Segmentation method and apparatus for medical images using diffusion propagation, pixel classification, and mathematical morphology

• US 6,785,409 B1
• Filed: 10/24/2000
• Issued: 08/31/2004
• Est. Priority Date: 10/24/2000
• Status: Expired due to Fees

• Current Assignee: Marconi Medical Systems, Inc. (Koninklijke Philips N.V.)
• Original Assignee: Koninklijke Philips Electronics N.V. (Koninklijke Philips N.V.)
• Inventors: Suri, Jasjit S.
• Primary Examiner(s): Mehta, Bhavesh M.
• Assistant Examiner(s): Choobin, Barry
• Application Number:
  US09/695,667
• Time in Patent Office: 1,407 Days
• Field of Search: 382/128, 382/130, 382/255, 382/308, 382/305, 382/131, 382/228, 382/132, 382/173, 382/170, 382/171, 382/194, 382/199, 382/203, 382/221, 382/224, 382/232, 382/242, 382/243, 382/266, 382/267, 382/238, 382/219, 382/106, 382/288, 128/922, 250/455, 345/723, 356/39, 377/10, 600/407, 600/425
• US Class Current: 382/128
• CPC Class Codes: G06T 2207/20132 Image cropping
• CPC Class Codes: G06T 2207/30016 Brain
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/155 involving morphological operators

Abstract

A method of digital imaging includes receiving image data and fitting a curve to boundaries within the image data. The curve is fit to the boundaries within the image data by extracting a region of interest from the image data and computing a signed distance transform in a narrow band within the region of interest. Finite difference equations including various variables are solved to determine a rate at which the distance transform changes. The distance transform is then diffused at that rate. The technique is based on region-based diffusion propagation, pixel classification, and mathematical morphology. The method is implemented to run in the narrow band of the region of interest specified by the user and the computations are implemented using a fast marching method in the narrow band. While idealized for distinguishing segments of white matter, gray matter, and cerebral spinal fluid in the brain, the algorithm can applied to find contours in any digital image.

Claims

1. A method of segmenting a medical image comprising:

determining a region of interest on the medical image;

estimating a contour corresponding to a segment of the image;

computing a propagation speed indicative of a rate at which the contour changes as the estimated contour approaches a final contour;

computing an altered contour within the region of interest based on a previous contour and the propagation speed; and

extracting the final contour from the region of interest.

2. The method of segmenting a medical image as set forth in claim 1, where the region of interest comprises a first set of pixels distinguishable from other sets of pixels in the medical image and the extracting a final contour step comprises repetitively adjusting the altered contour until the altered contour substantially circumscribes the first set of pixels.

3. The method of segmenting a medical image as set forth in claim 2, where the first set of pixels includes one of the set of white matter, gray matter, and cerebral spinal fluid.

5. The method of segmenting a medical image as set forth in claim 1, where the propagation speed includes curvature speed relating to curvature of the contour.

7. The method of segmenting a medical image as set forth in claim 1, where the propagation speed includes gradient speed relating to gradient information of the medical image.

8. The method of segmenting a medical image as set forth in claim 1, where the propagation speed includes shape speed relating to a shape of the contour.

11. The method of segmenting a medical image as set forth in claim 1, further comprising:

registering the medical image with the final contour; and

displaying the registered medical image and final contour.

4. A method of segmenting a medical image comprising:

determining a region of interest on the medical image, the region of interest including a first set of pixels having a defined similarity to each other;

computing a propagation speed indicative of a rate at which a contour changes;

computing an altered contour within the region of interest based on a previous contour and the propagation speed; and

extracting a final contour from the region of interest by repetitively adjusting the altered contour until the altered contour substantially circumscribes the first set of pixels.

6. A method of segmenting a medical image comprising:

selecting a region of interest on the medical image;

computing a propagation speed indicative of a rate at which a contour changes, the propagation speed includes including at least one of;

region speed relating to the determined region of interest,

gradient speed relating to gradient information of the medical image, and

shape speed relating to a shape of the contour;

computing an altered contour within the region of interest based on previous contour and the propagation speed; and

extracting a final contour from the region of interest.

