Gamma camera error correction using multiple point sources

ABSTRACT

A gamma camera system and method are described which use multiple point sources to detect inaccuracies in detector translational and rotational alignment. In practice of the method of the preferred embodiment, three capillary tubes, each containing a drop of an isotope, are located in different planes and locations with respect to the axis of rotation of the detectors. A SPECT acquisition is performed and the point source projection data is processed to calculate the point source coordinates, from which center-of-rotation correction factors may be calculated. These correction factors are applied by mechanical and software adjustments to the gantry and acquisition systems of the camera to correct for both translational and rotational inaccuracies.

INFORMATION

DETAILED DESCRIPTION OF THE INVENTION

There are a number of misalignment or COR errors that can develop in a gamma camera, as illustrated in FIGS. 1-. In these drawings two detectors heads, Head1 and Head2, are mounted for rotation by a gantry about a center-of-rotation indicated by crosshair which is coincident with the z axis. The detector heads can become misaligned in the axial direction as indicated by the misalignment of the dotted lines extending from the centers of the two detector heads in FIG. 1. The detector heads can become misaligned in the tangential direction as indicated by the lack of alignment of the dotted lines extending from the centers of the two detector heads in FIG. 1. Ideally, both dotted lines should intersect the z axis at the center of the crosshair . FIG. 1illustrates a radius error by the noncircular dotted line, indicating that the two detector heads are not following the identical concentric path when rotated by the gantry about the center of the crosshair . In FIG. 1the detector surface of the Head1 detector is not parallel to the z axis of rotation, referred to herein as a “tilt error.” In FIG. 1there is a rotation error of the Head1 detector; the axis of rotation of Head1 is misaligned in this example. This type of error is referred to as an “angle error.” FIG. 1is a plan view of the detector surface in which the detector head is rotated from its orthogonal alignment with the z axis, referred to herein as a “twist error.” The prior art approaches address only the translation errors of FIGS. 1and , and assume that the misalignment is constant for each angle setting of the detector heads.

The method of the present invention can be performed by mounting or suspending point sources in the field-of-view of a gamma camera as described more fully below. An acquisition sequence is then performed in which the detector head(s) are rotated about the COR by the gantry and gamma rays detected at each detector location. In a constructed embodiment there may by 64-128 such detector locations. The locations of the point sources in the field-of-view are detected from the gamma ray counts on the crystals and the position and angle of locational shifts of the point sources at different detector positions are measured. In a preferred embodiment these point source measurements are analyzed by a least squares approximation method to resolve the errors at each detector location. The detected errors are corrected by either mechanical adjustment of the head rotation or by correction factors in the reconstruction algorithms. For instance, the x,y coordinates of detected gamma ray events on the crystal can be corrected to x+Δx, y+Δy. Preferably these corrections are done on-the-fly during acquisition to correct locational shifts as the event data is acquired. An average correction value can be applied, but preferably event-by-event correction values are used.

A point source can be prepared as shown in FIG. 2. A syringe is used to put a tiny drop of a radionuclide in a capillary tube . As shown in the drawing, to approximate a point source in the typical camera image field, it is desirable to keep the size of the drop to 4 mm or less, and preferably less than 3 mm in all dimensions. A suitable radionuclide for many cameras is technetium (Tc-99m) with an activity level of 100 μCi or greater, and preferably in the range of 500 to 1000 μCi. A point source is easy and inexpensive to produce in this manner and can be properly disposed of relatively easily.

FIGS. 3and are axial and side views of an example of effective placement of the point sources 1, 2, and 3 in the field-of-view between camera detector heads and . The arrows indicate the direction of rotation of the detector heads during acquisition. The crosshairs mark the center of the x,y coordinates of the system, which also intersect the z axis of rotation. In the illustrated point source placement the sources 1, 2, 3 are seen by both detectors at every angle during the rotational ECT scan. None of the sources shares a transverse plane, that is, each source has its own unique z value. Preferably the z coordinates of the sources are separated by at least 10 cm and the sources are placed about 10 cm from the z axis of rotation in this embodiment. Preferably, the sources are distributed approximately evenly in space when viewed axially.

In a preferred method of the present invention a 360-degree SPECT study is performed with the point sources in the field-of-view using the largest matrix available. The number of stops should be 64 or greater. The result for each head is saved as a separate study.

The alignment errors for each detector head are determined based on the measured point source projection data. A flowchart illustrating this procedure is shown in FIG. . From the point source projection data the projection center of each point source (PSF) is determined (step ). A two-dimensional (2D) Gaussian surface fit program can be used to determine the projection center of each point source at each detector position. Using a 2D Gaussian surface fit can minimize the error in center determination due to asymmetry of the peak when the projection of a point source is close to an edge of the detector. FIG. 7 shows a flowchart of this procedure. The center of a point source projection, i.e., a peak, is accurately determined by two passes. First the approximate center is found by searching the position of the maximum pixel (step ). Assuming this is roughly the center of a peak, a rough estimation of the sigma (σ) of the peak is determined (step ). A square subset of the frame with the maximum intensity pixel at the center and of 6σ in size is extracted from the frame for the next pass, that is, the surface fit (step ). Using σ and the position of the maximum pixel as the initial values, the 2D Gaussian surface fit program determines the accurate center of the peak from the extracted sub-frame (steps , ). To find the center of the next peak, the surface fit of the current peak is subtracted from the frame before looping back to the first pass of the peak-searching program (step ). The peak-searching program generates a list of point source projection centers found from each frame ().

