COMPUTER-ASSISTED CLASSIFICATION SYSTEM AND METHOD FOR LOCAL LUNG STRUCTURE FOR DETECTING SMALL NODULE

摘要

PPROBLEM TO BE SOLVED: To reduce false positive numbers in a classification system and method for local lung structure with higher flexibility and lower calculation cost. PSOLUTION: In the method to classify the lung structure on a digital image, a step to provide an approximate objective structure location of one or more objective structures in the 3D digitized image, a step to create center positions of a more detailed 3D objective model and those of one or more objective structure by applying anisotropic gaussian model for the approximate objective structure location, a step to transform each 3D objective model to a 3D sphere, a step to comprise a boundary manifold for each transformed 3D sphere, and a step to classify one or more objective structure identifying the cluster on the boundary manifold. PCOPYRIGHT: (C)2007,JPO&INPIT