Method and system for automatic algorithm selection for segmenting lesions on pet images

• US 7,953,265 B2
• Filed: 11/22/2006
• Issued: 05/31/2011
• Est. Priority Date: 11/22/2006
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Ubelhart, Istvan, Cziria, Balazs, Sirohey, Saad Ahmed
• Primary Examiner(s): Perungavoor; Sath V
• Assistant Examiner(s): HEIDEMANN, JASON E
• Application Number:
  US11/603,861
• Publication Number:
  US 20080118134A1
• Time in Patent Office: 1,651 Days
• Field of Search: None
• US Class Current: 382/131
• CPC Class Codes: G06V 10/26 Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
• CPC Class Codes: G06V 2201/03 Recognition of patterns in medical or anatomical images

Abstract

A method and system is provided for automatically processing a volumetric diagnostic image dataset. A first seed point is defined within a lesion. The lesion is within a first image dataset representative of a subject. A first boundary is defined in three dimensions within the first image dataset and the first seed point and the lesion are within the first boundary. At least one parameter is determined based on the first image dataset. A first segmentation algorithm is selected from a plurality of segmentation algorithms based on the at least one parameter, and the lesion is segmented using the first segmentation algorithm.

Claims

1. A method for automatically processing a volumetric diagnostic image dataset, comprising:

defining a first seed point within a lesion, the lesion being within a first image dataset representative of a subject;

defining a first boundary in three dimensions within the first image dataset, the first seed point and the lesion being within the first boundary;

determining at least one parameter based on the first image dataset;

obtaining at least one preliminary lesion parameter;

selecting a first segmentation algorithm from a plurality of segmentation algorithms based on the at least one determined parameter and the at least one preliminary lesion parameter; and

segmenting the lesion using the first segmentation algorithm.

2. The method of claim 1, further comprising:

defining a second seed point and a second boundary on a second image dataset of the subject, the first and second image datasets comprising like anatomy, the second seed point and second boundary being positioned on the second image dataset based on positions of the first seed point and first boundary on the first image; and

segmenting image data within the second boundary based at least in part on the first image dataset, the first segmentation algorithm, and at least one subsequent parameter based on the second image dataset.

3. The method of claim 1, further comprising pre-segmenting the lesion within the first boundary, the at least one determined parameter or the at least one preliminary lesion parameter being based on a result of the pre-segmenting.

4. The method of claim 1, further comprising determining at least one second parameter based on at least one of an acquisition parameter used when acquiring the first image dataset, disease history based on the subject, treatment history of the subject, and expected results of the treatment of the subject, the segmenting being based at least partially on the at least one second parameter.

5. The method of claim 1, further comprising:

defining a second seed point and a second boundary on a second image dataset of the subject, the first and second image datasets comprising like anatomy, the second seed point and second boundary being positioned on the second image dataset based on positions of the first seed point and first boundary on the first image;

determining at least one subsequent parameter based on the second image dataset;

selecting a subsequent segmentation algorithm from the plurality of segmentation algorithms based on the at least one subsequent parameter, the at least one parameter of the first image dataset, and the first segmentation algorithm; and

segmenting image data within the second boundary using the subsequent segmentation algorithm.

6. The method of claim 1, further comprising segmenting a second image dataset of the subject based on a comparison of at least one parameter associated with the first image dataset and the second image dataset.

7. The method of claim 1, further comprising:

defining a second boundary within a second image dataset, the first and second image datasets comprising like anatomy of the subject, the first and second boundaries surrounding a sub-set of like anatomy of the subject;

selecting a subsequent segmentation algorithm based at least on image data within the second boundary and image data within the first boundary; and

segmenting the image data within the second boundary using the subsequent segmentation algorithm.

8. A system for automatically segmenting a lesion within a positron emission tomography (PET) image dataset, comprising:

a display for displaying a volumetric PET image dataset of a subject;

an input device for defining a seed point within a lesion on the image dataset displayed on the display;

a CPU coupled to the display and the input device, the CPU defining a three dimensional bounding box around the lesion within the image dataset; and

a segmentation algorithm selection module coupled to the CPU, the selection module being configured to;

detect at least one parameter associated with at least one of image data within the image dataset and data associated with the subject;

obtain at least one preliminary lesion parameter; and

select a segmentation algorithm from a predetermined list of segmentation algorithms based on the at least one determined parameter and the at least one preliminary lesion parameter, the CPU segmenting the lesion with the segmentation algorithm.

9. The system of claim 8, wherein the data associated with the subject comprises non-image data.

10. The system of claim 8, wherein the at least one parameter is based on image data within the bounding box.

11. The system of claim 8, further comprising:

the display displaying a subsequent image dataset comprising like anatomy with respect to the image dataset; and

the selection module automatically selecting the segmentation algorithm based at least in part on a comparison of the image dataset and the subsequent image dataset.

12. The system of claim 8, further comprising:

the display displaying a subsequent image dataset comprising like anatomy with respect to the image dataset; and

the selection module automatically defining a subsequent seed point and a subsequent bounding box within the subsequent image dataset based on the seed point and the bounding box of the image dataset, the selection module selecting the segmentation algorithm based at least in part on a comparison of the image data within the bounding box and image data within the subsequent bounding box.

13. The system of claim 8, further comprising:

the display displaying a subsequent image dataset comprising like anatomy with respect to the image dataset;

the selection module defining at least one second parameter based on at least a portion of the subsequent image dataset, the selection module modifying the segmentation algorithm based on the at least one second parameter to determine a modified segmentation algorithm; and

the CPU processing the subsequent image dataset with the modified segmentation algorithm.

14. The system of claim 8, further comprising:

the display displaying a subsequent image dataset comprising like anatomy with respect to the image dataset;

the selection module detecting a first and second parameter based on the image dataset and the subsequent image dataset, respectively, the selection module comparing the first and second parameters to identify a difference beyond a predetermined threshold associated with the parameter, the selection module selecting the segmentation algorithm based at least in part on the difference.