|
KMID : 1147120090150010001
|
|
Journal of the Korean Society of Imaging Informatics in Medicine
2009 Volume.15 No. 1 p.1 ~ p.7
|
|
|
Computer-Aided Diagnosis of Pure Nodular Ground-Glass Opacities in CT images: Classification of Benign and Malignant Lesions using K-means Algorithms
|
|
Son Woo-Ram
Park Sang-Jun
Park Chang-Min
Goo Jin-Mo
Kim Jong-Hyo
|
|
|
Abstract
|
|
|
|
Lung cancer is one of the most prevalent diseases in the world. Recently, proportion of pure nodular ground-glass opacity (PNGGO) has been reported to increase among for all CT-detected pulmonary nodules. Moreover, the malignancy rate of PNGGOs is a considerable proportion. Hence, accurate classification of benign or malignant PNGGO is an essential feature in lung cancer diagnosis of Computer-aided Diagnosis (CAD) in order to reduce the unnecessary biosies. In this reason, we are developing a CAD system to classify benign or malignant disease of PNGGOs from CT images. Here we present a classifier to reduce the inter- or intra-observer variation of the primary classification. Our system consists of following three steps. Segmentation of PNGGOs is performed by region-growing technique. After that, statistical features of segmented PNGGOs regions are extracted. Finally, Kmeans clustering algorithm as a classifier is applied. Experiment was performed employing 20 CT image sets and 0.83 (value of area under the ROC curve) was achieved.
|
|
|
KEYWORD
|
|
|
Lung cancer, ground-glass opacity, classification, k-means, computer-aided diagnosis
|
|
|
FullTexts / Linksout information
|
|
|
|
|
|
Listed journal information
|
|
|
|