Method for determination of 3-D structure in biplane angiography

What is claimed as new and desired to be secured by Letters Patent of the United States is:

1. A method for determination of 3-D structure of an object in biplane angiography, comprising:

providing first and second imaging systems including first and second x-ray sources each having a focal spot and respective first and second x-ray sensitive receivers each defining an image plane, said first x-ray source arranged to transmit x-rays from its focal spot in a first direction through said object to the image plane of said first receiver and said second x-ray source arranged to transmit x-rays from its focal spot in a second direction arbitrarily selected with respect to said first direction through said object to the image plane of said second receiver;

determining the distances (D, D') of perpendicular lines from the focal spot of each x-ray source to the image plane of the respective receiver;

determining the points on respective image planes where respective perpendicular lines from the respective focal spots to the respective image planes intersect the respective image planes and defining said points on respective image planes as the origins of respective two dimensional image coordinate systems (uv), (u'v') at the respective image planes, wherein said focal spots and the respective image planes define respective first and second three-dimensional coordinate systems having respective z axes coincident with said perpendicular lines (D, D') in the directions of respective image planes, x axes parallel to respective of the image plane axes (u, u') and y axes parallel to respective of the image plane axes (v,v'), where the relative geometry of said first three dimensional coordinate system with respect to said second three dimensional coordinate system is defined by

x.sub.1 '=[R]{x.sub.1 =-t},

where x.sub.1 ', is the position vector of the object point (x.sub.i ', y.sub.i ', z.sub.i ') in said second three dimensional coordinate system, x.sub.1 is the position vector of the same object point (x.sub.i, y.sub.i,z.sub.i) in said first three dimensional coordinate system, [R] is a rotation matrix defining the rotation in three-dimensional space between the first and second three-dimensional coordinate systems and t expresses, in the first coordinate system xyz, a unit translation vector that moves the origin of the first coordinate system xyz to the origin of the second coordinate system;

irradiating said object with x-rays from said x-ray sources and producing respective first and second images defined by digital image data based on the x-rays received by said first and second receivers;

determining from each of said first and second images the image coordinates ((u, v), (u', v')) in the respective coordinate systems of N objects points, where N.gtoreq.8, which correspond to the same object points in the object;

scaling the image coordinates (u.sub.i, v.sub.i), (u.sub.i ', v.sub.i ') of said eight points by respectively dividing said image coordinates by the respective distances (D, D') to obtain normalized image coordinates (.xi..sub.i, .eta..sub.i), (.xi..sub.i ', .eta..sub.i'));

constructing N linear equations, one for each object point, containing only normalized image coordinates (.xi..sub.i, .eta..sub.i), (.xi..sub.i ', .eta..sub.i) and nine unknown elements (q.sub.kl), where q.sub.kl represent an unknown relative geometry between the two imaging systems in terms of nonlinear combinations of the elements of the translation vector (t) and the rotation matrix [R];

solving the N linear equations for eight of the q.sub.kl values relative to the ninth q.sub.kl value to produce a matrix [Q*]; and

determining scaled three dimensional coordinate (x.sub.i, y.sub.i, z.sub.i) of said N object points expressed in units of a unit translation vector t from the product matrix [Q*].sup.T [A*].

2. The method according to claim 1, further comprising:

defining at least two of said N object points as object points separated by a know distance;

determining the scaled distance between said at least two object points from the scaled three dimensional coordinates x.sub.1, y.sub.1, z.sub.1 ; x.sub.2, y.sub.2, z.sub.2) of said at least two object points;

determining the ratio of the scaled distance between said at least two object points and the known separation distance therebetween to derive a scaling factor;

producing an absolute t vector based on the quotient of the unit t vector and said scaling factor; and

determining absolute three-dimensional coordinates of object points using said absolute t vector, said rotation matrix [R] and said image coordinates ((u.sub.i, v.sub.i), (ui', vi')) in said image planes.

3. The method according to claim 2, further comprising:

selecting plural of said N object points as bifurcation points between tracked vessels of a vascular tree of said object;

defining approximate centerlines of vessel segments between said bifurcation points in each of said biplane images;

determining corresponding points in the two biplane images along said centerlines; and

determining from said corresponding points, [R] and t three dimensional positions of the points along the vessel segments between said bifurcation points.

4. The method according to claim 3, comprising:

displaying said three dimensional positions of the points along the vessel between said bifurcation points.

