Vocal Folds Disorders Detection and Classification in Endoscopic Narrow-Band Images

The diagnosis of vocal folds (VF) diseases is error- prone due to the large variety of diseases that can affect them. VF lesions can be divided in nodular, e.g. nodules, polyps and cysts, and diffuse, e.g. hyperplastic laryngitis and carcinoma. By endoscopic examination, the clinician traditionally evaluates the presence of macroscopic formations and mucosal vessels alteration. Endoscopic narrow-band imaging (NBI) has recently started to be employed since it provides enhanced vessels contrast as compared to classical white-light endoscopy. This work presents a preliminary study on the development of an automatic diagnostic tool based on the assessment of vocal cords symmetry in NBI images. The objective is to identify possible protruding mass lesions on which subsequent vessels analysis may be performed. The method proposed here is based on the segmentation of the glottal area (GA) from the endoscopic images, based on which the right and the left portions of the vocal folds are detected and analyzed for the detection of protruding areas. The obtained information is then used to classify the VF edges as healthy or pathological. Results from the analysis of 22 endoscopic NBI images demonstrated that the proposed algorithm is robust and effective, providing a 100% success rate in the classification of VF edges as healthy or pathological. Such results support the investment in further research to expand and improve the algorithm presented here, potentially with the addition of vessels analysis to determine the pathological classification of detected protruding areas.