WO2000063831A1 - Procede et appareil d'agrandissement d'image par optimisation et segmentation de donnees - Google Patents

Procede et appareil d'agrandissement d'image par optimisation et segmentation de donnees Download PDF

Info

Publication number
WO2000063831A1
WO2000063831A1 PCT/US2000/010532 US0010532W WO0063831A1 WO 2000063831 A1 WO2000063831 A1 WO 2000063831A1 US 0010532 W US0010532 W US 0010532W WO 0063831 A1 WO0063831 A1 WO 0063831A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
projection
points
point
along
Prior art date
Application number
PCT/US2000/010532
Other languages
English (en)
Inventor
Dennis L. Parker
Andrew Lafayette Alexander
John Austin Roberts
Brian Earl Chapman
Original Assignee
University Of Utah Research Foundation
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by University Of Utah Research Foundation filed Critical University Of Utah Research Foundation
Priority to AU46480/00A priority Critical patent/AU4648000A/en
Priority to US09/959,236 priority patent/US6674894B1/en
Publication of WO2000063831A1 publication Critical patent/WO2000063831A1/fr
Priority to US10/692,133 priority patent/US7120290B2/en
Priority to US12/249,583 priority patent/USRE43225E1/en

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T11/00—2D [Two Dimensional] image generation
    • G06T11/003—Reconstruction from projections, e.g. tomography
    • G06T11/008—Specific post-processing after tomographic reconstruction, e.g. voxelisation, metal artifact correction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10081—Computed x-ray tomography [CT]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30101—Blood vessel; Artery; Vein; Vascular
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2211/00—Image generation
    • G06T2211/40—Computed tomography
    • G06T2211/404—Angiography

