Extraction Morphology of Primary Lung Cancer using Gray Level Co-occurrence Matrices, Zernike Moment and roundness
ABSTRACT: Lung cancer is one of the cancer types that has a high mortality rate in the world. To reduce the mortality rate, early detection is necessary so that the patients can be treated as soon as possible. One of the detection processes is conducted by using screening that is Computed Tomography (CT) scan. By using CT images, it can be determined the level of malignancy of a lesion. Before determining the level of malignancy, it is necessary to recognise and analyse morphological characteristics of the lesion. These characteristics will be used to determine the level of malignancy and the level of the patient�s risk. Previous research did not discuss about the determination of the level of malignancy but some research has made the determination of one of the characteristics of the lesion morphology, it is Ground Glass Opacity. Therefore, this research aims to do an introduction to the morphological characteristics of the lesion in the lung cancer cases of primary lesion Ground Glass Opacity (GGO) and shape that can be used as a parameter in determining the severity of a lesion. Recognition process carried out in several phases: cropping, preprocessing, feature extraction, classification and feature selection. Based on the research results, the accuracy of the classification process is 83,33% with sensitivity 76,4% and specificity 89,4% for GGO lesion and accuracy obtained 86,7% with sensitivity 95% and specificity 70% for shape lesion. After that, feature selection stage is used to search the most significant features and obtained some features for each lesion. Feature selection results are then classified and obtained an increase in accuracy for GGO lesion is 88,8% with sensitivity 87,5% and specificity 90% and accuracy obtained for lesion shape is 93,3% with sensitivity 95% and specificity 90%.
ABSTRACT: Lung cancer is one of the cancer types that has a high mortality rate in the world. To reduce the mortality rate, early detection is necessary so that the patients can be treated as soon as possible. One of the detection processes is conducted by using screening that is Computed Tomography (CT) scan. By using CT images, it can be determined the level of malignancy of a lesion. Before determining the level of malignancy, it is necessary to recognise and analyse morphological characteristics of the lesion. These characteristics will be used to determine the level of malignancy and the level of the patient�s risk. Previous research did not discuss about the determination of the level of malignancy but some research has made the determination of one of the characteristics of the lesion morphology, it is Ground Glass Opacity. Therefore, this research aims to do an introduction to the morphological characteristics of the lesion in the lung cancer cases of primary lesion Ground Glass Opacity (GGO) and shape that can be used as a parameter in determining the severity of a lesion. Recognition process carried out in several phases: cropping, preprocessing, feature extraction, classification and feature selection. Based on the research results, the accuracy of the classification process is 83,33% with sensitivity 76,4% and specificity 89,4% for GGO lesion and accuracy obtained 86,7% with sensitivity 95% and specificity 70% for shape lesion. After that, feature selection stage is used to search the most significant features and obtained some features for each lesion. Feature selection results are then classified and obtained an increase in accuracy for GGO lesion is 88,8% with sensitivity 87,5% and specificity 90% and accuracy obtained for lesion shape is 93,3% with sensitivity 95% and specificity 90%.
