Home Health Computational Anatomy Based on Whole Body Imaging: Basic Principles of Computer-Assisted Diagnosis and Therapy
Lung Nodule Detection
This section discusses CAD of lung cancer on CT images. A promising application of computational anatomical models is the automatic detection of pulmonary nodules as lung cancer candidates. Another application is to suggest malignancy or benignity of nodules to support radiologists/physicians in diagnosing the disease. Other applications, including characterization of time-interval changes of pulmonary nodules using follow-up CT images, and prognosis generation, are important topics in the implementation of CAD in lung cancer management.
Lung cancer is the most common cause of cancer death worldwide [8, 9] . CT is the modality of choice for lung imaging. There is abundant evidence that CT screening with a low-dose imaging protocol improves sensitivity for identification of pulmonary nodules compared with plain radiography [10-12]. Research results from the National Lung Screening Trial (NLST) revealed that screening for lung nodules with low-dose CT (LDCT) reduced lung cancer mortality in heavy smokers by 20% compared with plain chest radiography .
Recently, results of two lung cancer screening studies showed the ability of CT to differentiate malignant from benign nodules before invasive biopsy procedures are considered [14, 15]. In a study by Aberle et al., an analysis of NLST results found that by the third annual screening, radiologist can distinguish malignant from benign nodules detected on earlier screening rounds based on change over time . Since even benign nodules can increase in size, the quantitative tools in CAD system might assist radiologists to assess the growth ratio of nodules in annual screening. McWilliams et al. found that analyzing patient and nodule characteristics can be used to estimate the malignant potential of pulmonary nodules detected on baseline LDCT screening . The authors created two prediction models to determine whether a nodule detected on the first CT scan was cancerous or not. One model consisted of the limited malignancy predictors that were significant, and the other included additional variables thought to be associated with a higher risk of malignancy. Analysis of nodule characteristics included nodule type (solid, nonsolid, or part solid), whether nodule margins were smooth or spiculated, nodule location in the lung, and presence of visible emphysema. Analyses of these nodule characteristics were carried out by visual assessment. They concluded that predictors in the full model included older age, female sex, family history of lung cancer, emphysema, larger nodule size, location of the nodule in the upper lobe, part-solid nodule type, lower nodule count, and spiculation. The readers who are interested in the management of solitary and multiple pulmonary nodules can refer the statement from the Fleischner Society . Subjective radiologic descriptors of pulmonary nodules are replaced by quantitative metrics that enable statistical comparisons between features and clinical outcomes; computer-aided diagnosis for lung nodules (CADx) has become one of the most active research areas as well as CADe [17-25].
|< Prev||CONTENTS||Next >|