Definitions

  • the present invention relates generally to data image processing and, more specifically, to generating enhanced digital images utilizing image segmentation processes that involve depth buffer information in association with point selection projection images.
  • 3D imaging techniques such as magnetic resonance imaging (MRI), computed tomography (CT), and rotation X- ray CT angiography (XRCTA)
  • MRA magnetic resonance angiography
  • CTA 3D CT angiography
  • arterial beds are generally sparse, so there is typically a large amount of nonvascular image data present in a 3D angiographic image.
  • MRA examinations have increased recently, having adequate means to effectively and efficiently review the available image information is a significant challenge for radiologists.
  • the first technique is to display selected original cross-sectional MRA images.
  • the second technique is to display oblique planar reformatted images.
  • the last technique is to display a maximum intensity projection (MIP) image.
  • MIP maximum intensity projection
  • the original cross-sectional images also referred to as slices, contain the maximum amount of information on a local level.
  • Each image displays the transverse segments and cross sections of vessels that intersect the plane of the image.
  • the vascular detail of interest may be observed in a single image. More often, because of the intricate paths of the vessels, many images must be viewed, and the information from each image must be integrated to formulate an understanding of the structure of interest. This method is inefficient since it requires a single slice to be displayed, which slice contains very little global information about any single intricate vessel in the overall vascular network.
  • the oblique planar reformatted images provide improved efficiency over the original images by displaying image planes that follow a segment of a vessel of interest through the three-dimensional image data.
  • Volume rendering consists of the projection or rendering of an entire 3D image volume onto a single two-dimensional image.
  • the volume is projected along parallel or diverging lines through the three-dimensional volume onto a two-dimensional image.
  • the intensities along the projection line are transformed according to some specified transformation.
  • Such rendering in a variety of forms, has become very useful in assisting the observer in the rapid and efficient interpretation of the large amounts of image data originally obtained.
  • a simple form of volume rendering which is also intuitive, is an X- ray-like summation of the image densities or intensities.
  • the MIP algorithm is successful to the extent that the signal from vessels is greater than the signal of the surrounding tissues. In regions where vessels do not appear overlapped, the MEP algorithm is sufficient to display vessels that are hyperintense relative to the variations in the overall background. Unfortunately, the MIP algorithm does not provide any information about vessels that are hidden below the intensity of other structures. Because of the sparseness of vessels, there is a significant amount of vessel detail that is not hidden and the MIP performs very well in increasing the amount of vessel detail observed in a single image.
  • the MIP algorithm provides a large amount of useful information in a single display.
  • the information density, or information content per image element is much higher in a MIP image than in the original 3D source data.
  • the source image data contains more total information, including some small arteries that would not appear in the MIP image
  • the density of vessel information in the source image data i.e. the percentage of image elements associated with vessels
  • many investigators have tried to develop algorithms to overcome the limitations of the MIP algorithm, and, although these have been viewed as improvements, the improvements have not been sufficient for any of these algorithms to replace the MIP algorithm.
  • the advantages of the MIP algorithm typically outweigh the disadvantages found therein. These advantages include reduced dynamic range, generally consistent image display, and improved signal difference to noise ratio (SDNR or contrast to noise ratio) for vessels that appear in the MIP.
  • SDNR signal difference to noise ratio
  • the artifacts in the MIP although very real, are also well understood and can be "read through.”
  • the MEP display contains a large amount of global information about the vascular system.
  • the appearance of a MIP image is quite similar to that of an X-ray angiogram, but there are several differences. The MIP simply selects the image element with the maximum intensity along each projection line, while the X-ray projection is a summation of all densitometric information along each projection line.
  • the MIP image is therefore a flat display having no depth information and no information about the thickness of the vessel through which the projection line passed, while in X-ray angiography the vessel brightness or darkness depends directly on the length of the X-ray proj ection path through the vessel.
  • the signal intensity is not related to the proj ection path length through the vessel, an increase in vessel signal is not observed for foreshortened vessels.
  • Statistical nosie properties associated with the signal of the tissue background cause the signal level of the background in the MIP to increase with the projection volume thickness.
  • ZFI zero filled interpolation
  • MRA display capabilities are also useful in a number of intracranial applications. There are various disease processes for which the diagnosis can benefit from improved image display capabilities. These diagnostic capabilities include predicting the need for thrombolysis in stroke conditions, diagnosing vasculitis and other occlusive diseases, identifying intracranial tumors and arterial venous malformations, and performing preoperative assessments. Further, MRA and other 3D angiographic images provide useful assistance for surgical procedures. The display of catheter angiograms and/or MRA image data have been found to be important aids during aneurysm surgery. The usefulness of the display of 3D angiographic image data during surgery can be enhanced by the display of the angiographic images in conjunction with 3D images of other types of anatomy such as adjacent bone structures or critical organs.
  • a method and system are disclosed for enhancing and segmenting a multi-dimensional image based upon the depth buffer (or
  • Z-buffer that is generated in conjunction with a maximum intensity projection (or related point selection projection) operation through the multi-dimensional image.
  • This enhancement and segmentation is referred to as the depth-buffer segmentation (DBS) process.
  • DBS depth-buffer segmentation
  • the DBS process segments an MRA image volume into regions having a likelihood of vessel occupation and regions that are unlikely to contain vessels. The process reduces the total image volume to a much smaller volume so that the amount of territory that a user must cover is greatly reduced.
