Title :
A computer method for synthesizing lung nodules
Author :
Dajnowiec, Maciej ; Alirezaie, Javad
Author_Institution :
Dept. of Electr. & Comput. Eng., Ryerson Univ., Toronto, Ont., Canada
Abstract :
The presented technique is the first step to developing a program fully capable of producing databases of 2D, single slice, synthetic lung nodules. The synthetic lung nodules produced by this program are based on a 2D Gaussian structure. This is modeled on the study of the structure of real lung nodules. Some limited information from other research with regards to nodule structure is presented, and is a valuable tool in the future development of the program. A synthetic nodule is evaluated from both its appearance in a computed tomography (CT) image and from a 3D mesh plot. The 2D image has a similar intensity and size to the real nodule to which it is compared. The 3D mesh also has a close resemblance to the real nodule. Based on this comparison, the Gaussian structure appears to work well for modeling the nodule. Further development of the program is intended to allow it to handle different digital imaging and communications in medicine (DICOM) image formats, improve the synthetic lung nodule structure and more realistically insert the nodules into sets of CT image slices with different thicknesses rather than just individual images.
Keywords :
Gaussian processes; cancer; computerised tomography; image reconstruction; lung; medical image processing; 2D Gaussian structure; 3D mesh plot; CT image; computed tomography image; digital imaging and communications in medicine; image formats; lung cancer; nodule structure; synthetic lung nodules; Cancer detection; Computed tomography; Data engineering; Design engineering; Diagnostic radiography; Image segmentation; Java; Lungs; Robustness; Testing;
Conference_Titel :
Electrical and Computer Engineering, 2004. Canadian Conference on
Print_ISBN :
0-7803-8253-6
DOI :
10.1109/CCECE.2004.1349643