5. The method according to claim 1, comprising:

determining the absolute distance between the focal spots of said x-ray sources; and

multiplying the scaled three dimensional coordinates (x.sub.i, y.sub.i, z.sub.i) of said object points by the determined absolute distance between the focal spots to obtain absolute three-dimensional coordinates of said object points.

6. The method according to claim 5, further comprising:

selecting plural of said N object points as bifurcation points between track vessels of a vascular tree of said object;

defining approximate centerlines of vessel segments between said bifurcation points in each of said biplane images;

determining corresponding points in the two biplane images along said centerlines; and

determining from said corresponding points, [R] and t three dimensional positions of the points along the vessel segments between said bifurcation points.

7. The method according to claim 5, comprising:

displaying said three dimensional positions of the points along the vessel between said bifurcation points.

8. The method according to claim 1, further comprising:

selecting plural of said N object points as identifiable corresponding points of vascular structure in said first and second images;

tracking the approximate centerlines of the various vascular segments in both images between said identifiable corresponding points to define a complete vascular tree in both images;

determining polynominal fitting functions which represent the centerlines of said various vascular segments in both images;

selecting various points along the polynominal centerlines in one of said images;

for each of the selected various points, determining an auxiliary line which is a locus of points in the second of said images that represents the set of all possible points in the second image that correspond to the selected point in the first image;

determining mathematically the intersection of the corresponding polynominal centerlines in the second image with the auxiliary lines in the second image, in order to the determine the points in the said second image which corresponds with said selected various points in said first image;

determining from said corresponding points, [R] and t the absolute three dimensional positions of the said selected points along the vessel segments between said identifiable corresponding points.

9. The method according to claim 8, further comprising:

displaying the absolute three dimensional coordinates of all the selected points of said object;

displaying of absolute three dimensional information as a complete, connected vascular tree, composed of originally identified identifiable corresponding points, as well as selected points along vascular segment centerlines.

10. The method according to claim 8, further comprising:

displaying the absolute three dimensional coordinates of all the selected points of said object;

displaying of absolute three dimensional information as a complete, connected vascular tree, composed of originally identified identifiable corresponding points, as well as selected points along vascular segment centerlines.

11. A method for determination of 3-D structure of an object in biplane angiography, comprising:

providing first and second imaging systems including first and second x-ray sources having respective and second focal spots and respective first and second x-ray sensitive receivers each defining an image plane, said first x-ray source arranged to transmit x-ray from its focal spot in a first direction through said object to the image plane of said first receiver and said second x-ray source arranged to transmit x-rays from its focal spot in a second direction arbitrarily selected with respect to said first direction through said object to the image plane of said second receiver;

defining respective first and second three-dimensional coordinate systems, xyz and x'y'z', having respective origins located at respective of said first and second focal spots, and having respective z axes that are oriented toward the respective image planes and parallel to the respective line segments that are perpendicular to the respective image planes and intersecting the respective focal spots, where the relative geometry of said first three dimensional coordinate system with respect to said second three dimensional coordinate system is defined by

x.sub.i '=[R]{x.sub.i -t}

where, x.sub.i ' is the position vector of a point (x.sub.i ', y.sub.i ', z.sub.i ') in said second three dimensional coordinate system, x.sub.i is the position vector of the same point (x.sub.i, y.sub.i, z.sub.i) in said first three dimensional coordinate system, [R] is a rotation matrix defining the rotation in three-dimensional space between the first and second three-dimensional coordinate systems and t expresses, in the first coordinate system xyz, a unit translation vector that moves the origin of the first coordinate system xyz to the origin of the second coordinate system;

defining respective first and second image plane coordinate systems uvw and u'v'w', with origins located on the respective first and second image planes along the respective z and z' axes, and at distance D,and D', respectively, from the origins of xyz and x'y'z coordinate systems;

determining the distance D that separates the origin of the uvw coordinate system from the origin of the xyz coordinate system, and the distance D, that separates the origin of the u'v'w' coordinate system from the origin of the x'y'z' coordinate system, as being the respective perpendicular distances between said respective x-ray focal spots and image planes;

determining the positions on respective image planes where respective perpendicular lines from the respective focal spots to the respective image planes intersect the respective image planes and defining said points of intersection on respective image planes as the origins of the respective uvw and u'v'w' coordinate systems at the respective image planes;

