False positive reduction in lung nodule computer-aided detection based on 3D ranklet transform

摘要

The purpose of this study is to develop a technique for reducingudthe number of false positives affecting lung noduleudcomputer–aided detection in computed tomography (CT) images.udContiguous 2D regions of interest found on segmentedudlung areas from sections of a CT scan are merged to formudvolumes of interest (VOIs). Feature vectors are then computedudby submitting each VOI to the 3D ranklet transform,udi.e., a non–parametric, orientation–selective and multi–resolutionudtransform developed and evaluated herein. Finally, audsupport vector machine classifier is used to discriminate VOIsudcontaining nodules from those containing normal tissue. Theudproposed approach is evaluated on data consisting of 25 nodulesudmarked by experienced thoracic radiologists and 1048udnon–nodules randomly selected within the segmented lungudvolume of healthy patients. By achieving 96% of sensitivityudat 1% of false positive fraction, leave–one–out performancesudseem to be promising.