After the point source centers are found for the last frame, the integrity of number of source points found in all the frames is checked (step ). FIG. 6 shows a flowchart of this procedure. A check is performed of the number of centers found in a frame (step ). If there is a points missing for a frame, it may be an indication that either the point projection is outside the field-of-view or the source is too weak or the acquisition time is too short. If the frame is missing a center it is marked (step ). If there are more points found in a frame, it may be due to high noise level in the acquisition. Weak point sources or short acquisition times are likely causes of the high noise level. Centers that do not match points in other frames are discarded (step ). The procedure is continued until all frames have been checked.

After the point source projection centers are found for all the frames, the coordinates of the point sources can be determined using Equation (16) found in the Appendix below (step ). The center-of-rotation alignment errors for each frame can be determined using Equation (18) of the Appendix (step ). The alignment errors are stored in a table of rotational alignment data (Step ).

The alignment errors which have been determined are then used to correct the alignment geometry, either through mechanical adjustment to the gantry or correction factors in the processing software. In one typical embodiment, the translation errors Et and Es were corrected in the acquisition software. The detector angle error scan be corrected with control mechanisms in the detector head mounts. Rotational errors ζ and ξ are preferably adjusted mechanically by trained service personnel.

Appendix

The equations shown below are based upon two coordinate systems, one of spatial coordinates and the other of coordinates of the detector of the gamma camera. These are defined as follows:

Spatial coordinates include x-, y- and z-axis. When a patient is in a head-in, supine position on the patient table, the x-axis points horizontally to the left of the patient; the y-axis points upright; the z-axis points horizontally to the feet of the patient. The origin of the spatial coordinate system is fixed relative to the patient table.

Detector coordinates include t-, s-axis and detector angle. When a patient is in a head-in position on the table, and is viewed from the feet of the patient, 12 o'clock is the 0 degree of the detector rotation angle and clockwise rotation is positive. When a detector head is at 0-degree position, the t-axis of the detector is parallel to the x-axis of the spatial coordinates. The s-axis of the detector is always parallel to the z-axis of the spatial coordinates. The t-axis in the detector coordinate system is fixed relative to the detector head.

The distribution of activities is presented in spatial coordinates. The measured projections are presented in detector coordinates.

Ideal Alignment

Assume an ideal point source is positioned at (x0, y0, z0) in the spatial coordinates. The spatial activity distribution can be represented as:

()=δ(000)  (1)

When this point source is measured with a parallel-beam collimator on a detector, a projection generated in the detector coordinates should be:

(,θ)=(0 cos θ+0 sin θ,00 sin θ−y0 cos θ)  (2)

where h (t,s,d) is the point spread function at source-to-collimator distance d, R is the detector radius defined from the collimator surface to the center of rotation.

Assuming that h (t,s,d) is symmetric to the t- and s-axis, the resulting projection would be symmetric to the center of the spot generated by the point source in the detector space. Therefore the centroid coordinates can be used to represent the projection.

Equation (3) can be rewritten as

where r0={square root over (x2+y02)} and θ0=tan−1(x0/y0), or equivalently, x0=r0 sin θ0 and y0=r0 cos θ0.

Translation Error in Axial Direction

When an axial translation error is present, the measured t and s for the point source would be

where Es(θ) is the translation of the detector in the s direction as a function of detector angle θ. A positive Es causes the detector the shift in the s or z direction, as illustrated by Head1 in FIG. 1

Translation Error in Tangential Direction

With a translation error in the tangential direction, we have

where Et(θ) is the translation of the detector in the t direction as a function of detector angle θ. FIG. 1illustrates a positive Et of Head1.

Translation Error in Radius Direction

The effect of a translation error in the radius direction is negligible in parallel-beam SPECT studies. This error can be measured and needs to be addressed in fan-beam, cone-beam or MCD geometry systems.

Detector not Parallel to Rotation Axis (Tilt Error)

With tilt errors, we have

where ζ(θ) is the detector rotation angle relative to the z-axis as a function of detector angle θ and R is the radius of rotation measured from the collimator surface. u0 is defined as:

00 sin θ+0 cos θ  (8)

The rotation axis of the tilt error is on the surface of the collimator and is perpendicular to the gantry rotation axis. A positive ζ causes a positive s on the detector surface to tilt towards the gantry rotation axis. FIG. 1shows that Head1 has a positive ζ.

Inaccurate Angular Positioning (Angle Error)

With angle errors, we have

where ε(θ) is the error that causes the detector surface not to be facing the axis of rotation. The rotation axis of the angle error is defined on the surface of the collimator and is parallel to the gantry rotation axis. The direction of ε defined in the same way as θ. FIG. 1shows that Head1 has a positive ε.