9. A method of segmenting a medical image comprising:

determining a region of interest on the medical image;

computing a propagation speed indicative of a rate at which a contour changes;

computing an altered contour within the region of interest based on a previous contour and the propagation speed;

computing a signed distance transform of a previous contour using a curve layering method in a band surrounding the contour; and

extracting a final contour of the region of interest.

10. The method of segmenting a medical image as set forth in claim 9, where the computing a signed distance transform step comprises:

determining an accepted set of pixels;

tagging a trial set of pixels and a far set of pixels; and

calculating distances of the trial set from the accepted set and calculating distances of far set from the accepted set.

12. A method of circumscribing an object in a medical image comprising:

determining an accepted set of pixels around a contour on the image;

tagging a trial set of pixels and a far set of pixels;

calculating distances of the trial set from the accepted set and calculating distances of the far set from the accepted set; and

iteratively computing an altered contour on the image based on the distances calculated including determining a rate of contour change by calculating a speed function controlled by regional, gradient, and curvature change speeds.

13. A medical imaging apparatus comprising:

a source of digitized medical image data;

a user interface which allow_s a user to input an initial contour onto a representation of the digitized medical image data;

a segmentation processor which adjusts the contour to boundaries within the digitized medical image data, the segmentation processor including;

a mathematical morphology algorithm which identifies a region of interest in the image data;

a fast marching method algorithm which computes a signed distance transform (SDT) in a narrow band within the region of interest;

a speed control algorithm which determines a rate at which the SDT changes;

a propagation algorithm which diffuses the SDT at the rate determined; and

,

a processor which processes the contour and image data for display.

14. The medical imaging apparatus as set forth in claim 13, where the speed control algorithm comprises:

a finite difference equation solver which solves a difference equation including terms of regional speed, gradient speed, curvature speed and a speed based on contour shape.

15. The medical imaging apparatus as set forth in claim 13, where the fast marching method algorithm comprises:

flood filling a contour yielding a regional image;

obtaining an accepted set of contour points and distances in the narrow band from the regional image;

tagging a trial set and a far set of points defining a revised contour;

computing trial set distances from the accepted set; and

computing far set distances from the accepted set.

16. The medical imaging apparatus as set forth in claim 15, where the obtaining an accepted set comprises:

checking a location of each pixel point in the image relative to a central pixel; and

computing a fractional distance of each pixel point from the central pixel.

17. A method of determining a contour for pixel groups within a diagnostic image comprising:

sorting pixels in the image into a defined number of classes;

iteratively, fitting a curve to pixels of a like class within the image; and

registering at least the curve and the image for display, the fitting a curve step including;

extracting a region of interest in the image;

computing a signed distance transform contour in a narrow band within the region of interest;

determining a rate at which the contour changes; and

expanding the contour at the rate determined.

18. The method of determining a contour as set forth in claim 17, where the determining a rate step comprises:

solving the finite difference equation including terms of regional speed, gradient speed, curvature speed and, a speed based on organ shape.

19. A method of determining a contour comprising:

sorting pixels in a diagnostic image into a defined number of classes;

iteratively, fitting a contour to pixels of a like class within the diagnostic image;

flood filling the contour yielding a regional image;

obtaining an accepted set of contour points and distances in a narrow band;

tagging a trial set and a far set of points defining a revised contour;

computing trial set distances from an accepted set; and

computing far set distances from the accepted set.

20. The method of determining a contour as set forth in claim 19 where the obtaining an accepted set comprises:

checking a sign of each pixel point in the image as immediately surrounding a central pixel;

computing a fractional distance of each pixel point from the central pixel.

21. An apparatus for segmenting medical images comprising:

a display on which a region of interest is displayed, the region of interest including at least one segment of pixels that have a defined similarity to each other;

a contour altering software routine which iteratively alters an input estimated contour to an output estimated contour, the contour altering software route calculating a speed at which the contour changes;

an initial estimate software routine which inputs an initial estimated contour of the at least one segment into the contour altering routine;

a final contour extracting software routine for iteratively inputting the output estimated contour into the contour altering software routine until a final contour which substantial circumscribes the at least one segment is extracted.

22. The apparatus as set forth in claim 21 wherein the contour altering routine includes:

a region speed software routine which calculates a rate at which the region changes;

a shape speed software routine which calculates a rate at which a shape of the contour is changing;

a gradient speed software routine which calculates a rate at which an edge of the contour is changing.