  • projection images of the vascular information found within the data become much more refined and visible when the background regions have been removed.
  • the DBS process merges the global properties of the MIP process with the local properties of continuity to achieve a high specificity in the segmentation of vessel voxels from background voxels while maintaining a high sensitivity.
  • the segmentation provides an accurate and robust view of the vessel structure and filters out most of the non-vessel regions in the image.
  • the vascular detail can then be converted back to image data for display - for example, as a densitometric summation of the MRA image data resembling that of an X-ray angiogram.
  • the dynamic range of the display is reduced by hollowing out the vessel structures and displaying the vascular details as X-ray projection through hollow tubes.
  • the image may be modified by adding a degree of shading so that the 3D vessel structure is apparent in a manner not visible in X-ray images. Such image shading was not possible in the MIP process.
  • the method generates a reduced dimensionality image data set from a multi-dimensional image by formulating a set of projection paths through image points selected from the multi-dimensional image, selecting an image point along each projection path, analyzing each image point to determine spacial similarities with at least one other point adjacent to the selected image point in a given dimension, and grouping the image point with the adjacent point or spacial similarities between the points is found thereby defining the data set.
  • the method within the analyzing step, the step of determining similarity of brightness between the image point and the adjacent point. Further, the analyzing step also determines similarity of smoothness between the image point and the adjacent point. In one example, the smoothness is determined by using a least squares fit of adjacent image points.
  • the method further includes selecting another image point along the projection path and performing the analyzing and grouping steps on the newly selected image point. Further still, the method may also convert the grouped and ungrouped image points into a multi- dimensional image and then perform region growing within the converted multidimensional image or perform hollo wing-out of the multi-dimensional image for image enhancement. Another image enhancement steps include removing all pixels that are surrounded on each side by an adjacent pixel prior to displaying the image of the merged and unmerged image points.
  • the system further defines an image processing apparatus that comprises means for defining a set of projection paths through a multi-dimensional image, means for selecting at least one point, along each projection path, based upon a specified criterion, means for formulating an array of projection values corresponding to the positions of selected points along their respective projection path, and means for grouping a selected number of projection values based upon their proximity to other projection values.
  • the apparatus essentially is a programmable computer system that loads a particular software program capable of implementing the steps of the method claims within a computer architecture environment for manipulating the data and processing it for imaging, whether the imaging is in a printed image or a displayed image, such as on a computer monitor.
  • the method may be implemented in a computer program product for sale or distribution, whether that product is in a portable medium or in a fixed medium or remote medium that is communicated via a communications source, such as the Internet or other communication means known to those skilled in the art.
  • Figure 1 illustrates a block diagram of the imaging system utilizing depth-buffer (or Z-buffer) segmentation in accordance with the present invention
  • Figure 2 illustrates a flow diagram of the method steps implemented by the imaging apparatus of Figure 1 in rendering DBS images in accordance with the present invention
  • Figure 3 illustrates a standard axial MIP image of a patient with a large ophthalmic artery aneurysm
  • Figure 4 illustrates a depth-buffer or Z-buffer array corresponding to Figure 3 in which brightness is proportional to the "Z" location of each point selected for the MIP along each proj ection line ;
  • Figure 5 depicts a plot of the intensity values of the data from the projection lines for the 30 points indicated by the short line that crosses the middle cerebral artery as shown in Figure 4.
  • Figures 6 through 8 illustrate an expanded view of the small region of the Z- buffer array of Figure 4 where Figure 6 illustrates example line segments used for the least squares fit roughness calculation, Figure 7 illustrates image brightness inversely proportional to the roughness values in accordance with the present invention, and Figure 8 illustrates the same region with brightness set proportional to group size after grouping low roughness pixels that have similar depth values - i.e. have connectivity in the Z- buffer;
  • Figures 9 through 11 illustrate a series of images where each image represents where Figure 9 is a brightness inversely proportional to the roughness value in accordance with the present invention, Figure 10 is a brightness based upon number of points connected in the Z-buffer, where connectivity is based upon local smoothness and proximity in the depth or "Z" direction, and Figure 11 is a brightness proportional to number of points connected after connectivity has been enhanced by the 3D region growing process;
  • Figures 12 through 14 illustrate part of an original 3D image with the corresponding segmented image and also the segmented image overlayed on the original image.
  • Figure 15 illustrates the performance of the segmentation based on a vessel and non- vessel classification of the segmented image of Figures 12 through 14 as performed by an independent observer;
  • Figures 16 through 18 illustrate stereo pair X-ray-like densitometric projections (DPs) of the DBS segmentation corresponding to the image of Figure 3 , where Figure 16 is the DP of all segmented vessels, Figure 17 is the anterior circulation, and Figure 18 is a combination of DP and shaded surface display of a hollow representation of the DBS segmentation of the anterior and posterior circulations;
  • DPs stereo pair X-ray-like densitometric projections
  • Figures 19 through 22 illustrate stereo pair X-ray-like DPs, based on the principles of the present invention, of the intracranial circulation of a patient with a posterior communicating artery aneurysm and a basilar tip aneurysm ;
  • Figures 23 through 24 illustrate stereo pair X-ray-like DPs of a contrast enhanced abdominal aorta in accordance with the present invention.
  • Figures 25 through 26 illustrate a set of CTA images including (A) axial collapse of segmented data structures and (B) shaded hollow-body reprojection in accordance with the present invention.
