The role of computer-assisted analysis in the evaluation of nuclear characteristics for the diagnosis of precancerous and cancerous lesions by contact laryngoscopy

Abstract

Contact endoscopy (CE) through the direct contact with the surface of the mucosa enables in vivo visualization of upper epithelial layers. There is a broad spectrum of laryngeal pathologies, as has been confirmed by earlier CE reports. The aim of the study was to resolve some of the limitations of CE through the application of computer-assisted image analysis. Quantitative and qualitative evaluation of nuclei was applied in the diagnosis of precancerous and cancerous lesions. Fifty four patients with various laryngeal pathologies were included in the study. Paraffin section histopathology showed 15 benign lesions, 12 precancerous lesions (5 mild and 7 severe dysplasias) and 27 invasive squamous cell cancers (SCC). After staining the mucous with 1% methylen blue, examination with contact endoscope (Karl Storz, Germany) connected to the C-7070 Wide Zoom Olympus high-resolution camera was performed. The most discriminative parameters were revealed to be as follows: nucleus area (p<0.001), nuclei density index (p<0.001), elongation coefficient (p<0.05), nucleus area to equivalent area ratio (p<0.05). Computer-assisted image analysis composed with data mining techniques is presented for nuclei categorization. We established that computer-aided image analysis can indicate, with a high level of reliability, cases of severe dysplasia and carcinoma. By implementing the technique described in this paper, we can substantially increase the sensitivity of CE.