Elimination of clavicle shadows to help automatic lung nodule detection on chest radiographs

Lung nodule detection is one of the most important goals of large scale screening of chest radiographs. The success of nodule detection can be increased if it is possible to suppress the bony structures from the chest radiographs. While one possible way to do this is to use dual-energy imaging, most of the commercial X-ray machines do not offer this technology. Finding an alternative approach is an important task. This paper proposes a new solution for it, where first the contours of bones -especially the contours of clavicles- are detected then the shadow of the bones is removed. The performance of the clavicle suppression algorithm is then evaluated by the use of a nodule detection system. The algorithms were tested on images from the JSRT database. The nodule detection algorithm was looking for up to 20 suspicious areas in the original images. According to our first results in the images with suppressed clavicle shadows 20 of 380 false positives were removed, while all true positives were preserved. 69 of the original 380 false positives were around the clavicle, which would mean 24% decrease in the concerned region.

[1]  A. Macovski,et al.  Generalized image combinations in dual KVP digital radiography. , 1981, Medical physics.

[2]  Elena Casiraghi,et al.  A Fully Automated Method for Lung Nodule Detection From Postero-Anterior Chest Radiographs , 2006, IEEE Transactions on Medical Imaging.

[3]  Hidefumi Kobatake,et al.  Detection of cancerous tumors on chest X-ray images -candidate detection filter and its evaluation , 1999, Proceedings 1999 International Conference on Image Processing (Cat. 99CH36348).

[4]  Terry E. Weymouth,et al.  Using Dynamic Programming For Minimizing The Energy Of Active Contours In The Presence Of Hard Constraints , 1988, [1988 Proceedings] Second International Conference on Computer Vision.

[5]  Carlos S. Pereira,et al.  Evaluation of Contrast Enhancement Filters for Lung Nodule Detection , 2007, ICIAR.

[6]  Demetri Terzopoulos,et al.  Snakes: Active contour models , 2004, International Journal of Computer Vision.