irradiating said object with x-rays from said x-ray sources and producing respective first and second images defined by digital image data based on the x-rays received by said first and second receivers;

determining from each of said first and second images the image coordinates of N object points ((u.sub.i, v.sub.i), (u.sub.i ', v.sub.i ')) in terms of the respective image coordinate systems, where N >8, which correspond to the same object points in the object;

scaling the said first and second image coordinates ((ui, vi), (ui', vi')) of said N points by respective dividing said image coordinates by the respective distances (D, D') to obtain normalized image coordinates ((.xi..sub.i, .eta..sub.i), (.xi..sub.i ', .eta..sub.i ');

constructing N linear equations, one of each object point, containing only normalized image coordinates (.xi..sub.i, .eta..sub.i), (.xi..sub.i ', .xi..sub.i ') and nine unknown elements (q.sub.kl), and solving for the rotation matrix [R] and the unit translation vector t; and

determining the three-dimensional coordinates of said N object points (x.sub.1, y.sub.1, z.sub.1 ; x.sub.2, y.sub.2, z.sub.2), scaled to the length of the translation vector from the normalized image coordinates ((.xi..sub.i, .eta..sub.i), (.xi..sub.i ', .eta..sub.i ')) the rotation matrix [R] and unit translation vector t.

12. The method according to claim 11, further comprising:

defining at least two of said N object points as object points separated by a know distance;

determining the scaled distance between said at least two object points from the scaled three dimensional coordinates (x.sub.1, y.sub.1, z.sub.1 ; x.sub.2, y.sub.2 ; , z.sub.2) of said at least two object points;

determining the ration of the scaled distance between said at least two object points and the known separation distance therebetween to derive a scaling factor;

producing an absolute t vector based on the quotient of the t vector and said scaling factor; and

determining absolute three-dimensional coordinates of object points using said absolute t vector, said rotation matrix [R] and said image coordinates ((u.sub.i, vi), (u.sub.i 'm vi.sub.i ')) in said image planes.

13. The method according to claim 11, further comprising:

selecting plural of said N object points as identifiable corresponding points of vascular structure in said first and second images;

tracking the approximate centerlines of the various vascular segments in both images between said identifiable corresponding points to define a complete vascular tree in both images;

determining polynominal fitting functions which represent the said centerlines of said various vascular segments in both images;

selecting various points along the said polynominal centerlines in one of said images;

for each of the selected various points, determining an auxiliary line which is a locus of points in the second of said images that represents the set of all possible points in the second image that correspond to the selected point in the first image;

determining mathematically the intersection of the corresponding polynominal centerlines in the second image with the said auxiliary lines in the second image, in order to the determine the points in the said second image which corresponds with said selected various points in said first image;

determining from said corresponding points, [R] and t the absolute three dimensional positions of the said selected points along the vessel segments between said identifiable corresponding points.

Description Submit all comments and votes

BACKGROUND OF THE INVENTION

1. Field of the Invention

This invention relates to a method for determining 3-dimensional structure in biplane angiography from two biplane images obtained at arbitrary orientations with respect to an object.

2. Discussion of Background

The development of digital imaging techniques in the last twenty years has greatly expanded the field of radiology. Digital subtraction angiography (DSA) makes use of the digital format of the vascular images by the subtraction of a mask frame from an image containing contrast-filled vessels. The result is an image in which intervening structures and background have been removed. At present, DSA images are widely used in the diagnosis and treatment planning of most diseases of vessels, including atherosclerosis, aneurysms, arteriovenous malformations, etc.

The digital format of DSA also lends itself well to quantitative measurements of the vascular system. Many researchers have developed methods using single DSA images to quantify physical parameters such as vessel size, the amount of narrowing (or stenosis) of a vessel, or the rate of blood flow in a given vessel or supplying a given tissue. The application of all such quantitative methods is complicated by the fact that a single projection image of the vasculature provides little information concerning the true 3-dimensional vascular structure. Thus, the magnification of a vessel, which is a function of its relative 3-dimensional position between the x-ray source and the imaging plane, is difficult to derive from a single image. In calculations of vessel size and blood flow rate, the magnification of the vessel enters as the first and third power, respectively. (LE Fencil, et al, Accurate Analysis of Blood Flow and Stenotic Lesions by Using Stereoscopic DSA System, Medical Physics, 1987, 14, p. 460, presented at AAPM, 1987). In addition, the 3-dimensional orientation of the vessel with respect to the imaging plane is difficult or impossible to infer from a single image. Knowledge of the orientation of vessels is important for quantitative blood flow measurement, and is also important for the diagnosis of vessel malformations and for surgical planning.