  • FIG. 1 illustrates a block diagram of an imaging apparatus 10 in accordance with the present invention.
  • Imaging apparatus 10 comprises an imaging device 12, a Z- buffer MIP processor 14, an image processor 16, a storage system 18, output device 20, and image viewing device 22.
  • Each element is connected with at least one other element within apparatus 10 or, alternatively, each element is connected to a common bus 23 that allows communication to and from each element with any other element within apparatus 10.
  • Imaging device 12 typically is a magnetic resonance imaging (MRI) system or a computed tomography (CT) system or a rotational X-ray CT angiography (XRCTA) system used in generating 3D images in a non-invasive manner of a subject, such as an object or a patient, which images can be stored in computer manipulable form for subsequent processing and display.
  • Z-Buffer MIP processor 14 and storage system 18 are typically selected as a personal computer-type system or work station.
  • One such system includes a personal computer based on the Advanced Micro Devices (AMD) Athlon 650 megahertz CPU with a storage system comprising 128 megabytes of RAM and a 20 gigabyte long-term storage hard drive.
  • AMD Advanced Micro Devices
  • Athlon 650 megahertz CPU with a storage system comprising 128 megabytes of RAM and a 20 gigabyte long-term storage hard drive.
  • the computer system can be selected from other known computer systems as previously mentioned, such as for example, a work station like that provided by Sun Computers, such as their UltraSpark Platform.
  • the computer system further incorporates the image processor 16.
  • One image processor suitable for use is provided by Mitsubishi and is known as the VP 5003D graphic accelerator.
  • the personal computer system operates in a Windows NT operating system environment.
  • Other operating systems include Windows 98 and Windows 2000, which are both provided by Microsoft of Redmond, Washington.
  • the invention can be implemented on an Apple compatible system running under OS 9 and configured in a comparable manner to the personal computer mentioned above.
  • Output device 20 can include such items as a printer system or removable storage device.
  • the printing systems typically utilized as output devices offer a high quality output such as those found in digital photography printers. Other printers, including low resolution laser printers, would also be acceptable in some situations. Typically, however, the use of better resolution in printing is desirable.
  • Image viewing device 22 typically is a video monitor. The monitor typically is a high resolution monitor given the large amount of image detail that is of interest.
  • the depth-buffer segmentation process makes it feasible to include a graphical user interface to allow a user to select different viewing options, different angles of view, rotational viewing options, and other animation-type viewing options, which enhance and optimize the viewing needs of the user.
  • apparatus 10 As well as the process for converting raw image data information into a useful depth-buffer segmentation (DBS) image, is shown in the flow diagram of Figure 2 and is presented in greater detail below in accordance with the present invention.
  • DBS depth-buffer segmentation
  • An image In order to begin processing an image, an image must be generated or taken of the subject matter to be examined.
  • the image or images can come from various imaging modalities. These modalities can be selected from magnetic resonance imaging systems or other 3D imaging systems. Other imaging modalities include computed tomography (CT), computed tomography angiography (CTA), and X-ray CT angiography
  • CT computed tomography
  • CTA computed tomography angiography
  • image data acquisition was performed as part of an aneurysm imaging protocol.
  • the imaging was performed on a 1.5 Tesla (T) General Electric model Signa MRI scanner with actively shielded gradients.
  • the gradient system operates with a maximum strength of 23 milli- Tesla/meter (mT/m) and a slew rate of 77 milli-Tesla/meter/millisecond (mT/m/ms) on all three axes for the MRA sequence.
  • An optimized cylindrical birdcage transmit/receive coil with an RF mirror endcap is used.
  • the image is acquired in any desired technique, one of which is illustrated in the acquisition of step 212, which uses a three- dimensional time-of-flight (TOF) sequence.
  • TOF time-of-flight
  • the 3D TOF sequence incorporates imaging techniques such as abbreviated magnetization transfer, flow compensation in the frequency and slice-encode directions, and RF spoiling, which are understood by those skilled in the art.
  • imaging techniques such as abbreviated magnetization transfer, flow compensation in the frequency and slice-encode directions, and RF spoiling, which are understood by those skilled in the art.
  • step 214 applies zero filled interpolation in all 3 image directions.
  • step 216 applies MIP processing to the 3D image data.
  • MIP processing involves generating a 2D image from the 3D image, where the image value selected for each point in the 2D image is the maximum (or minimum) image value found along a corresponding line through the 3D image.
  • the MIP projection lines may all be parallel, as in a parallel ray MIP image, or the lines may be non-parallel and for example might diverge from a single source point as in a divergent ray MIP image or in curvilinear space as well.
  • an array is generated from the original 3D image data by determining the depth or Z-positions of the points included in the MIP image.
  • the value of each element is related to the distance of the corresponding point in the MIP image from the 2D image plane (i.e. the location of the maximum (or minimum) intensity value along each projection line).
  • Figure 3 illustrates a standard axial MIP image of a patient having an ophthalmic artery aneurysm. The image was acquired in a 512 x 256 x 64 image format and then interpolated to 1024 x 1024 x 128, which is shown in step 214. The image is generally considered to have high quality, but the artifacts inherent in the MIP algorithm, such as vessels obscured by other vessels, are visible.
  • Figure 4 displays the depth-buffer (or Z-buffer) array corresponding to the axial MIP image of
  • the Z-buffer array values displayed here as levels of brightness, are proportional to the depth or "Z" positions of the points that appear in the MIP image of Figure 3.
  • the generation of the MIP with the Z-buffer may actually be formed in any direction through the original 3D MRA image data.
  • a convenient direction that is chosen in the samples previously mentioned and that is used in the current embodiment is to perform the MIP in the original slice selection direction.
  • the path will be along a defined line taken for evaluation, whether it be a straight line, a divergent line, or a curvilinear line, as determined by the operator at the time the evaluation is performed.
  • Figure 5 illustrates a plot of signal intensities along the projection lines in the original 3D image data for 30 points taken from Figure 3.
