Ground glass opacity lesion morphology extraction in primary lung cancer

Abstract

In determining the level of tumour malignancy in lung cancer, several characteristics of lesion in the lungs need to be recognised. The characteristics include several components, namely tumour size, enhancement, irregular spiculated edge, lobulated, air bronchograms, ground glass opacity (GGO) and density. This study identifies GGO lesion characteristics using CT image datasets obtained from Sardjito Public Hospital, Indonesia. The initial stage conducted is a cropping process performed by a radiologist so that the research's focus is merely on the lesion. The next process is the feature extraction by using grey level co-occurrence matrices (GLCM) with four features, namely energy, contrast, correlation and homogeneity. The classification stage is carried out after the extraction stage which is followed by features selection. Having selected two most dominant features from total of 16 features, the proposed method achieves accuracy of 88.8%, sensitivity of 87.5% and specificity of 90%.