EP1714250A2

EP1714250A2 - Automatic segmentation of tissues by dynamic change characterization

Info

Publication number: EP1714250A2
Application number: EP05702601A
Authority: EP
Inventors: Ori Hay; Armin Marcovitch; Ifat Lev
Original assignee: Koninklijke Philips Electronics NV
Current assignee: Koninklijke Philips NV
Priority date: 2004-01-29
Filing date: 2005-01-06
Publication date: 2006-10-25
Also published as: WO2005073915A3; CN1914642A; WO2005073915A2; JP2007521867A

Abstract

A reconstruction processor (24) reconstructs diagnostic data from a diagnostic imaging device, such as a CT scanner (10), starting before a contrast agent reaches a region of interest (50), as the concentration of contrast agent in the region of interest builds (52), and at a contrast agent peak (56). The plurality of images generated while the contrast agent concentration is building are aligned (78). A change map is generated indicative of a rate-of-change (62) gradient or a time-to-peak (64) for corresponding pixels or voxels of the images generated during the time the contrast agent is building to the peak. A segmentation processor (70) uses the change map in segmenting the diagnostic images generated without contrast agent or at the contrast agent concentration peak.

Description

AUTOMATIC SEGMENTATION OF TISSUES BY DYNAMIC CHANGE CHARACTERIZATION DESCRIPTION The present invention relates to the diagnostic imaging arts. It finds particular application in conjunction with the segmentation of contrast agent enhanced diagnostic images into different tissue regions or organs and will be described with particular reference thereto. In many diagnostic imaging situations, the diagnostician tries to determine the boundary between various tissues of interest, for example, the boundaries between the blood and the vessel wall. Often, the diagnostician relies on differences in intensity or gray scale to differentiate among the various tissues. Rather than relying on the human eye to make this differentiation, various computer programs have been developed for automatically segmenting or differentiating the various tissues of a diagnostic image. Typically, there are tissues, plaques, and other structures of interest which are not easily differentiated by intensity or gray scale. One technique for assisting this differentiation, particularly when one of the tissues is blood, is through the use of contrast agents. A contrast agent is a pharmaceutical agent which is typically injected into the blood to alter its response to the diagnostic imaging technique. In a CT scanner, the contrast agent is more highly x-ray absorptive than blood, causing the ^"blood which carries the contrast agent to appear bright. One can also determine the rate at which blood is absorbed by various tissues by watching the rate at which the contrast agent, hence the blood, permeates the tissue of interest. By comparing the same image with and without the contrast agent, segmentation of the image is facilitated. To maximize the effect of the contrast agent, a contrast agent enhanced image is generally collected when the contrast agent is at its peak. The contrast agent is injected into the blood. When the bolus contrast agent reaches the imaged region, the intensity of the blood increases along a relatively steep curve until it reaches a peak. The curve decays along a more shallow slope as the bolus of contrast agent passes from the region of interest. Ideally, the generation of the contrast agent enhanced image(s) is timed such that the imaging process commences just before the peak and continues until just after the peak. One technique for locating the peak includes generating a number of images of the region of interest after the contrast agent has been injected. The intensity of the blood remains flat until the bolus reaches the imaged region. By monitoring the increase in intensity along the steeply rising portion of the Gaussian curve and knowing the general shape of the curve, and knowing the general shape of the curve, the peak and the commencement time for imaging can be predicted. Although effective, there are still tissues or other substances of interest which are difficult to segment even using the combination of a contrast enhanced image and a corresponding image without a contrast agent. The present invention overcomes the above-referenced problems and others.

