Segmentation of the lungs within magnetic resonance (MR) scans is a necessary preprocessing step in the computerized analysis of thoracic MR images. This task is complicated by potentially significant cardiac and pulmonary motion artifacts, partial volume effect, and morphological deformation from disease. We have developed an automated segmentation method to account for these complications. First, the thorax is segmented using a threshold obtained from analysis of the cumulative gray-level histogram constructed along a diagonal line through the center of the image. Next two separate lung-thresholded images are created. The first lung-thresholded image is created using histogram-based gray-level thresholding techniques applied to the segmented thorax. To include lung areas that may be adversely affected by artifact or disease, a second lung-thresholded image is created by applying a grayscale erosion operator to the first lung-thresholded image. After a rolling ball filter is applied to the lung contour to eliminate non-lung pixels from the thresholded lung regions, a logical OR operation is used to combine the two lung-thresholded images into the final segmented lung regions. Modifications to this approach were required to properly segment sections in the lung bases. In a preliminary evaluation, the automated method was applied to 10 MR scans, an observer evaluated the segmented lung regions using a five-point scale (“highly accurate segmentation” to “highly inaccurate segmentation”). Eighty-five percent of the segmented lung regions were rated as highly or moderately accurate.
[1]
M D Morgan,et al.
Detection and quantification of pulmonary emphysema by computed tomography: a window of opportunity.
,
1992,
Thorax.
[2]
J W Sayre,et al.
Knowledge-based segmentation of thoracic computed tomography images for assessment of split lung function.
,
2000,
Medical physics.
[3]
S. Armato,et al.
Automated detection of lung nodules in CT scans: preliminary results.
,
2001,
Medical physics.
[4]
E. Hoffman,et al.
Characterization of the interstitial lung diseases via density-based and texture-based analysis of computed tomography images of lung structure and function.
,
2003,
Academic radiology.
[5]
Michael F. McNitt-Gray,et al.
Method for segmenting chest CT image data using an anatomical model: preliminary results
,
1997,
IEEE Transactions on Medical Imaging.
[6]
W. Bradley,et al.
MRI: The Basics
,
1997
.
[7]
Scott T. Acton,et al.
Merging parametric active contours within homogeneous image regions for MRI-based lung segmentation
,
2003,
IEEE Transactions on Medical Imaging.
[8]
W. Kalender,et al.
Semiautomatic evaluation procedures for quantitative CT of the lung.
,
1991,
Journal of computer assisted tomography.
[9]
S. Armato,et al.
Automated lung segmentation for thoracic CT impact on computer-aided diagnosis.
,
2004,
Academic radiology.
[10]
Ian Middleton,et al.
Segmentation of magnetic resonance images using a combination of neural networks and active contour models.
,
2004,
Medical engineering & physics.