XIV: Images 1
Journal of Japan Society of Computer Aided Surgery · 2014
This paper describes a method of restriction of lymph node (LN) existence using cascaded classifiers for automated lymph node detection from abdominal CT image.LN analysis is one of important parts of lymphadenectomy, which is a surgical procedure to remove one or several LNs in the tumor area.Therefore, it is important to detect LNs automatically from CT scan.We proposed a method based on RST analysis for automated LN detection.However, the computational time of RST analysis is very high due to complexity of RST computation.Therefore, this paper proposes a method of restriction of LN existence using cascaded classifiers, which consist of AdaBoost and SVM classifiers, in order to reduce computational time of automated LN detection.AdaBoost classifiers utilize 3D Haar-like features to extract a bounding box contains LN.After that the proposed method eliminates FPs using a SVM classifier based on RST analysis, average intensity and standard deviation of voxels in the bounding box.The experimental result revealed that the sensitivity of the proposed method was 86.2% with 55 FPs per case.