In accordance with one aspect of the present invention, an apparatus is provided for segmenting diagnostic images. A means generates images of a region of interest of a subject at least during a period of time when a contrast agent concentration is building toward a peak in the region of interest. Another means generates a change map or image in which each pixel or voxel has a value indicative of change in the corresponding pixel or voxel in the images generated during the period of time in which the contrast agent concentration is building toward the peak. In accordance with another aspect of the present invention, a method of segmenting diagnostic images is provided. Images are generated of a region of interest of a subject at least during a period of time when a contrast agent concentration is building toward a peak in the region of interest. A change map or image is generated in which each pixel or voxel has a value indicative of change in the corresponding pixel or value in the images generated during the time that the contrast agent is building toward the peak. One advantage of the present invention is that segmentation and identification of tissues and organs is made possible even when there is little or no difference in the gray scale or intensity of surrounding or adjoining tissues. Another advantage of the present invention is that segmentation of tissues and organs with intensity or gray scale values that change with a contrast agent is possible when other techniques based on fixed differences fail to segment all of an organ or tissue region of interest. Still further advantages of the present invention will be appreciated to those of ordinary skill in the art upon reading and understand the following detailed description.

The invention may take form in various components and arrangements of components, and in various steps and arrangements of steps. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. FIGURE 1 is a diagrammatic illustration of a diagnostic imaging system in accordance with the present invention; and, FIGURE 2 is illustrative of a blood intensity curve.

With reference to FIGURE 1, a diagnostic imaging device, such as a CT scanner 10, non-invasively examines a subject in an imaging region 12 and generates diagnostic image data representative of a region of interest of the subject disposed within the examination region. In the illustrated CT scanner embodiment, an x-ray tube 14 or other source of radiation generates a beam of radiation, preferably a multi-slice or cone-beam, which is directed through the examination region 12 to a radiation detector 16. A motor 18 and an associated drive (not shown) rotates the x-ray beam around the examination region, typically by rotating the x-ray source and detector. The subject is supported on a patient support 20 such as a patient couch. A motor 22 and associated gearing and connecting linkage (not shown) advance the subject longitudinally through the examination region to pass the region of interest of the subject through the examination region. For spiral scanning, the motor 22 moves the patient support continuously through the examination region as the motor 18 rotates the x-ray beam, such that the subject is scanned along a spiral trajectory. A reconstruction processor 24 reconstructs the generated image data into image representations, preferably three-dimensional image representations, which are stored in an image memory 30. More specifically, prior to injection of contrast agent, a diagnostic imaging sequence is conducted and a basis or reference diagnostic image is generated of the region of interest with no contrast agent, which image is stored in a basis image memory section 32. Once the contrast agent has been administered, a series of images are generated and stored in series of post-contrast agent injection image memory sections 34. When the contrast agent is at its peak, a peak contrast agent enhanced diagnostic image is generated and stored in a peak diagnostic image memory section 36. Optionally, additional series of diagnostic images are generated after the contrast agent peak and stored in a post-peak diagnostic image memory section 38. A portion of the basis or reference image is displayed on a monitor 40. The operator manipulates the displayed image to optimize the display of the region of interest. The operator uses a user input means 42 to designate a region or sub-region of the displayed image which is to be monitored for contrast agent build-up and for which diagnostic images are to be generated. Various subregion selecting means are contemplated, such as a keyboard, mouse, or other graphic input which enables the user to define preselected shapes at preselected points on the display, such as circles of a selectable diameter. As another option, the operator can draw free-hand on the display using a mouse, touch sensitive screen, or other electronic drawing device. As another option, the cursor can be manipulated to designate a single voxel or pixel, such as the center point of a blood vessel which is to be monitored alone or with a limited number of surrounding pixels or voxels. A plurality of subregions of interest may be designated to be monitored for the contrast agent peak. More specifically, a scanner controller 44 controls the scanner to take the basis image and the series of images after injection of the contrast agent. With continuing reference to FIGURE 1 and further reference to FIGURE 2, a contrast agent monitoring means 46 monitors the pixels or voxel in the designated subregion of the change images. Before the contrast agent reaches the designated region, the intensity will be relatively low as indicated by sampling points 50. As the bolus of contrast agent reaches the designated region, the intensity in the designated region increases as indicated by sampling points 52. Based on the sampling points in the increasing intensity region of the contrast agent curve and the knowledge of the typical shape of contrast-agent curves in the designated region, a peak predicting means or processor 54 predicts when the peak will be reached. More specifically, the peak predicting means causes the control means 44 to initiate the contrast agent peak data acquisition at a time 56 just slightly before the contrast agent peak such that a sampling window 58 for the peak intensity diagnostic image is substantially centered on the peak. It will be noted that in the illustrated embodiment, samplings 50, 52 are taken at regular intervals, but the exact timing of sampling point 56 is adjusted dynamically in order to capture the peak. It is further to be appreciated that image data may be taken at a higher rate such that many more samplings are acquired prior to the contrast agent peak being reached. A change determining means 60 determines a change map or image in which each voxel or pixel represents a change of the corresponding voxel or pixel in the change images from the post-contrast agent image memory 34. In the preferred embodiment, the change determining means 60 includes a gradient or rate-of-change determining means or processor 62 and a time-to-peak means or processor 64. The gradient or rate-of-change means determines a steepness or rate of change of the intensity of each voxel or pixel. The time-to-peak means determines time over which the intensity builds to the peak. The change map or image can be displayed as an image on the monitor 40. An automatic segmentation means 70 automatically segments each region designated by the user input means 42 for segmentation into different types of tissue or structures. The segmentation region can be designated with respect to the pixels around one or more seed cursors, an encircled region, or the like. Of course, rather than limiting the segmentation to the designated regions of interest, the segmentation can be conducted over the picture as a whole. In the preferred embodiment, the segmentation means looks at the change map segments of the region based on the change map as well as on the gray scale or intensity of the corresponding voxels of a diagnostic image from one of memories 32, 36, 38, or a difference between the peak image and contrast agent-free image. This is as opposed to segmenting based only on the gray scale or intensity of each voxel or pixel of the segmentation region with or without contrast agent or the difference between images with and without the contrast agent. By segmenting based on the change, the resultant image is segmented based on the rate at which the various tissue types take up the contrast agent. Segmenting based on this contrast agent uptake can also help to segment diseased or damaged tissue from healthy tissue. As another example, when the contrast agent is injected into the blood, the arteries quickly change intensity. . However, the intensity change of the viens vessel takes longer to peak than the intensity change in the arteries. Thus, although both arteries and veins may reach the same peak intensity, the time-to-peak varies and provides a basis for segmentation. Further, some tissues or substances in the body, such as plaque, may be imaged equally bright with the contrast agent enhanced blood, but can be differentiated due to its different or lack of change in intensity relative to the contrast agent enhanced pixels or voxels. Optionally, the segmenting means 70 can segment based on both the time-to- peak and the rate-of-change for each voxel. As yet another alternative, the segmenting means can segment based on one or both of the time-to-peak and the rate-of-change for each voxel, as well as on the traditional pre-contrast agent grayscale levels, post-contrast agent grayscale levels, a difference between the peak and pre-contrast agent levels, and the like. Preferably, the segmentation means 70 includes programs, processes, or other means for implementing automatic segmentation techniques including simple threshold techniques 72, region-growing techniques 74, active contours 76, and the like. An alignment processor 78 looks at each of the images in memory sections

