Classification System Vocal Cords Disease Using Digital Image Processing

This paper presents the study of vocal cords disease classification using digital image processing. Stroboscopy and laryngoscopy are common tools used by doctors to observe the state of the vocal cords directly. The vocal cords disease can be characterized by characters change the shape of glottis contour on the vocal cords. There are six classifications of vocal cords, normal, paralysis, nodule, papilloma, cyst, and granuloma which present in this paper. Before the classification process, extraction image vocal cords to get the characteristics or information of objects in the image. In this study used feature descriptions with the Speeded Up Robust Features (SURF) algorithm and shape measurement to extract the glottis contour of the vocal cords that can be analyzed and classified. The process of measuring the glottis contour of vocal cords requires the vocal image in the binary image. To get binary image this study used a method to automatically obtain the glottis area segmentation without user initialization. The segmentation is mainly based on active contour, Chan-Vese algorithm. The results of this study can optimize glottis contour extraction and results of the classification training process using K-Nearest Neighbor get an accuracy of 96.7%.