Pulmonary Vessel Area Extraction with Application to Reduce False Positive of Nodule Detection in 3D Chest CT Data(International Forum on Medical Imaging in Asia 2009 (IFMIA 2009))

概要

論文の詳細を見る
In this paper, we propose a novel approach to extract pulmonary vessels. Also we present its application to reduce the number of false positive (FP) regions by subtracting vessel regions from nodule candidate regions in an existing nodule detection method. We use a line structure enhancement filter and a blob structure enhancement filter based on Hessian matrix to emphasize the vessel-like regions and blob-like regions, respectively. Their output values are used to construct a speed function of the Level-set method. The fast marching algorithm, which is one of Level-set segmentation methods, uses this speed function to extracts the blood vessel region with suppressing extraction of the nodules by propagating its front to vessel-like regions. A quantitative performance evaluation of the vessel extraction method was performed by reduction rate of FPs. The result shows the proposed method can reduce the numbers of false positives occurred in our previous nodule detection method by 80%.
社団法人電子情報通信学会の論文
2009-01-12