32-38 and makes appropriate spatial adjustments to bring them into alignment for segmentation, as well as to be sure that the same voxel is followed by the gradient determining processor 62 and the time-to-peak processor 64. The alignment processor can register the images based on body contour, bones, or various other methods of registration as are known in the art. By way of example, one common technique for extracting a contour or surface using the active contours technique minimizes the energy. More specifically, based on grayscale change or a surface or boundary roughly indicated by the operator, the surface S(X) is extracted by:

where, X(s) =(x(s), y(s), z(s), ... n), in an n-dimensional contour or surface. Rather than limiting the n spatial dimensions, the present segmentation technique minimizes to an n+1 dimensional surface, the additional dimension being time. The surface function X(s) can now be written as: X(s)=(x(s), y(s), z(s), ..., t), where t is time. For example, rather than minimizing energy in an image generated without contrast agent or within images generated with and without the contrast agent, the energy is also minimized with respect to the change map or image and/or the time-to-peak map or image. In region growing, the region growing techniques examine a neighborhood of the current pixels, neighboring being in the sense of space and time. The neighboring pixels are examined starting at a seed pixel or voxel and moving outward to contiguous voxels having like time-to-peak, gradient, grayscale, and like characteristics. The segmented images are processed by a processor 80 to create a user- selected display format, as is convention in the art, for display on the monitor 40. The invention has been described with reference to the preferred embodiments. Modifications and alterations may occur to others upon reading and understanding the preceding detailed description. It is intended that the invention be constructed as including all such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.

