IMAGE FUSION BASED LUNG NODULE DETECTION USING STRUCTURAL SIMILARITY AND MAX RULE

Abstract

The uncontrollable cells in the lungs are the main cause of lung cancer that reduces the ability to breathe. In this study, fusion of Computed Tomography (CT) lung image and Positron Emission Tomography (PET) lung image using their structural similarity is presented. The fused image has more information compared to individual CT and PET lung images which helps radiologists to make decision quickly. Initially, the CT and PET images are divided into blocks of predefined size in an overlapping manner. The structural similarity between each block of CT and PET are computed for fusion. Image fusion is performed using a combination of structural similarity and MAX rule. If the structural similarity between CT and PET block is greater than a particular threshold, the MAX rule is applied; otherwise the pixel intensities in CT image are used. A simple thresholding approach is employed to detect the lung nodule from the fused image. The qualitative analyses show that the fusion approach provides more information with accurate detection of lung nodules.