Computer aided diagnosis of lung ground glass opacity nodules and large lung cancers in CT

摘要: Diagnosis of lung nodules and cancers is a critical urgent problem in clinical diagnosis. This thesis to design build computer aided ground glass opacity (GGO) large diagnosis system which aims quantify the volumetric change GGO between pre-treatment post-treatment. In order over time, we need segmentation registration methods determine same nodule or cancer posttreatment, as well detection methods. For method, segmented pulmonary tubular objects will act landmarks. We extract centerlines 3D using improved ridge-based for with fully automatic bifurcation points. The points ensures continuity objects. Since medical images contain anatomical structures various shapes, first perform pre-selection method identify region containing by applying intensity ridge tracing method. These steps are based on eigenanalysis Hessian matrix, provides an estimation elongated direction cross-sectional planes orthogonal While objects, automatically detected from applying scan-conversion Adaboost algorithm specially designed steerable filters. For develop 3D-3D model rigid organs radius both planning respiration-correlated CT (RCCT) images. detects filters. then apply minimizes least square error corresponding For segmentation, propose novel detect segment chest detection, classifier boosting k-Nearest Neighbor, whose distance measure Euclidean nonparametric density estimates two regions. clustering regions nodules. analyzing texture likelihood map region. also present statistical validation proposed (10 datasets contains 10 nodules)…