In short, an accurate 3-dimensional (3-D) representation of the vascular structure would be very useful in many areas of medicine.

Several methods have been developed which derive 3-D information from two digital images. Stereoscopic digital angiography has been used in the calculation of 3-D position and orientation information of vessels (LE Fencil et al., Investigative Radiology, December 1987; and KR Hoffman et al., SPIE Medical Imaging, Vol. 767, p. 449, 1987). However, stereoscopic determination of 3-D vessel position becomes less accurate if the orientation of the vessel is close to the direction of the stereoscopic shift. Thus, the reliability of this method in determining 3-D vascular structure depends on the orientations of the vessels themselves.

Szirtes in U.S. Pat. No. 4,630,203 describes a technique for the 3-D localization of linear contours appearing in two stereoscopic images. However, this method also suffers from the limitation that the contour must not lie in the direction of the stereoscopic shift. In addition, a separate calibration step is required in this method to determine the 3-D locations of the x-ray sources relative to the imaging plane.

Several workers have developed methods to derive 3-D structure from two radiographic images that are obtained in exactly orthogonal directions (A. Dwata et al., World Congress in Medical Physics and Biomedical Engineering, 1985; and JHC Reiber et al., Digital Imaging in Cardiovascular Radioloqy, Georg Thiem Verlag, 1983). The 3-D information obtained with these techniques in binary: i.e., no gray levels remain in the reconstructed image of the 3-D object. Secondly, the images must be obtained in exactly orthogonal directions, which may be difficult to achieve in conventional biplane radiography systems. Also, determination of the positions of vessel segments which run in a direction perpendicular to one of the imaging planes is difficult or impossible with these methods.

To eliminate these problems, a method has been developed that allows calculation of 3-D vascular structure from two images obtained at arbitrary orientations (see JM Rubin et al., Investigative Radiology, Vol. 13, p. 362, 1978; and SA MacKay et al., Computers and Biomedical Research, Vol. 15, p. 455, (1982)). This method is important, but requires a separate, somewhat cumbersome calibration step which is executed either before or after imaging the patient. Specifically, a calibration object of known dimensions is imaged in the same biplane configuration as is used to image the patient. Data are then collected from the images of the calibration object, and parameters are calculated to provide the 3-D positions of selected points in the vascular images.

A different technique has been described in the field of computer vision to determine 3-D positions of object points from two arbitrary views without a calibration step. This method was described in two independent theoretical papers. (HC Longuet-Higgins, Nature, Vol. 293, p. 133, 1981; and RY Tsai, TS Huang, Technical Report, Coordinated Science Laboratory, University of Illinois, Nov. 12, 1981). It does not appear, however, that this approach has ever been successfully applied in the field of radiology.

SUMMARY OF THE INVENTION

Accordingly, one object of this invention is to provide a novel method for determination of 3-D vascular structure from two biplane images, wherein the biplane images can be obtained at arbitrary orientations with respect to an object.

Another object is to provide such a method which is rapid, requires minimal additional equipment, and is inexpensive and easy to implement in existing digital angiographic systems.

Yet another object of this invention is to provide a new and improved method for determination of 3-D structure in biplane angiography with minimal prior knowledge concerning the biplane imaging geometry and using biplane images of arbitrary relative orientation.

These and other objects are achieved according to the present invention by providing a novel method for determination of 3-D structure in biplane angiography, including determining the distances of perpendicular lines from the focal spots of respective x-ray sources to respective image planes and defining the origin of each biplane image as the point of intersection with the perpendicular line thereto, obtaining two biplane digital images at arbitrary orientations with respect to an object, identifying at least 8 points in both images which correspond to respective points in the object, determining the image coordinates of the 8 identified object points in the respective biplane unknowns based on the image coordinates of the object points and based on the known focal spot to image plane distances for the two biplane images; solving the linear equations to yield the 8 unknowns, which represent the fundamental geometric parameters of the biplane imaging system; using the fundamental parameters to calculate the 3-dimensional positions of the object points identified in the biplane images; and determination of the 3-D positions of the vessel segments between the object points. Thus, the entire 3-D structure of the vasculature is determined.

