Detection of Lung Cancer with the Fusion of Computed and Positron Emmision Tomography

Abstract

Computed Tomography (CT) is an imaging technique which gives the structural information of the internal parts of the body. Positron Emission Tomography(PET) gives the functional information of internal parts of body. Fusion of CT scan and PET scan is mainly done by machines in many hospitals. As the installation machinery is costly, all hospitals may not afford it. Wavelet based fusion is the method to obtain the fused PET/CT without machines. Database has been collected from PGIMER, Chandigarh. Database included 70 abnormal and 100 normal images of both PET and CT scans. Fusion of CT scan and PET scan of lung images and classification of PET/CT images have been done in this dissertation. Fusion has been done by wavelet transform (using haar as a mother wavelet) and PET/CT images have been obtained which have higher sensitivity & specificity. This technique of fusion is cheaper than the machine fusion and also it is easy to use as compared to machines. Segmentation of ROIs has been done to segment normal and abnormal ROIs by manual segmentation technique. Further, features have been calculated for both normal and abnormal PET/CT images by Gray Level Co-occurrence Matrix (GLCM) technique. PET/CT images have been classified using support vector machine. The results of classification shows that the accuracy of PET/CT is greater than the accuracy of CT images. The accuracy comes out to be 98.2%, which is 3.2% greater than the accuracy of CT images.