Akademik Veri Yönetim Sistemi
23rd IEEE International Workshop on Machine Learning for Signal Processing (MLSP), Southampton, Birleşik Krallık, 22 - 25 Eylül 2013
Physical examination of larynx by using videolaryngostroboscopy provides valuable information for diagnosis of vocal fold pathologies. However difficulty of differentiate nodules and cysts using clinical resources alone motivates physicians to research new strategies. In this paper, we propose a novel approach that performs nodule-cyst classification exploiting visible blood vessels on the superior surface of vocal folds. We first detected the region of vocal folds on videolaryngostroboscopy images and then extracted centerlines of vessel network on vocal folds. We used orientation pattern of vessels for classification. The performance of the proposed system was evaluated using laryngeal images of 21 patients. True positive rates of 76% and 74% were obtained for nodule and cyst classes respectively. These results indicate that visible vessels of vocal folds may play a critical role in more effective diagnosis of vocal fold pathologies like nodule and cyst which may be difficult to differentiate.