DETEKSI KANKER PARU MENGGUNAKAN SEGMENTASI CITRA CT- SCAN DENGAN ALGORITMA KLUSTER K-MEANS

Abstract

Imaging  technology  such  as  Computed  Tomography  (CT)  plays  an  important  role  in  producing high-resolution images.  CT makes it easy for an oncologist to diagnose patients for malignant and benign  tumors.  However,  the  technology  requires  image  engineering  techniques  to  make  the diagnosis  more  accurate.  The  technique  to  draw  that  is  segmentation.  Segmentation  plays  an important  role  to  distinguish  between  healthy  tissue,  benign  and  malignant  tumors.  There  are segmentation techniques based on mass density (pixel), contour (color) etc. This article will review and  explain  the  development  of  segmentation  techniques  specific  to  lung  cancer  images. Thresholding  segmentation  method  with  k-means  clustering  using  a  dataset  from  The  Cancer Imaging Archive (TCIA) named SPIEAAPM Lung CT Challenge dataset is performed in this study. The image data is divided into a number of k clusters that are mutually exclusive of each other. In CT  images  the  value  of  k  is  considered  as  2  as  there  are  2  regions,  one  is  lung  and  other  is background. This k-means clustering process will be done in two phases, the first phase k clusters will  be  formed  by  taking  each  pixel  intensity  value  to  the  nearest  centroid  by  calculating  the distance  between  pixel  intensity  values  to  each  centroid  using  different  distance  calculation methods.  The  segmentation  results  show  that  the  accuracy,  sensitivity  is  achieved  very  well  to differentiate and pinpoint the clusters.