Automatic segmentation of prostate, rectum and urethra in ultrasound imaging
AssigneeAlpha Intervention Technology, Inc.
InventorsGang Gheng, Haisong Liu, Yan Yu
Life of this patent
Estimating the term…
Abstract
An ultrasound image of a transverse cross-sectional outline of the prostate gland is acquired. For three-dimensional imaging, a series of images can be taken, one of each slice of the prostate. The initial ultrasound images are pre-processed to remove the noise and increase the contrast. The rectum edge is located from the bottom of the images. The key points on the prostate boundaries are located and connected under the iterative training of a knowledge-based model until the shape of the boundary reaches a stable state. The urethra is segmented near the center of the prostate.
Claims48
A method of determining boundaries of an object, the method comprising:
- (a) acquiring an image of a transverse cross-sectional outline of the object;
- (b) pre-processing the image acquired in step (a) to remove noise and increase contrast so as to produced a pre-processed image;
- (c) finding an edge of the object from a point in the pre-processed image;
- and
- (d) locating key points of the boundaries of the object and connecting the key points under iterative training of a knowledge-based model until a shape of the boundaries of the object reaches a stable state.
A method for mean filtering of an image, the method comprising, for every pixel in the image:
- (a) connecting a straight line from the pixel to a point in the image;
- (b) computing an angle α of the straight line and a distance d from the pixel to the point;
- (c) setting a window size s for local mean filtering according to s=kd, where k>0;
- (d) locating a mean filtering window on the image, the mean filtering window centered on the pixel and perpendicular to the straight line;
- and
- (e) performing the mean filtering within the mean filtering window.
A system for determining boundaries of an object, the system comprising:
- (a) an imaging device for acquiring an image of a transverse cross-sectional outline of the object;
- and
- (b) a computing device, receiving the image from the imaging device, for pre-processing the image to remove noise and increase contrast so as to produced a pre-processed image, finding an edge of the object from a point in the pre-processed image, and locating key points of the boundaries of the object and connecting the key points under iterative training of a knowledge-based model until a shape of the boundaries of the object reaches a stable state.
Dependencies are inferred from the claim text.