Fast segmentation of cardiac images

• US 5,903,664 A
• Filed: 02/05/1997
• Issued: 05/11/1999
• Est. Priority Date: 11/01/1996
• Status: Expired due to Fees

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Hartley, Richard Ian, Cline, Harvey Ellis, Curwen, Rupert William Meldrum
• Primary Examiner(s): Couso, Jose L.
• Assistant Examiner(s): Dang, Duy M.
• Application Number:
  US08/794,981
• Time in Patent Office: 825 Days
• Field of Search: 382/154, 382/171, 382/172, 382/173, 382/128, 382/130, 382/131, 382/132, 382/260, 382/261, 382/272, 382/270, 382/204, 324/309, 324/307, 364/413.3, 364/418, 364/518, 364/555
• US Class Current: 382/154
• CPC Class Codes: G06T 2207/20012 Locally adaptive
• CPC Class Codes: G06T 2207/20101 Interactive definition of point of interest, landmark or seed
• CPC Class Codes: G06T 2207/20104 Interactive definition of region of interest [ROI]
• CPC Class Codes: G06T 2207/30048 Heart; Cardiac
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/11 Region-based segmentation

Abstract

A cardiac segmentation system acquires a series of images acquires as slices through a volume, and as images at different time periods throughout a cardiac cycle. It displays an image to an operator which interactively selects a region of interest (ROI) of the image to be segmented, such as the left ventricle. A seed point is also selected within the ROI and the structure desired to be segmented. The image is then thresholded by a masking device classifying points within the ROI as above the threshold, or not above the threshold. A 3D connectivity device identifies points within the ROI having the same classification as an expanded seed point which are also contiguous with the seed point as the segmented structure. The segmented structure is expanded and a histogram is constructed. A new threshold is selected which separates modes of the histogram, and used to carry out a revised, final, segmentation of the current image. The centroid of the current image is used as a seed point in segmenting adjacent images. Similarly, the current threshold is used as an initial threshold for adjacent images. The previous seed point and ROI may also be used. This is repeated for a number of images to result in segmented structures may then be stored, displayed and used in calculating heart functionality.

Claims

1. An image segmentation system for segmenting a set of volumetric images acquired over a plurality of phases during a cycle into structures comprising:

a) an image memory having a said prestored set of volumetric data representing "slices" of images acquired over a plurality of phases;

b) a masking device for interactively receiving an operator-defined threshold, and an image slice operating to create a threshold image displayed to an operator showing pixels having a value either above or below the threshold;

c) a 2D connectivity device coupled to masking device, for receiving a threshold from the masking device, an operator-defined seed point and region of interest (ROI), and an image slice, operating to segment the volumetric images by determining contiguous points of the image starting in a predetermined vicinity of the seed point which are within the ROI and have values above the threshold, being a segmented structure;

d) an image selector coupled to the image memory and the masking device for selecting one of the image slices as a current image from the image memory and passing it to the masking device to produce a threshold image;

e) a threshold-update device coupled to the masking device and the image selector, operating to receive the current image slice, a plurality of images adjacent to the current image slice in time or in space, the segmented structure and the ROI, analyze the intensities of pixels in the segmented structure and in a neighborhood of said segmented structure and compute an updated threshold, which is provided to the 2D connectivity device to segment the current image slice a second time, said updated threshold being also used as an initial threshold for adjacent image slices; and

f) a threshold update device coupled to the masking device and the image selector, operating to receive a plurality of adjacent image slices of the segmented structure for a plurality of phases, operating to expand the segmented structure by a predetermined amount, create a histogram of pixel values of at least one of the slices of the expanded segmented structure, and determine a threshold which will separate peaks of the histogram, and provide this threshold to the masking device for subsequent segmentation of an adjacent slice.

2. The image segmentation system of claim 1 in wherein the threshold-update device operates to receive a plurality of image slices adjacent in space, expand the segmented structure by a predetermined amount, create a histogram of the pixels of the expanded segmented structure effectively space filtering, determine an updated threshold which separates peaks of the histogram, and provide this adjusted threshold to the masking device to be used in segmenting the image a second time.

3. The image segmentation system of claim 1 further comprising:

a control panel coupled to the masking device allowing operator definable input, including the initial threshold, the seed point, and the ROI.

4. The image segmentation system of claim 1 further comprising:

a) a display device for displaying visual images of graphics signals provided to it;

b) a graphics engine coupled to the display device, the masking device, and the 2D connectivity device for interactively creating a graphics signal of the threshold image, and the segmented structure based upon the operator defined input, and displaying these signals on the display device.

5. The image segmentation system of claim 1 in wherein the threshold-update device operates to receive a plurality of image slices adjacent in time, expand the segmented structure by a predetermined amount, create a histogram of the pixels of the expanded segmented structure effectively time-filtering, determine an updated threshold which separates peaks of the histogram, and provide this adjusted threshold to the masking device to be used in segmenting the image a second time.

6. A method of segmenting a plurality of slice images of a desired volume acquired over a plurality of phases of a period cycle, into anatomical structures comprising the steps of:

a) selecting an initial image to be the current image acquired at an initial slice and phase;

b) selecting an initial threshold;

c) displaying the set of points having values either above, or below the threshold, as a threshold image;

d) selecting an initial region of interest (ROI) and a seed point within the ROI on the threshold image;

d) classifying point of the image within the ROI as being above the threshold and not above the threshold;

f) identifying a region of contiguous points having the same classification as the seed point, being connected to the seed point through points all having this same classification, defining a segmented structure in this image;

g) analyzing pixels in the segmented structure and in a neighborhood of the segmented structure to determine an updated threshold by;

i. expanding the segmented structure in a plurality of images for the same phase/slice by a predetermined amount to result in an expanded region; and

ii. determining a histogram of the intensities of the expanded region in the plurality of images effectively filtering in time/space, respectively;

iii. selecting an adjusted threshold that separates modes of the histogram;

h) segmenting the current image again by repeating steps "e" and "g" to compute an updated threshold;

i) selecting an adjacent image, in phase or in space, next to the current image; and

j) repeating steps "e" →

"i" using each successive updated threshold for a plurality of the images to result in a 3D segmented structure over a plurality of phases.

7. The method of segmenting a plurality of slice images of claim 6 in which the seed point and initial threshold used to segment all images except the first are the chosen comprising the steps of:

a) choosing a seed point for an image adjacent in time or space, to be a centroid of the updated segmented structure; and

b) choosing the initial threshold of the adjacent image as the updated threshold of the current image.

8. The method of segmenting a plurality of slice images of claim 7 in which the updated threshold is the same as the initial threshold.