The present invention allows the determination of complete 3-D vascular structure from two biplane images obtained by arbitrary orientations. This method does not require a separate calibration step for each imaging geometry, and requires only minimal prior knowledge concerning the imaging system. Some of the mathematics utilized in this technique, although independently derived, bear strong similarity to the mathematics in the Longuet-Higgins paper above-noted and described hereinafter. However, the present method represents the first successful application of this theoretical method to the radiological field. Also described hereinafter is the solution of several problems which arise in the application of this technique to angiographic images. In addition, this technique is extended to one which will reconstruct the entire vasculature appearing in the two images. Previous theoretical reports describe a method for localizing points in 3 dimensions, but not for determining entire connected vascular structures.

The technique for the complete reconstruction of vascular structure from two biplane images according to the invention has many advantages. First, the use of biplane images (instead of stereo images, for example) is helpful because there are many more biplane digital systems presently installed than there are sterescopic digital systems. Secondly, because the present technique utilizes images from biplane systems that allow almost arbitrary positioning of the imaging projections, imaging geometries may be chosen for optimal visualization and reconstruction of the vessels of interest. This is a distinct advantage over methods employing stereoscopic or orthogonal projections. Varying the imaging geometry can ensure that vessel segments of interest will not run in a direction exactly perpendicular to one of the imaging planes, which is a problem in any 3-D reconstruction method employing two imaging projections. Third, the high spatial and temporal resolution of DSA makes the 3-D information derived from these images superior to information from tomographic techniques such as MRI and CT. In general, the pixel size in DSA images is smaller than that in MRI and CT images. Also, the time to acquire DSA images is generally equal to or shorter than that required for MRI or CT images. This is important in cardiac radiography, where moving vessels cause blurring of the image. Fourth, the hardware implementation of this method in, for example, digital cardiac angiography equipment allows real-time 3-D reconstruction of moving coronary vessels, and would be helpful in interventional procedures such as coronary angioplasty.

BRIEF DESCRIPTION OF THE DRAWINGS

A more complete appreciation of the invention and many of the attendant advantages thereof will be readily obtained as the same becomes better understood by reference to the following detailed description when considered in connection with the accompanying drawings, wherein:

FIG. 1 is a schematic block diagram illustrating an overview of the processing steps performed according to the invention;

FIG. 2 is a flow chart providing an overview of processing steps to obtain 3-D positions of object points and the geometric imaging parameters;

FIG. 3 is a flow chart illustrating in detail the processing steps performed in the determination of 3-D positions of object points and of the independent geometric imaging parameters;

FIG. 4 is a flow chart illustrating the processing steps in the determination of a complete 3-D vascular structure of step 4 shown schematically in FIG. 1;

FIG. 5 :s a schematic perspective view illustrating the geometry and coordinate systems of the biplane imaging system;

FIG. 6 is an illustration of two coronary angiographic images, with 8 (A-G) vessel bifurcation points selected, as well as an example of the calculated 3-D positions of a set of selected object points.

FIG. 7 is an illustration of the images shown in FIG. 6 with a vessel segment illustrated between two selected bifurcations and in which corresponding centerline points are indicated as well as an example of how the vessel segment might appear in 3 dimensions.

FIG. 8 is a schematic block diagram of an implementation of the invention for the determination of 3-D structure in a biplane digital angiographic system;

FIG. 9 is a schematic perspective view illustrating one of the biplane views shown in FIG. 5;

FIG. 10 is a schematic illustration of the inherent ambiguity of object scale, if the scale of the translation vector t is unknown;

FIGS. 11 and 12 are flow charts illustrating in more detail the steps in determining the exact and least squares solutions of the q vector, respectively, as shown in FIG. 10;

FIG. 13 is a schematic diagram of one x-ray tube and image-intensifier (I.I.) TV system in the biplane imaging system;

FIG. 14 is a schematic diagram showing the incorporation of the apparatus utilized to determine D, D' and the origins of the image coordinate systems, into the imaging system shown in FIG. 13.