  • the vertical marks on the 30 plots of Figure 5 show the positions and values of the points that projected into the MIP image in Figure 3.
  • Each of these lines corresponds to a point in the vertical bar that crosses the right middle cerebral artery shown in the display of the Z-buffer array of Figure 4.
  • the first is that the vessels exhibit very high continuity in position in the "Z" location. This continuity is used in a primitive vessel detection algorithm that generates a value based on the smoothness in at least one direction around a point.
  • the intensities in the image of Figure 7 are inversely related to the minimum ⁇ 2 value obtained from the first order fit of 5 points in the four principal directions around each point of Figure 6, which is a magnified view of the region indicated with a small rectangle in Figure 4.
  • the four principal directions are defined as 0, 45, 90 and 135 degrees from horizontal in the 2D image plane. For those points where the ⁇ 2 value is low in at least one direction (smooth neighborhood), the image is bright.
  • the vessel segments visible in the MIP of Figure 3 are all characterized by smoothness and continuity.
  • the process of performing depth-buffer segmentation consists in grouping image points based on this smoothness and continuity.
  • the major vessels are characterized by large groups of points.
  • the apparatus and process of the present invention measures the local roughness for each element within the Z-buffer array in accordance with step 220.
  • the process measures the roughness (or smoothness), as described below, in the Z-buffer array.
  • the process performs a low order polynomial least squares fit to Z-buffer array values (MIP image Z-locations) in the four principal directions around each point in the Z-buffer array. The four directions are shown in the upper left hand quadrant 600 of Figure 6. Beginning with the brightest point in the MIP image, every point in the MIP image is tested for classification as vessel based upon local smoothness in the corresponding Z-buffer array.
  • the process tests all eight neighbors in two dimensions for additional possible vessel elements. If the neighboring elements are also "smooth” and if they are close in “Z,” they are added to the group and their neighbors are added to the list for testing. The process continues until all neighbors to the group are too “rough” or too separated in "Z" to be connected.
  • the process then considers the next nonclassified, brightest point in the MIP image and tests whether it is a "smooth" element. If this element is smooth, it is defined as the next group and the method repeats the process of looking for additional members to add to this new group. The process continues until all elements in the Z-buffer array have been tested.
  • Figure 6 illustrates an expanded view of a small region of the Z-buffer array of Figure 4.
  • the four small line segments, segments 610-616 show the length (5 pixels) of the segment used for the first order least squares fit.
  • Points 610 and 614 are outside vessels and yield a high ⁇ 2 value for the fit in any direction.
  • Point 612 is within a vessel, but yields a high value except for the direction parallel to the vessel.
  • Point 616 is within the vessel and yields a low value for three out of four possible directions.
  • Figure 7 illustrates the same region as Figure 6, where the image brightness of each pixel is inversely proportional to its minimum calculated roughness value, or ⁇ 2 value, as specified below.
  • the P ijk corresponds to a predicted value for the same element determined from a low order polynomial fit to the values along the line.
  • V i5 is
  • the constants A and B are empirically determined.
  • the brightness values displayed in Figure 7 tend to increase with the likelihood that the element is part of a vessel.
  • Other methods of measuring smoothness in the MIP Z-buffer could be used.
  • the current method involves performing a low order least squares fit to a small number of image elements centered at a particular point.
  • the process utilizes a first order fit along five points, and an example of applying the fit in four principal directions is shown in Figure 6.
  • a different number of points, a different order fit, and a different number of directions may be used in measuring local roughness or smoothness as desired by the system designer.
  • the process could be omitted completely in some implementations.
  • Figures 12 through 14 illustrates points in the Z-buffer that are grouped together based upon their low roughness values and proximity.
  • the brightness shown in Figures 12 through 14 is proportional to the number of points in each group.
  • step 222 the process performs a grouping operation where each data point is considered in a specific order.
  • the use of the MIP image implies that the bright parts of the original three dimensional image data were the most important.
  • the process performs the connectivity operations by selecting the brightest image element in the MIP image. This element is tested for a low value in the corresponding minimum ⁇ 2 array. If the minimum ⁇ 2 is below a predefined threshold, the point is selected as a possible vessel and the neighboring points in the 2D MIP image are tested. The brightest element and all of its neighbors and their neighbors, and so forth, which satisfy the connectivity criterion, are added to the group. The process then considers the next non-classified brightest point and determines which remaining neighboring points satisfy the connectivity criterion to form another group. The process continues until all points have been tested.
  • the minimum ⁇ 2 value must be below a given threshold and the Z-position must be within a small distance of the currently considered point.
  • the threshold value for ⁇ 2 equaled 1.0 and a step of +/- 2 in Z was allowed.
  • Other values may be selected by the designer according to the needs of the designer, but this step size recognizes that larger vessels occasionally project from different regions within the vessel, and the larger value improves the character of the connectedness.
  • Figures 16 through 18 illustrates an image of the points that have been grouped based upon the
  • SU3STITUTE SHEET (RULE 26) proximity in “Z” and minimum roughness.
  • the intensity of the display is proportional to the number of points in each group. It is shown that some vessels, although obviously “connected” in real life, are not connected by this implementation of the process (e.g. the middle cerebral artery is not connected to the trifurcation area). These disconnects happen because only one point in a thick vessel is projected by the MIP algorithm, and the thickness of the vessel allows larger jumps in "Z” than are accepted by the connectivity criteria used in this example.
