Sclerotic rib metastases detection on routine CT images

This work presents a computer-aided detection (CAD) system to aid radiologists in finding sclerotic bone metastases in the ribs on routine body imaging CT protocols with 5mm chest/abdomen CT images. First, the spine is segmented to locate the ribs using thresholding, region growing and a vertebra template. The centerlines of ribs are then traced by a progressive algorithm based on maximizing the cross-section overlap. The profile of the ribs along the centerline is analyzed to generate the initial detections. A set of quantitative features is then extracted and sent to a support vector machine for classification. Novel visualization techniques are also developed to display all ribs and the distribution of bone lesions in one single view. The system was tested on 10 data sets with 185 sclerotic lesions identified by an expert. A leave-one-out cross-validation results in 75.4% sensitivity at an average of 5.6 false positives per case.