FIG. 15a, 15b, 16a, 16b and 16c are schematic perspective views illustrating various geometrical parameters used as variables in a computer simulation study performed to test the feasibility and accuracy of the method of the invention;

FIG. 17 is a table showing the results of the computer simulation study; and

FIGS. 18 and 19 are graphs illustrating the effect of the number of object points and angle of rotation, respectively, on the average distance error in the computer simulation study, respectively.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

Referring now to the drawings, and more particularly to FIG. 1 thereof, implementation of this invention involves five basic steps:

Step (1) Initial calibration of the origins of the image coordinate systems uv and u'v', and of the distance between the x-ray source (x-ray focal spot), and the image plane, D and D'. This is accomplished using the calibration device described in relation to FIGS. 13 and 14 hereinafter. Note that although this calibration procedure is required, it is only done once for a given imaging system. It does not need to be repeated for different imaging geometries, as long as the relative positions of the x-ray focal spot and the image plane are maintained. The need for calibration may be obviated altogether in digital systems which are designed to incorporate the present invention.

Step (2) Acquistion of two digital biplane images of the object. This is accomplished using the setup illustrated in FIG. 8, wherein the lateral (LAT.) x-ray generator control, 122, and the AP (P) x-ray generator control, 124, control x-ray production in x-ray tubes, 102 and 104, respectively. Images of the object (or patient), 100, are acquired by image intensifier-TV (I.I.-TV) systems, 106 and 108. These images are digitized by A/D converters, 110 and 112, and are stored in the host computer, 114.

Step (3) Referring again to FIG. 1, identification of 8 (or more) points in the object which appear in both images, 3.1. These points may be selected manually, or with some automated detection scheme yet to be developed. These points would likely correspond to bifurcation points in the vasculature or to easily identifiable features on the vessels. From the N selected point, N linear equations are derived in terms of nine unknown q values, 3.2 The resultant [Q] matrix is then solved 3.3, followed by determination of the t vector (axis translation vector described hereinafter), 3.4, followed by determination of the [R] matrix (rotation matrix), 3.5, also described hereinafter. Step 3 is then completed by determining the 3-D positions of the object points, 3.6.

Subsequently, the 8 independent parameters which describe the biplane imaging geometry, as well as the 3-D positions of the object points, are automatically determined. The mathematics of this step are described in complete detail later in this document. Knowledge of the 8 independent parameters allows calculation of the 3-D positions of any object points whose coordinates can be identified in both images.

Step (4) Automated tracking of the vascular structure between the bifurcation points selected in Step 3. Subsequent to this, the automated correlation of many points along the vasculature in the two images, followed by calculation of the 3-D positions of each of these correlated points.

Step (5) Combination of the 3-D structure data of the object obtained in Steps 3 and 4. Referring now to FIG. 8, the presentation of the entire connected vascular structure on CRT displays, 118 and 120, in a projection format, from any desired orientation is possible. The vascular structure data could also be presented on a more advanced, true 3-D display, 116, for example a vibrating mirror display. Other manipulations of the 3-D data, such as rotation of the vessel structure in time on the CRT displays, 118 and 120, are possible.

The operations of Steps 2 and 3 are shown in the flow chart of FIG. 2 and in somewhat greater detail in the flow chart of FIG. 3. Step 4 is diagrammed in FIG. 4. Devices utilized for the 3-D structure display in Step 5 are diagrammed in FIG. 8. Step 1 will now be described in more detail.

Step (1) Referring to FIG. 1, the procedure and device utilized provide two types of information that are required for performance of the method of the present invention. First, the perpendicular distances D, D' between the respective imaging planes and the respective x-ray source, or focal spot, are determined. (See FIG. 5 which shows the geometry and coordinate systems.) D, D' may also be measured using a "yardstick" which is often present on digital imaging devices. Once the distance D, D' are determined, they usually can be kept constant for most imaging situations and do not have to be measured again. Secondly, the points on the imaging planes, where a perpendicular line extending from the x-ray focal spot to the imaging plane intersects the imaging plane, 35 and 36 in FIG. 5, are determined. These points are equivalent to the origins of the coordinate systems of the digital images. These also need be measured only once for a given imaging system. In imaging systems incorporating the present invention, the construction of the digital system could be designed to ensure that the origins of the coordinate systems would be in the center of the image matrix, and this calibration could be eliminated.

Referring now to FIGS. 13 and 14, imaging system utilized is an x-ray tube, 1301, and image-intensifier (I.I.)-TV system 1302. The x-ray tube and the I.I.-TV system are attached to the two ends of a "C-arm" apparatus, 1303. The x-ray tube mount, 1304, includes micrometer screws, 1305, which allow the x-ray tube to be moved in a plane parallel to the I.I.-TV input plane, 1306.

