Kernel Based Fuzzy C-Means Clustering for Chronic Sinusitis Classification

9th Annual Basic Science International Conference 2019

自然科学（总论）

Putri, Rezki Aulia^1 ; Rustam, Zuherman^1 ; Pandelaki, Jacub^2

Department of Mathematics, Faculty of Mathmatics and Natural Sciences, University of Indonesia, Kampus UI Depok, Depok

16424, Indonesia^1

Department of Radiology, Cipto Mangunkusumo National General Hospital, DKI, Jakarta

10430, Indonesia^2

关键词: Chronic sinusitis;    Clustering techniques;    Euclidean distance;    Fuzzy C-means algorithms;    Fuzzy c-means clustering algorithms;    General hospitals;    High dimensional spaces;    Kernel-based fuzzy c-means clustering;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/546/5/052060/pdf
DOI  :  10.1088/1757-899X/546/5/052060

学科分类：自然科学（综合）

来源: IOP

【 摘 要 】

Sinusitis is an inflammation of the sinus wall, a small cavity interconnected through the airways in the skull bones. It is located on the back of the forehead, inside the cheek bone structure, on both sides of the nose, and behind the eyes. Chronic sinusitis is caused by infection, growth of nasal polyps, or irregularities of the nasal septum. This condition can affect teenagers, adults, and even children. To classify sinusitis we use Kernel Based Fuzzy C-Means (FCM) Clustering Algorithm, which is the development of Fuzzy C-Means (FCM) Algorithm. FCM is one of the widely used clustering technique. FCM algorithm comprises of sample points used to make whole and sub vector spaces according to the size of the distance. However, when non-linear data is separated, the convergence is inaccurate and slow. To overcome this problem, a Kernel-Based Fuzzy C-Means algorithm that makes use of kernel functions as a substitute for Euclidean distance utilized. It maps out samples to high-dimensional space to increase the differences between cluster centres, so they can overcome FCM deficiencies and improve linear machine capabilities. Data was obtained from the laboratory of Radiology at Cipto Mangunkusumo National General Hospital, Indonesia, with a 100% accuracy.