Positron Emission Tomography scan images are extensively used in radiotherapyplanning, clinical diagnosis, assessment of growth and treatment of a tumor.These all rely on fidelity and speed of detection and delineation algorithm.Despite intensive research, segmentation remained a challenging problem due tothe diverse image content, resolution, shape, and noise. This paper presents afast positron emission tomography tumor segmentation method in whichsuperpixels are extracted first from the input image. Principal componentanalysis is then applied on the superpixels and also on their average. Distancevector of each superpixel from the average is computed in principal componentscoordinate system. Finally, k-means clustering is applied on distance vector torecognize tumor and non-tumor superpixels. The proposed approach is implementedin MATLAB 2016 which resulted in an average Dice similarity of 84.2% on thedataset. Additionally, a very fast execution time was achieved as the number ofsuperpixels and the size of distance vector on which clustering was done wasvery small compared to the number of raw pixels in dataset images.
展开▼
