GGN segmentation in pulmonary images for accuracy and consistency
AssigneeSiemens Medical Solutions USA, Inc.
InventorsLi Zhang, Tiantian Zhang
Life of this patent
Estimating the term…
Abstract
A method and system for ground glass nodule (GGN) segmentation. The invention comprises methods and systems for GGN segmentation using adaptive intensity models in conjunction with Markov random field (MRF) based modeling. The adaptive intensity model overcomes problems associated with low-contrast GGN images. Vessels included by MRF segmentation are removed by shape analysis based on distance maps of the segmentation.
Claims22
A method for nodule segmentation comprising:
- selecting an image series group of image slices for a nodule;
- initializing segmentation comprising:
- selecting one image slice from the image series group where a nodule periphery is the most distinguishable as a center slice;
- for the center slice, calculating major and minor axes of an ellipse model by searching for image intensity peaks from intensity gradients representing the nodule;
- calculating mean and standard deviation intensity estimates from the ellipse model for beginning mean and standard deviation intensity estimate values;
- selecting an adaptive intensity model based on the beginning mean and standard deviation intensity estimate values;
- performing an Iterated Condition Mode (ICM) comprising:
- pre-segmenting the nodule on the center slice using the adaptive intensity model in conjunction with Markov Random Field (MRF) based modeling to calculate conditional intensity probability mean and standard deviation values for the nodule and background intensities;
- updating the adaptive intensity model with the conditional intensity probability mean and standard deviation values for the nodule and background intensities obtained from pre-segmentation;
- segmenting the nodule using the updated adaptive intensity model in conjunction with MRF based modeling to calculate conditional intensity probability mean and standard deviation values for the nodule and background intensities;
- updating the adaptive intensity model with the conditional intensity probability mean and standard deviation values for the nodule and background intensities obtained from segmentation;
- and
- calculating a growth rate measurement for the nodule.
The method according to 6 further comprising converting a plurality of the segmented image slices for the nodule image series group into a 3D rendering.
A system for nodule segmentation comprising:
- a processor and a storage device having embodied therein a program for execution by the processor, the program providing a segmentation framework configured to:
- select an image series group of image slices for a nodule;
- initializing segmentation comprising:
- selecting one image slice from the image series group where a nodule periphery is the most distinguishable as a center slice;
- for the center slice, calculating major and minor axes of an ellipse model by searching for image intensity peaks from intensity gradients representing the nodule;
- calculating mean and standard deviation intensity estimates from the ellipse model for beginning mean and standard deviation intensity estimate values;
- selecting an adaptive intensity model based on the beginning mean and standard deviation intensity estimate values;
- performing an Iterated Condition Mode (ICM) comprising:
- pre-segmenting the nodule on the center slice using the adaptive intensity model in conjunction with Markov Random Field (MRF) based modeling to calculate conditional intensity probability mean and standard deviation values for the nodule and background intensities;
- updating the adaptive intensity model with the conditional intensity probability mean and standard deviation values for the nodule and background intensities obtained from pre-segmentation;
- segmenting the nodule using the updated adaptive intensity model in conjunction with Markov Random Field (MRF) based modeling to calculate conditional intensity probability mean and standard deviation values for the nodule and background intensities;
- updating the adaptive intensity model with the conditional intensity probability values mean and standard deviation values for the nodule and background intensities obtained from segmentation;
- and
- calculating a growth rate measurement for the nodule.
The system according to 17 further comprising converting a plurality of the segmented image slices for the nodule image series group into a 3D rendering.
Dependencies are inferred from the claim text.