudinal and perpendicular, with an association of the perpendicular pattern with malignancy, and the longitudinal pattern with benign conditions. In the present study, the vascular patterns reproduced by Image1 S endoscopy in benign and malignant lesions were highly consistent with those defined in the ELS classification. The prediction of the pathological diagnosis strongly correlated with the definitive pathological diagnosis of the lesions. Moreover, the evaluations of the patterns were highly consistent among the evaluators, irrespective of their work history, or their previous experience with image enhancement systems. These consistencies stated above, are directly related to improved visibility of the vasculatures by contrast enhancement. Therefore, Image1 S is considered to be an effective and practical technique for evaluation of the vascular characteristics of laryngeal lesions to predict malignancy.
Stanikova et al. used both NBI and the Image1 S system to investigate the relationship between the vascular patterns and the pathological diagnosis of laryngeal lesions (7). They found the sensitivity and specificity values for NBI endoscopy of 83% and 98% respectively, whereas those values for Image1 S were 86% and 96%, respectively. In diagnostic use, the high specificity of the test indicated a strong association of benign lesions with the longitudinal pattern and a weak association with the perpendicular pattern, suggesting an ability to detect the patterns in benign lesions. However, in our study, Image 1 endoscopy showed a higher sensitivity (96%) and a lower specificity (72%) values, compared to the previous study. In diagnostic use, the high sensitivity of the test is related to the improved detection of vascular patterns in dysplastic and malignant lesions, and it indicates a strong association between the perpendicular pattern and malignant lesions and a weak association between the longitudinal pattern and malignancy. Thus, malignant and dysplastic lesions were predicted with certainty with the Image1 S technique and, as a result, relatively few cases of cancer were missed in the endoscopic examination. Our study was conducted on a relatively small number of subjects; however, the sample size was sufficient because malignant and premalignant subjects were most prevalent in the whole sample. In addition, a highly sensitive test with many true positives requires a minimum sample size for statistically significant results. Image1 S endoscopy, therefore, proved to be a sensitive and reliable screening test for the prediction of malignancy in suspicious lesions presenting with a perpendicular vascular pattern.
The false positive and false negative cases and the cases excluded from the survey did not have any significant effect on the statistical parameters, but they are clinically important when screening to predict malignancy. The only false negative case was a low-grade dysplasia associated with a longitudinal pattern; this case may reflect the small extent of neo-angiogenesis in low grade dysplastic lesions (8). Among the false positive cases, one case with benign granular cell tumor and two cases with granulation tissue, were negatively associated with a perpendicular vascularization pattern. A perpendicular pattern in benign tumors has also been reported by Arens et al. for HPV-related papilloma. That study used contact endoscopy with NBI system and emphasized the differences in turning angle of the vessels within the vascular loop; where wide long turns in the papilloma differed from short, acute turns in terms of carcinogenesis (2). Distinction of these differences was not possible in our endoscopic examination, but coupling the image1 S system with contact endoscopy, which would result in a nearly 60-fold gain in magnification, may help to clarify these distinctions. Two other false positive cases with granulation tissue after laser surgery appeared to be related to the wound healing processes, which can make vascular evaluation difficult.
The cases excluded from the survey were those presenting with leukoplakia and granulation tissue, and they featured with surface characteristics that limited the visualization of the vasculatures. The limitation in the leukoplakia case was the umbrella effect of keratosis (9). Likewise, the vasculatures in the granulation tissue cases were obscured due to the exudative material that covered the surface. For these reasons, endoscopic image enhancement techniques may have a limited value in keratotic and exudative lesions. Nevertheless, a high rate of accuracy with many true positives and true negatives was achieved in our study using Image1 S enhancement, apart from the few exceptions mentioned above.
This study also evaluated the practical utility of the ELS concept of vascular patterns, by conducting an electronic survey, on a group of otolaryngologists with a varying degrees of work experience. This part of the study had two major limitations: one was that digital images provide less information than actual video-endoscopic examination, and the other was the varying degree of knowledge among the evaluators regarding the ELS classification is variable. Despite these limitations, the survey results were consistent between the authors and the participants regardless of previous work experience and quality of the images, i.e., the vascular patterns were identified with high accuracy by all the evaluators. This study confirms the feasibility of using the ELS classification in routine endoscopic examination and for e-screening and e-learning purposes.
In the survey, we also questioned whether the spectral light images are preferred over the white light images for discrimination of the vascular patterns. Technically, white light provides sharpness and local brightness, whereas spectral modes provide the highest image contrast. The spectral modes have been used previously in several situations, including urinary bladder imaging, for improved visualization of the parathyroid gland in endoscopic thyroidectomy, during endonasal sinus surgery, and for treatment of oral cavity and oropharynx lesions in combination with contact endoscopy (10-13). Based on the empirical observations of the users in the present study, the spectral modes, which produced high contrast images, were apparently preferred over white light when examining laryngeal lesions.