The apparatus utilized to determine D, D' and the origins of uv and u'v' coordinate systems is illustrated systems is illustrated in FIG. 14. The apparatus includes a top plate ,1401, and a bottom plate, 1402, which are parallel to each other. A similar device has used for of x-ray tube focal spots (see: doi eral., Medical Physics 1975; 2: 268). The distance between the two plates is adjustable with a micrometer screw, 1403, and is indicated on a ruler attached beside the micrometer screw, 1404. The top plate, made of 3 mm thick aluminum, contains 5 pinholes, each 0.5 mm in diameter, 1405. Four pinholes make a 5 cm square, and one pinhole is at the center of the square. The bottom plate is made of 3 mm thick plastic, and contains 5 lead shots 1.0 mm in diameter, 1406. Four shots make a 10 cm square, and one shot is at the center of the square. Thus, the large square in the bottom plate is twice the size of the small square in the top plate. The structure holding the plates is constructed to ensure that the centers of the top and bottom squares lie on a line perpendicular to the plates and that the sides of the squares on the top and bottom plates are parallel. This alignment device is attached to the input surface of the I.I. by holding clamps. The position of the pair of plates parallel to the I.I. input plane can be adjusted by two micrometer screws attached to the base of the apparatus, 1407.

The sliding x-ray tube mount and the apparatus are used in a procedure that involves two steps.

(1) First, the location in the image of the origin of the image coordinate system is determined. This origin corresponds to the location on the I.I.-TV input plane where a line perpendicular to the I.I.-TV input plane intersects the I.I.1-TV input plane. This may be accomplished initially by observation of fluoroscopic images or digital images obtained with the x-ray tube operated at approximately 80 kVp in the pulse mode. When the initial x-ray exposure is made, the image displayed on a CRT monitor shows five bright spots corresponding to the x-ray beams transmitted through the five pinholes, and also five dark spots corresponding to the shadows cast by the five lead shots. Since the half-value layer of an 80 kVp x-ray beam in aluminum is approximately 3 mm, the x-ray intensity that passes through the pinholes to form the bright spots will be approximately twice the x-ray intensity that is responsible for the uniform background observed on the CRT monitor. On the other hand, due to x-ray absorption of the lead spheres, the x-ray intensity responsible for the dark spots observed on the CRT monitor will be only about 0.05% of the x-ray intensity that produces the uniform background. Therefore, the bright and dark spots can be distinguished clearly from the uniform background and from each other.

The origin of the image coordinate system is determined by adjustment of either the micrometer screws at the base of the alignment device, or the micrometer screws in the x-ray tube mount, so that the bright spot from the central pinhole is exactly superimposed on the shadow cast by the central lead shot. The location of superposition in the digital image matrix is the origin of the image coordinate system. When these adjustments are completed, the focal spot is located on the line perpendicular to the I.I. input plane that passes through the element of the digital image matrix on which the central bright and dark spots are superimposed.

The second step determines the perpendicular distance between the x-ray tube focal spot and the I.I. input plane. The vertical position of the top plate of the alignment device is adjusted to match the locations of the four non-central bright spots to the locations of the four non-central dark spots. When these adjustments are completed, the distance from the focal spot to the bottom plate is equal to twice the distance from the top plate to the bottom plate. If the distance from the bottom plate to the I.I. input plane is not negligibly small, it can be estimated separately and added to the distance from the focal spot to the bottom plate to yield distance from the focal spot to the I.I. input plane, D.

This device and procedure described must be applied to both x-ray tube and I.I.-TV system pairs in the digital imaging system. In this way, the distances D and D', as well as the origins of the image coordinate systems uv and u'v', are determined.

In the following, and referring again to FIG. 1, is a qualitative description of the processes involved in Step 3.

After the two digital images are obtained and stored in the image memory, 8 (or more) easily identifiable object points appearing in both images are selected. As an example, an operator may select points using a cursor to specify the image coordinates of each object point. An example of selected points in two radiographic images, is shown in FIG. 6. The image coordinates of the selected points are relative to the origin of the image coordinate system, which was measured in Step 1 described hereinabove. If more than 8 points are selected, the mathematics in the solution of the [Q] matrix, 3.3 in FIG. 1, are overdetermined, and a least-squares solution (See FIG. 3, 308) is employed. This least-squares solution reduces the effects of measurement errors in the selected image coordinates of individual points. If the selected points correspond to bifurcation points on the vessels, subsequent determination of the entire vascular structure in step 4 is easier to accomplish.