  • step 224 to complete the process of connecting all vessels, the groups are mapped back to the three dimensional image space, and all points from the two dimensional groups are considered again in the same order. All neighboring points in the 3D data are tested based upon their intensity being a selected (e.g. two standard deviations) above the average background value.
  • Points that are two standard deviations or more above the average background would have projected in the MIP image had they not been obscured by other brighter structures, such as vessels.
  • the process fills out the vessels in 3D and also connects most branches that were "disconnected" by being obscured by other vessels.
  • An image of points connected by the process in step 224 is illustrated in 11.
  • extraneous groups are removed from the set of grouped image points created in step 224.
  • noise groups can be removed automatically by eliminating all groups with very small numbers of elements.
  • all groups with less than 45 elements were considered to be noise and were removed.
  • All the major vessels subtend more than 45 elements, and very few noise groups have more than 45 elements (voxels).
  • groups with large numbers of voxels are not of interest, such as regions of fat or bone or organ tissue in angiographic imaging, they can be eliminated by a manual step, that allows the user to select the groups to be removed and cause those groups not to display.
  • a qualitative evaluation of the results of the segmentation process can be performed by comparing the elements segmented as belonging to vessels with those seen in the original cross-sectional images.
  • the DBS process performs well in classifying vessel and non-vessel voxels.
  • the misclassifications consist of the smallest vessels that did not appear in the MIP and the centers of the carotid arteries that were dark in the original 3D data.
  • Figures 12 through 14 show an example of points in the original 3D image that, as shown in Figure 15, were manually classified as vessel and non- vessel by an independent observer.
  • the points segmented by the DBS process of the present invention as vessels and non- vessels are shown as white and black regions of points in
  • Figures 12 through 14 The manually classified points appear as white points in these same figures.
  • Figure 15 illustrates graphs of the performance of the segmentation based on the vessel and non- vessel classification of the segmented image of Figures 12 through 14 as performed by an independent observer. The graphs represents the vessel inclusion sensitivity as measured by the number of voxels in a group of a few hundred and of xlOOO.
  • Line 910 represents very small vessels means only see in local projections.
  • Line 912 represents small vessels as seen in the MIP images.
  • Line 914 represents the medium size secondary bifurcations, M2, A2.
  • Line 916 represents large vessels, such as internal carotid and middle cerebral arteries.
  • Figure 16 is a stereo pair of the full intracranial circulation while Figure 17 illustrates a stereo pair of vessels connected to the anterior and middle cerebral arteries. It is difficult to show the full dynamic range of intensities contained in the images of Figures 16 and 17. The dynamic range is reduced by removing all points internal to the DBS segmented structures.
  • X-ray-like densitometric projection using the hollow DBS process in accordance with the present invention is shown in Figure 18. In Figure 18 some surface shading is also added to the X-ray-like projection image. In
  • the intracranial carotid artery element 1010 is visible below the aneurysm.
  • the process manipulates the dynamic range of information for display as shown in step 228 of Figure 2 to enhance the resolution of the vessel structures of interest during the display.
  • one example of performing dynamic range reduction is to eliminate all data points within the image that have neighbors in every direction. This results in an image where the vessels appear as hollow tubes.
  • Typical displays of densitometric projections through the hollow vessel images are shown in Figure 18 and in Figures 19 through 26.
  • the characterization of the image may be further modified by adding an amount of shading to each vessel surface thereby enabling the observer to determine whether the vessels are near or far, and the vessel orientation is more discernable as well.
  • the process displays the processed image on an imaging device, such as a video monitor and/or prints the image on a printing apparatus.
  • an imaging device such as a video monitor
  • Optional display processes are contemplated that may yield more visual information.
  • the present embodiment of the invention may not eliminate all noise, but the noise is reduced to the point that the useful data is easily recognized over these noise points.
  • Figures 19 through 22 illustrate stereo pair X-ray-like densitometric reproj ections of the results of the DBS segmentation of points from the 3D image of a patient with a communicating artery (PCOM) aneurysm and a basilar tip aneurysm.
  • Figure 19 is the cranial view of the PCOM aneurysm while Figure 20 is the caudal view.
  • Figure 21 highlights the posterior communicating artery aneurysm 1108 while Figure 22 illustrates the highlight of a small basilar tip aneurysm, which is behind tip 1110 and is shown from two different orientation.
  • Figures 23 through 24 illustrates a stereoscopic X-ray-like densitometric reproj ection through a renal artery study in accordance with the present invention.
  • FIG 23 A shows the anterior view while Figure 24 illustrates the posterior view.
  • Figures 25 through 26 represents an example of CTA images acquired at relatively low resolution on a helical CT scanner. Segmentation is performed with the DBS process in accordance with the present invention.
  • Figure 25 depicts the axial collapse of the segmented data structures
  • Figure 26 illustrates shaded hollow-body reproj ections of the aortic aneurysm.
  • the DBS process as described and presented in the present invention results in an image segmentation process that is readily applicable to magnetic resonance angiography (MRA), computed tomography angiography (CTA), rotational X-ray angiography (XRCTA), and other medical and non-medical applications. Since the DBS process is based upon the generic MIP algorithm, the application of the DBS process can be extended wherever the MIP algorithm is currently being used. The utility of the DBS process can be enhanced to include display options that allow the user to toggle between the DBS process and the preliminary MIP process, as well as other forms of volume rendering. The DBS process is also applicable to such fields as computer assisted screening and image interpretation based upon segmented anatomy.
  • MRA magnetic resonance angiography
  • CTA computed tomography angiography
  • XRCTA rotational X-ray angiography
  • the DBS process is also applicable to such fields as computer assisted screening and image interpretation based upon segmented anatomy.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Health & Medical Sciences (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)
  • Apparatus For Radiation Diagnosis (AREA)

