Fuzzy logic based algorithm to classify tumor categories with position from brain MRI images

• Published in: EICT 2017 : 3rd International Conference on Electrical Information and Communication Technology : 7-9 December 2017 pp. 1 - 6
• Main Authors: Ahmmed, Rasel, Rahman, Md Asadur, Hossain, Md Foisal
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2017
• Subjects: Classification algorithms; Clustering algorithms; defuzzified values; Feature extraction; Fuzzy logic; fuzzy logic system; Image segmentation; Magnetic resonance imaging; Magnetic resonance imaging (MRI); temper based k-means and modified fuzzy c-means clustering (TKFCM); Tumors
• ISBN: 9781538623053; 1538623056
• DOI: 10.1109/EICT.2017.8275232

Early tumor detection is a vital issue and tumor position, tumor area, and tumor categories evaluation are also mandatory concerns for the proper medication. This paper proposes fuzzy logic based tumor classification method which can identify the tumor position as well. A database is prepared with normal and tumor affected brain MRI images. The integration of Temper based K-means and modified Fuzzy C-means (TKFCM) clustering algorithm is used to segment the MRI images regarding gray level intensity in small portion of brain images. The values of K which signifies the number of classifying contour in Temper based K-means algorithm is more than the conventional one and automatically updated membership of FCM eradicates the contouring problem between these two methods. Then, from the segmented images two types of features, i.e., first order statistic features and region property based features are extracted. The first features are used to detect and isolate tumor and second kind features are used to design Fuzzy expert logic with 93 rules to classify the tumor. In this process three inputs and one output variable are used with several membership functions to obtain the six categories of tumor. The orientation of the tumor provides position of tumors. The performance parameters of the proposed algorithm show substantial results which are effective in classifying tumors in multiple intensity based brain MRI image.