Method and apparatus for detecting anatomical elements

• US 9,940,545 B2
• Filed: 09/20/2013
• Issued: 04/10/2018
• Est. Priority Date: 09/20/2013
• Status: Active Grant

• Current Assignee: Change Healthcare Holdings, LLC (UnitedHealth Group Incorporated)
• Original Assignee: Change Healthcare LLC (UnitedHealth Group Incorporated)
• Inventors: Rezaee, Mahmoud Ramze, Top, Andrew, Brown, Colin, Hamarneh, Ghassan
• Primary Examiner(s): Seth, Manav
• Application Number:
  US14/032,791
• Publication Number:
  US 20150086091A1
• Time in Patent Office: 1,663 Days
• CPC Class Codes: A61B 5/7264 Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems neural networks per se G06N3/00; expert systems per se G06N5/00
• CPC Class Codes: G06F 18/2323 based on graph theory, e.g. minimum spanning trees [MST] or graph cuts
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06F 18/24323 Tree-organised classifiers
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06V 10/446 using Haar-like filters, e.g. using integral image techniques
• CPC Class Codes: G06V 10/7635 based on graphs, e.g. graph cuts or spectral clustering
• CPC Class Codes: G06V 2201/03 Recognition of patterns in medical or anatomical images

Abstract

A method, apparatus and computer program product are hereby provided to detect anatomical elements in a medical image. In this regard, the method, apparatus, and computer program product may receive a test image and generate a classified image by applying an image classifier to the test image. The image classifier may include at least one decision tree for evaluating at least one pixel value of the test image and the classified image may include a plurality of pixel values. Each pixel value may be associated with a probability that an anatomical element is located at the pixel location. The method, apparatus, and computer program product may also evaluate the classified image using an anatomical model to detect at least one anatomical element within the classified image.

Claims

1. A method for detecting anatomical elements comprising:

receiving a test image;

generating one or more feature vectors, wherein generating each feature vector comprises convolving a respective feature with the test image;

generating a classified image by using the one or more feature vectors to apply an image classifier to the test image, the image classifier comprising at least one decision tree, wherein application of the image classifier to the test image comprises feeding the one or more feature vectors through the at least one decision tree to generate at least one pixel value of the test image, the classified image comprising a plurality of pixel values, wherein generation of each pixel value assigns the pixel value a probability that its associated pixel is related to an anatomical element; and

evaluating, using a processor, the classified image using an anatomical model to detect at least one anatomical element within the classified image.

2. The method of claim 1, further comprising labeling the detected at least one anatomical element within the test image in response to detecting the at least one anatomical element within the classified image.

3. The method of claim 2, wherein a location of the at least one anatomical element within the test image corresponds to a location in which the at least one anatomical element was detected within the classified image.

4. The method of claim 1, further comprising generating the image classifier by:

receiving a set of training images, the set of training images comprising at least one target image and at least one source image;

determining at least one image feature;

transforming the at least one source image using the at least one image feature to generate at least one feature image;

generating at least one decision tree corresponding to the at least one image feature using at least the at least one feature image and the at least one target image; and

using the generated at least one decision tree as the image classifier.

5. The method of claim 4, further comprising:

generating a plurality of decision trees, each of the decision trees corresponding to a set of image features;

evaluating the plurality of decision trees to determine an accuracy value for each decision tree; and

selecting at least one of the plurality of decision trees with the highest accuracy value as the image classifier.

6. The method of claim 4, wherein the at least one image feature is a Haar-like feature.

7. The method of claim 4, wherein the at least one decision tree is generated by a process comprising:

determining a set of pixel values associated with a particular node of the decision tree;

determining a feature that results in a minimum variance in target pixel values associated with the set of pixel values; and

assigning the feature that results in a minimum variance in target pixel values as a decision feature for the particular node of the tree.

8. The method of claim 7, further comprising:

determining a threshold feature value associated with the feature that results in the minimum variance, wherein the threshold feature value results in a split in the set of pixel values when applied to the set of pixel values; and

assigning the threshold feature value to the particular node of the decision tree.

9. The method of claim 1, further comprising generating the anatomical model by evaluating a set of anatomical data.

10. The method of claim 1, wherein the anatomical model defines at least one of a size of an anatomical element, a shape of an anatomical element, or an offset between two or more anatomical elements.

11. The method of claim 1, wherein the anatomical elements are spinal vertebrae, and wherein the anatomical model defines an offset between adjacent vertebrae.

12. An apparatus comprising processing circuitry configured to:

receive a test image;

generate one or more feature vectors, wherein generating each feature vector comprises convolving a respective feature with the test image;

generate a classified image by using the one or more feature vectors to apply an image classifier to the test image, the image classifier comprising at least one decision tree, wherein application of the image classifier to the test image comprises feeding the one or more feature vectors through the at least one decision tree to generate at least one pixel value of the test image, the classified image comprising a plurality of pixel values, wherein generation of each pixel value assigns the pixel value a probability that its associated pixel is related to an anatomical element; and

evaluate the classified image using an anatomical model to detect at least one anatomical element within the classified image.

13. The apparatus of claim 12, further configured to label the detected at least one anatomical element within the test image in response to detecting the at least one anatomical element within the classified image.

14. The apparatus of claim 13, wherein a location of the at least one anatomical element within the test image corresponds to a location in which the at least one anatomical element was detected within the classified image.

15. The apparatus of claim 12, further configured to:

receive a set of training images, the set of training images comprising at least one target image and at least one source image;

determine at least one image feature;

transform the at least one source image using the at least one image feature to generate at least one feature image;

generate at least one decision tree corresponding to the at least one image feature using at least the at least one feature image and the at least one target image; and

use the generated at least one decision tree as the image classifier.

16. The apparatus of claim 15, further configured to:

generate a plurality of decision trees, each of the decision trees corresponding to a set of image features;

evaluate the plurality of decision trees to determine an accuracy value for each decision tree; and

select at least one of the plurality of decision trees with the highest accuracy value as the image classifier.

17. The apparatus of claim 15, wherein apparatus is further configured to:

determine a set of pixel values associated with a particular node of the decision tree;

determine a feature that results in a minimum variance in target pixel values associated with the set of pixel values; and

assign the feature that results in a minimum variance in target pixel values as a decision feature for the particular node of the tree.

18. The apparatus of claim 17, further configured to:

determine a threshold feature value associated with the feature that results in the minimum variance, wherein the threshold feature value results in a split in the set of pixel values when applied to the set of pixel values; and

assign the threshold feature value to the particular node of the decision tree.

19. The apparatus of claim 12, wherein the anatomical elements are spinal vertebrae, and wherein the anatomical model defines an offset between adjacent vertebrae.

20. A computer program product comprising at least one non-transitory computer-readable storage medium bearing computer program instructions embodied therein for use with a computer, the computer program instructions comprising program instructions configured to:

receive a test image;

generate one or more feature vectors, wherein generating each feature vector comprises convolving a respective feature with the test image;

generate a classified image by using the one or more feature vectors to apply an image classifier to the test image, the image classifier comprising at least one decision tree, wherein application of the image classifier to the test image comprises feeding the one or more feature vectors through the at least one decision tree to generate at least one pixel value of the test image, the classified image comprising a plurality of pixel values, wherein generation of each pixel value assigns the pixel value a probability that its associated pixel is related to an anatomical element; and

evaluate the classified image using an anatomical model to detect at least one anatomical element within the classified image.