Method and apparatus for automated delineation of structure shape for image guided treatment planning

• US 9,122,959 B2
• Filed: 05/03/2013
• Issued: 09/01/2015
• Est. Priority Date: 05/03/2013
• Status: Active Grant

• Current Assignee: Elekta Incorporated (Elekta AB)
• Original Assignee: IMPAC Medical Systems, Inc. (Elekta AB)
• Inventors: Zhou, Yan, Han, Xiao, Hibbard, Lyndon S., Willcut, Virgil M.
• Primary Examiner(s): Cunningham, Gregory F
• Application Number:
  US13/886,853
• Publication Number:
  US 20140328527A1
• Time in Patent Office: 851 Days
• Field of Search: 382/190
• US Class Current: 1/1
• CPC Class Codes: G06T 2207/10072 Tomographic images
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 2207/20128 Atlas-based segmentation
• CPC Class Codes: G06T 2207/30081 Prostate
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/149 involving deformable models, e.g. active contour models

Abstract

Disclosed herein are techniques for performing automated structure delineation on image data using trained landmark detectors and a shape refinement tool. The landmark detectors can be trained to detect a landmark in the image data based on image features that are indicative of intensity variations over a plurality of windows of the image data points. A machine-learning algorithm can be used to train the landmark detectors. The landmarks in the image data that are detected by the trained landmark detects can be used to initialize an iterative shape refinement to thereby compute a refined shape estimate for a structure of interest such as a prostate.

Claims

1. An apparatus for delineating a structure of interest within image data, the image data comprising a subject image of a region of interest, the region of interest including the structure of interest, the image comprising a plurality of data points, the data points comprising a plurality of intensity values, the apparatus comprising:

a processor configured to;

compute a plurality of features for a plurality of the data points, the features being indicative of intensity variations over a plurality of windows of the data points,

detect a plurality of locations for a plurality of landmarks within the image based on an application of the computed features to a trained landmark detector,

generate a shape estimate for the structure of interest based on the detected landmark locations, and

iteratively refine, based on a sparse shape model from a shape dictionary, the shape estimate for the structure of interest.

2. The apparatus of claim 1, wherein the trained landmark detector comprises a plurality of trained landmark detectors, each trained landmark detector being configured to detect a different one of the landmarks.

3. The apparatus of claim 1, wherein the processor is further configured to process atlas image data using a machine-learning algorithm to train the trained landmark detector, wherein the atlas image data includes an identification of where a landmark is located in a plurality of atlas images.

4. The apparatus of claim 3, wherein the machine-learning algorithm comprises a LogitBoost machine-learning algorithm.

5. The apparatus of claim 3, wherein the atlas images comprise three-dimensional (3D) atlas images, and wherein the processor is further configured to process the atlas image data processing step by:

extracting a plurality of two-dimensional (2D) atlas image slices from the 3D atlas images,

aligning the 2D atlas image slices according to an image registration technique,

collecting a plurality of positive samples and a plurality of negative samples from each of a plurality of the atlas image slices,

computing the features for the collected positive and negative samples for each of a plurality of the atlas image slices, and

applying the collected positive and negative samples and the features computed therefor to the machine-learning algorithm to train a landmark detector to detect a landmark of interest.

6. The apparatus of claim 1, wherein the processor is further configured to perform the detection operation by:

applying the computed features to the trained landmark detector to score a plurality of candidate data points,

comparing the scores with a threshold, and

classifying the candidate data points for eligibility with respect to a landmark of interest based on the comparison.

7. The apparatus of claim 6, wherein the processor is further configured to perform the detection operation by generating a probability map for a plurality of locations for the landmarks other than the landmark of interest based on a location for a candidate data point classified as the landmark of interest and a landmark distribution model.

8. The apparatus of claim 7, wherein the processor is further configured to perform the detection operation by determining which locations on the probability map score above another threshold, and determining the landmark locations by averaging the locations from the probability map that score above the another threshold.

9. The apparatus of claim 6, wherein the processor is further configured to select a subset of the data points, and limit the applying operation to the selected subset of data points.

10. The apparatus of claim 1, wherein the refined shape estimate comprises a contour of the structure of interest, and wherein the processor is further configured to define a radiotherapy treatment plan for the structure of interest based on the contour.

11. The apparatus of claim 1, wherein the image comprises a computed tomography image.

12. The apparatus of claim 11, wherein the structure of interest comprises a prostate.

13. An apparatus for training a landmark detector using a plurality of atlas images, the atlas images including location information for a landmark with respect to a structure of interest, the apparatus comprising:

a processor configured to;

collect a plurality of positive samples and a plurality of negative samples from the atlas images,

compute a plurality of sample features for the collected positive and negative samples, wherein the sample features indicate intensity variations over a plurality of the positive and negative samples, and

apply the computed sample features and location data associated with the computed sample features to a machine-learning algorithm to train a landmark detector to detect the landmark.

14. An apparatus delineating a structure of interest within image data, the apparatus comprising:

a processor configured to;

detect a plurality of landmarks with respect to a structure in an image;

generate, for each detected landmark, at least one probability map mapping a location probability for each other detected landmark;

determine, based on the probability maps, the location of the detected landmarks; and

generate, based on the determined landmark locations, a shape estimate for the structure.

15. The apparatus of claim 14, wherein the processor is further configured to refine the shape estimate according to a shape refinement tool to thereby compute a refined shape estimate for the structure of interest, wherein the shape refinement tool comprises a trained boundary detector, and wherein the processor is further configured to perform the refinement operation by estimating boundaries for the shape estimate using the trained boundary detector.

16. The apparatus of claim 15, wherein the processor is further configured to process atlas image data using a machine-learning algorithm to train the boundary detector, wherein the atlas image data includes an identification of where a boundary for the structure of interest is located in a plurality of atlas images.

17. The apparatus of claim 16, wherein the machine-learning algorithm comprises a random forests machine-learning algorithm.

18. The apparatus of claim 15, wherein the shape refinement tool further comprises a shape dictionary, and wherein the processor is further configured to perform the refinement operation iteratively refining the shape estimate based on a sparse shape model.

19. The apparatus of claim 18, wherein the processor is further configured to online update the shape dictionary based on the refined shape estimate for the structure of interest.

20. The apparatus of claim 14, wherein each probability map is defined according to a distribution model for the respective landmark.

21. The apparatus of claim 14, wherein the processor is further configured to:

detect a plurality of additional landmarks;

generate, for each of the detected additional landmarks, at least one probability map mapping a location probability for each other detected additional landmark;

determine, based on the probability maps for the additional landmarks, the location of the additional landmarks; and

refine, based on the determined additional landmark locations, the shape estimate for the structure.