Delineating low-count defective-contour SPECT lung scans for PE diagnosis using adaptive dual exponential thresholding and active contours

✍ Scribed by Alex Wang; Hong Yan

• Book ID: 102866031
• Publisher: John Wiley and Sons
• Year: 2010
• Tongue: English
• Weight: 515 KB
• Volume: 20
• Category: Article
• ISSN: 0899-9457
• DOI: 10.1002/ima.20222

✦ Synopsis

Abstract

To develop an automated lung delineation method that greatly improves the accuracy of contour extraction in SPECT lung scans with low maximum count, total counts, and/or defective contour edges, by combining deformable contour and planar image processing techniques with our existing adaptive dual exponential thresholding method. Two datasets consisting of 90 Monte Carlo simulations and 35 SPECT scans of real patients both with low maximum/total counts were used as the basis of this study. After preliminary contour extraction using dual exponential thresholding, a deformable contour algorithm was applied, combined with edge enhancement using image processing techniques such as Gaussian transformation and Sobel edge detection. Both quantitative validation via statistical measures and qualitative verification by an experienced practitioner were implemented to evaluate the method. An average of 83% congruency was achieved for 45 simulations with maximum count values below 31. This is a significant 33% improvement on the results of our previous study using dual exponential thresholding alone. Overall congruence averaged 95% for the entire simulation dataset. With the set of 35 actual SPECT scans with relatively low maximum/total counts, congruency again averaged near 95% with a few outliers ranging between 75 and 90%. We have successfully combined active contour with adaptive dual exponential thresholding to delineate lung contours accurately for SPECT scans with low maximum/total counts and/or defective contours due to various cardiopulmonary disorders. © 2010 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 20, 149–154, 2010