Abstract

L'invention porte sur un procédé et un système permettant de générer de meilleures images de données multidimensionnelles à l'aide d'un procédé (218) de segmentation de tampon de profondeur. Ce procédé et ce système utilisés dans un système informatique modifient l'image en générant un ensemble de données de dimensions réduites à partir d'une image multidimensionnelle par formulation d'un ensemble de voies de projection passant par des points d'image sélectionnés dans l'image multidimensionnelle (212), sélectionner un point d'image le long de chaque voie de projection, analyser chaque point d'image pour déterminer des similarités spatiales avec au moins un autre point adjacent au point d'image sélectionné dans une dimension donnée, et grouper le point d'image avec le point adjacent (222) ou des similarités spatiales entre les points pour obtenir l'ensemble de données.
PCT/US2000/010532 1999-04-20 2000-04-19 Procede et appareil d'agrandissement d'image par optimisation et segmentation de donnees WO2000063831A1 (fr)

Priority Applications (4)

Application Number Priority Date Filing Date Title
AU46480/00A AU4648000A (en) 1999-04-20 2000-04-19 Method and apparatus for enhancing an image using data optimization and segmentation
US09/959,236 US6674894B1 (en) 1999-04-20 2000-04-19 Method and apparatus for enhancing an image using data optimization and segmentation
US10/692,133 US7120290B2 (en) 1999-04-20 2003-10-23 Method and apparatus for enhancing an image using data optimization and segmentation
US12/249,583 USRE43225E1 (en) 1999-04-20 2008-10-10 Method and apparatus for enhancing an image using data optimization and segmentation

