Lung nodule segmentation and estimation using thresholding on local thickness

Abstract

A simple-yet-effective method, utilizing local thickness and auto-thresholding for manual lung nodule segmentation and estimation is presented. By utilizing local thickness algorithm, we can achieve the field of local thickness for the respective region of interest, which can be used for further assessment in thresholding. The auto-thresholding technique, IsoData, is used on local thickness field for segmenting lung nodule, which can remove blood vessels or features that are difficult to be removed by conventional mean and must be excluded in order to ensure correct segmentation and measurement. By using correlation comparison on the same lung nodule dataset between our proposed method and Philips Extended Brilliance Workspace, we can see that the correlation is 98.9%.