Learning-based pulmonary nodule detection from multislice CT data

An automatic computer-aided detection system is developed for detecting pulmonary nodules from high resolution CT data. The system is based on the concept of machine learning. A rotation-invariant feature, Radial Volume Distribution (RVD), is proposed to extract the volume intensity distribution along the radial directions of the 3D volume sample. For computational reasons, a variant of RVD is provided as an approximation. The volume intensity is utilized and the 3D characteristics of the volume is incorporated to extract features of more discriminating power. Support Vector Machine is used as the classifier. The preliminary experimental results show the promising performance.

• Type: Conference
• Year: 2004
• Where: CARS
• Authors: Xiaoguang Lu, Guo-Qing Wei, Jian Zhong Qian, Anil K. Jain
• Keywords: CARS 2004, Medical Imaging, Radial Volume Distribution, Volume Intensity, Volume Intensity Distribution