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US13022699P 1999-04-20 1999-04-20
US60/130,226 1999-04-20

Related Child Applications (3)

Application Number Title Priority Date Filing Date
US09/959,236 A-371-Of-International US6674894B1 (en) 1999-04-20 2000-04-19 Method and apparatus for enhancing an image using data optimization and segmentation
US09959236 A-371-Of-International 2000-04-19
US10/692,133 Continuation US7120290B2 (en) 1999-04-20 2003-10-23 Method and apparatus for enhancing an image using data optimization and segmentation

Publications (1)

Publication Number Publication Date
WO2000063831A1 true WO2000063831A1 (fr) 2000-10-26

Family

ID=22443664

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2000/010532 WO2000063831A1 (fr) 1999-04-20 2000-04-19 Procede et appareil d'agrandissement d'image par optimisation et segmentation de donnees

Country Status (2)

Country Link
AU (1) AU4648000A (fr)
WO (1) WO2000063831A1 (fr)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6028955A (en) * 1996-02-16 2000-02-22 Microsoft Corporation Determining a vantage point of an image
US6031941A (en) * 1995-12-27 2000-02-29 Canon Kabushiki Kaisha Three-dimensional model data forming apparatus

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6031941A (en) * 1995-12-27 2000-02-29 Canon Kabushiki Kaisha Three-dimensional model data forming apparatus
US6028955A (en) * 1996-02-16 2000-02-22 Microsoft Corporation Determining a vantage point of an image

Also Published As

Publication number Publication date
AU4648000A (en) 2000-11-02

Similar Documents

Publication Publication Date Title
US6674894B1 (en) Method and apparatus for enhancing an image using data optimization and segmentation
Wilson et al. An adaptive segmentation algorithm for time-of-flight MRA data
US4945478A (en) Noninvasive medical imaging system and method for the identification and 3-D display of atherosclerosis and the like
EP3035287B1 (fr) Appareil et procédé de traitement d'images
US7676257B2 (en) Method and apparatus for segmenting structure in CT angiography
US7567696B2 (en) System and method for detecting the aortic valve using a model-based segmentation technique
EP1315125B1 (fr) Procédé de traitement d'image et système pour détecter des maladies
US7492968B2 (en) System and method for segmenting a structure of interest using an interpolation of a separating surface in an area of attachment to a structure having similar properties
US7209581B2 (en) System and method for ground glass nodule (GGN) segmentation
EP2116973B1 (fr) Procédé pour déterminer interactivement une surface liante pour segmenter une lésion dans une image médicale
US20040096088A1 (en) Method for coloring of voxels and image data processing and visualization system
US7684602B2 (en) Method and system for local visualization for tubular structures
US7136516B2 (en) Method and system for segmenting magnetic resonance images
CN1452089A (zh) Pet肿瘤图像的半自动分割算法
JP2010517601A (ja) 医療画像における病変部の特定及び分析
Gerig et al. Medical imaging and computer vision: An integrated approach for diagnosis and planning
US10546205B2 (en) System and method for multi-modality segmentation of internal tissue with live feedback
US7653225B2 (en) Method and system for ground glass nodule (GGN) segmentation with shape analysis
Masutani et al. Automated segmentation and visualization of the pulmonary vascular tree in spiral CT angiography: an anatomy-oriented approach based on three-dimensional image analysis
WO2000063831A1 (fr) Procede et appareil d'agrandissement d'image par optimisation et segmentation de donnees
Maduskar et al. Automatic identification of intracranial hemorrhage in non-contrast CT with large slice thickness for trauma cases
JP2007534352A (ja) すりガラス様小結節（ｇｇｎ）セグメンテーションを行うためのシステムおよび方法
Steen et al. Combined visualization of tumor and vessel geometry
Iacono et al. A method for morphological characterization of dural ectasia in Marfan syndrome
Chapman et al. Diagnostic fidelity of the Z-buffer segmentation algorithm: preliminary assessment based on intracranial aneurysm detection

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY CA CH CN CR CU CZ DE DK DM DZ EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT TZ UA UG US UZ VN YU ZA ZW

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): GH GM KE LS MW SD SL SZ TZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
WWE Wipo information: entry into national phase

Ref document number: 09959236

Country of ref document: US

REG Reference to national code

Ref country code: DE

Ref legal event code: 8642

122 Ep: pct application non-entry in european phase
NENP Non-entry into the national phase

Ref country code: JP