Claims

CLAIMS Having thus described the preferred embodiments, the invention is now claimed to be: 1. An apparatus for segmenting diagnostic images, the apparatus comprising: a means (10, 24, 44) for generating images of a region of interest of a subject at least during a period of time when a contrast agent concentration is building toward a peak in the region of interest; a means (60) for generating a change map or image in which each pixel or voxel has a value indicative of change in the corresponding pixel or voxel in the images generated during the period of time the contrast agent concentration is building toward the peak.

2. The apparatus according to claim 1, further including: a means (70) for segmenting generated diagnostic images in accordance with the change map or image.

3. The apparatus according to claim 2, wherein the diagnostic images are generated at least one of: (a) before contrast agent reaches the region of interest and (b) after contrast agent concentration peaks in the region of interest, and the segmentation means (70) segments the diagnostic image based on the change map or image.

4. The apparatus according to claim 3, further including: a display means (40) for displaying at least one of the segmented image and the change map or image.

5. The apparatus according to claim 2, wherein the change means (60) includes: a means (62) for determining a rate-of-change in each corresponding pixel or voxel of the change images.

6. The apparatus according to claim 2, wherein the change means (60) includes: a means (64) for determining a time for the contrast agent to build to a peak in each voxel or pixel of the change image.

7. The apparatus according to claim 2, wherein the segmentation means (70) includes at least one of a threshold segmentation means (72), a region growing segmentation means (74), and an active contours segmentation means (76).

8. The apparatus according to claim 2, wherein the imaging means (10, 24, 44) generates a diagnostic image at least one of: before contrast agent reaches the region of interest and when a contrast agent concentration in the region of interest is at a peak and wherein the segmentation means (70) segments the diagnostic image based on a combination of intensity values of the diagnostic image and the change map.

9. The apparatus according to claim 2, further including an alignment means (78) for aligning diagnostic images taken: before the contrast agent reaches the region of interest, when contrast agent is at the peak, and during the time the contrast agent is building toward the peak.

10. The apparatus according to claim 2, further including a region designating means (42) for designating one or more regions of interest of an imaged volume to be segmented.

11. A method of segmenting diagnostic images, the method comprising: generating images of a region of interest of a subject at least during a period of time when a contrast agent concentration is building toward a peak in the region of interest; generating a change map or image in which each pixel or voxel has a value indicative of change in the corresponding pixel or voxel in the images generated at least during the time that the contrast agent is building toward the peak.

12. The method according to claim 11, further including: segmenting generated diagnostic images in accordance with the change map or image.

13. The method according to claim 12, wherein the diagnostic images are generated at least one before contrast agent reaches the region of interest, and after contrast agent concentration peaks in the region of interest, and the segmentation step segments the diagnostic image based on the change map or image.

14. The method according to claim 13, further including: displaying at least one of the segmented image and the change map.

15. The method according to claim 12, wherein the change determining step includes at least one of: determining a rate-of-change in each corresponding pixel or voxel of the images generated during the period of time that the contrast agent is building to the peak; and determining a time for the contrast agent to reach the peak in each pixel or voxel.

16. The method according to claim 12, wherein the segmentation step includes at least one of threshold, a region growing, and an active contours segmenting.

17. The method according to claim 12, wherein the imaging generating step includes generating a diagnostic image at least one of before contrast agent reaches the region of interest and when a contrast agent concentration in the region of interest is at the peak and wherein the segmentation step segments the diagnostic image based on a combination of mtensity values of the diagnostic image and the change map.

18. The method according to claim 12, further including: aligning images generated before the contrast agent reaches the region of interest, when contrast agent is at a peak concentration in the region of interest, and while the contrast agent is building toward the peak.

19. The method according to claim 12, further including: designating one or more regions of an imaged volume to be segmented.

20. In a diagnostic imaging apparatus (10) which generates diagnostic images of a subject, a segmentation processor (70) programmed to segment the diagnostic images according to the method of claim 12.