Multi-scale nodule detection in chest radiographs

Abstract

Early detection is the most promising way to enhance a patient's chance for survival of lung cancer. In this work, a novel computer algorithm for nodule detection in chest radiographs is presented that takes into account the wide size range for lung nodules through the use of multi-scale image processing techniques. The method consists of: i) Lung field segmentation with an Active Shape Model [1]; ii) Nodule candidate detection by Lindeberg's multi-scale blob detector [2] and quadratic classification; iii) Blob segmentation by multi-scale edge focusing; iv) k Nearest neighbor classification. Experiments on the complete JSRT database [3] show that by accepting on average 2 false positives per image, 50.6% of all nodules are detected. For 10 false positives, this increases to 69.5%. © Springer-Verlag Berlin Heidelberg 2003.