As mentioned above, a complete description of the concepts and the mathematics behind the determination of the 8 independent parameters from the image coordinates of 8 (or more) points is provided hereinafter. However, it is helpful to summarize the basic concepts here. To begin, it is first observed that there exists an inherent redundancy in biplane imaging. Specifically, for every object point of unknown 3-D position there exist 4 knowns and 3 unknowns the 4 knowns being the image coordinates (u, v) and (u', v') of the point appearing in the two images, and the 3 unknowns being the x, y, z coordinates of the object in 3-D space. This observation allows one to conclude that some information concerning the geometry of the biplane imaging system may be gleaned from the 4 image coordinates of an unknown object point.

The second observation is that, if the input quantities described in Step 1 above are known, the biplane imaging system may be completely described by 8 basic parameters. These parameters describe the rotation and translation involved in transforming one of the imaging views into the other. Thus, the goal of Step 3 is to determine these 8 parameters using the redundancy described above.

If one redundancy is obtained with each object point identified in both views, then it follows that a minimum of 8 object points must be identified to determine the 8 basic parameters of the biplane imaging system. If more than 8 points are identified, the 8 basic parameters are themselves overdetermined, and may be calculated using a least-squares method. Once the 8 parameters are determined, the biplane imaging system is described completely, and the 3-D position of any object point which appears in both views can be easily calculated.

Next described is the method for determination of the 8 basic parameters that are determined from the known image coordinates of 8 (or more) object points in Step 3. In brief, the image coordinates of each object point are first scaled using the distance between the x-ray focal spot and the imaging plane, which was measured in Step 1. Combinations of the scaled image coordinates of a single point then serve as coefficients in a single linear homogeneous equation in 9 unknowns. The 9 unknowns are themselves combinations of the 8 basic parameters. At least 8 object points are selected, so at least 8 linear homogeneous equations are formulated. The matrix of the coefficients for the 9 unknowns for all the selected object points is referred to hereafter as the [A] matrix. Because all of these equations are equal to zero, it is mathematically possible to solve only for 8 unknowns relative to the 9th. referring to FIG. his is accomplished using an exact solution, 307, or a least squares solution, 308. Once the 8 unknowns are determined relative to the 9th, the 8 basic parameters of the system are determined from these 8 unknowns, and the 3-D positions of the selected object points may be easily calculated from their image coordinates and the 8 basic parameters.

The following is a detailed description of the techniques employed according to the invention to determine 3-D positions of object points without prior knowledge of the biplane imaging geometry.

Here is presented the detailed mathematical analysis of the method used to determine the 3-D positions of eight or more object points as well as to determine the geometry of the biplane imaging system, from two two-dimensional (2-D) radiographic projection images made with x-ray sources located at different positions, as in biplane angiography.

The technique is based on the following assumptions:

(1) Different x-ray source positions--and for the general case, different orientations of the image recording plane--are used for the two 2-D images. The use of different recording planes for the two images distinguishes "biplane" radiography from "stereo" radiography, in which both images are recorded in the same plane.

(2) The perpendicular distances from the x-ray source positions to the recording planes of the 2-D images are known. In general, these distances may be different. The method for determining these distances is described in Step 1 hereinabove and in relation to FIGS. 13 and 14.

(3) In each of the two 2-D images, one can determine the point at which the line from the x-ray source perpendicular to the image recording plane intersects that plane. The method of determining this point in each respective image recording plane is described in Step 1 hereinabove and in relation to FIGS. 13 and 14.

(4) Within each image plane, the orientation of two orthogonal coordinate axes (defined by the orthogonal vectors u and v in one image and u' and v' in the other) may be chosen arbitrarily, but in such a way that the vector cross products w 32 u x v and w=u'x v' point away from the x-ray sources.

(5) For each of 8 or more points in the object, one can determine the corresponding image point coordinates in both 2-D images, where corresponding image point coordinates are defined hereinabove. If the images are formed by an image intensifier (I.I.)-TV system, for example, the image coordinates are assumed to have been corrected for distortion, such as "pincushion" distortion, where pincushion distortion is defined by nonlinear warping of the image coordinates resulting from a curved image recording plane, such as the curved surface of an I.I. The image point coordinates must be expressed in the same units as the distance between the source and the image recording plane; hence, in digital imaging, the pixel dimensions of the image must be known.

If and only if the absolute scale of the 3-D object is to be determined, one must