Detector Rotated Relative to the Norm of the Detector Surface (Twist Error)

With twist errors, we have

where ξ(θ) is the detector rotation angle relative to the norm of the detector surface as a function of detector angle θ. The rotation axis of twist error is at the center of the detector surface and is perpendicular to the surface. When viewed from the gantry rotation axis towards the detector, a positive twist error causes a clockwise rotation of the detector surface.

The actual measurement should include a combination of all the errors:

Note that we dropped θ from the variables that are a function of θ to shorten the expression.

The task of measuring the misalignments is to determine Et(θ), Es(θ), ξ(θ), ζ(θ), and ε(θ) with measurements that can be performed in a clinical environment.

Estimate the Spatial Coordinates of a Point Source

Assume all the rotation errors are small. The following conditions are true:

Equation (10) is simplified to:

where d0=R−u0. If we put n point sources in the field of view and find the center {circumflex over (t)}mi and ŝmi for each of them at each angle position θ, we can estimate the position of the point sources (xi, yi, zi) by minimizing the least squared error for each point for all angles:

where Np is the number of projection angles, and θj is the projection angle at the projection j. We have

Resolving the above equations produces

Estimate Alignment Errors at Each Projection Angle

We can then calculate (t0, s0, u0) using Equations (3) and (8) for each point source i (ti, si, ui). tmi and smi are then derived from (ti, si, ui) using Equation (13). The errors can be estimated by minimizing the following

where tmi and smi are derived from (xi, yi, zi) using Equation (13). Therefore we have

It is straightforward to resolve the above linear equations and determine Et, Es, ξ, ζ, and ε for each angle θ. While these equations are optimized for parallel-beam geometry, similar equations can be readily derived in the same manner for fan-beam, cone-beam and coincidence mode geometries.

In the drawings:

FIGS. 1-illustrate the types of misalignment of center-of-rotation (COR) errors that may be encountered with a two-detector-head gamma camera;

FIG. 2 illustrates the preparation of a point source sample;

FIGS. 3and illustrate the placement of the point sources in the field-of-view;

FIG. 4 is a flowchart of a method for determining the COR values for a detector head from its point source projection;

FIG. 5 is a flowchart of a method for determining the centers of a point source projection in a frame; and

FIG. 6 is a flowchart of a method for checking the integrity of a number of point projections in a frame.

CLAIMS

1. A method for adjusting the accuracy of detected event locations in a gamma camera system comprising: positioning a plurality of point sources in the field-of-view of the camera; acquiring event data from the point sources from a plurality of detector positions; processing the event data to produce translational and rotational alignment adjustments; and producing gamma camera image information adjusted in accordance with the translational and rotational alignment adjustments.

2. The method of claim 1, wherein positioning comprises positioning three point sources in the field-of-view of the camera.

3. The method of claim 1, wherein positioning comprises positioning three capillary tubes each containing a radionuclide sample in the field-of-view of the camera.

4. The method of claim 1, wherein processing comprises processing the event data to produce a translational adjustment in at least one of an axial direction and a tangential direction, and to produce a rotational adjustment for at least one of a tilt error, an angle error, or a twist error.

5. The method of claim 1, wherein positioning comprises positioning a plurality of point sources in the field-of-view of the camera so that no two point sources share the same transverse plane and no three sources share an oblique plane parallel to the axis of rotation.

6. The method of claim 1, wherein acquiring comprises performing a SPECT acquisition to produce point source projection data for a plurality of detector positions.

7. The method of claim 1, wherein processing comprises extracting the center of each point source in a frame by Gaussian surface fitting.

8. The method of claim 7, wherein processing further comprises determining the spatial coordinates of points sources by fitting the centers to a sine curve in a transverse direction or a constant in the axial direction.

9. The method of claim 1, wherein processing comprises producing translational and rotational alignment errors by a least squared fit based on the gamma camera geometry.

10. A gamma camera system which can be compensated for errors in detector head translation and rotation comprising: a plurality of detector heads defining a field-of-view; a plurality of point sources positioned in the field-of-view; a gantry which acts to move the detector heads to different detector head positions during acquisition of event data from the point sources; an event data processor which acts to process the event data from the plurality of point sources to produce translational and rotational error information; and an image processor which acts to produce a gamma camera image compensated by the error information.

11. The gamma camera system of claim 10, wherein the plurality of point sources comprise three point sources.

12. The gamma camera system of claim 10, wherein the point sources are positioned such that no two point sources share the same transverse plane and no three sources share an oblique plane parallel to the axis of rotation.

13. The gamma camera system of claim 10, wherein the event data processor acts to produce translational and rotational error information for each of a plurality of different detector head positions.

14. The gamma camera system of claim 10, wherein the event data processor comprises a point source projection data processor.

15. The gamma camera system of claim 10, wherein the event data processor comprises a surface or curve fitting processor.

16. The gamma camera system of claim 10, wherein the event data processor comprises a processor which produces at least one of an axial direction error term and a tangential direction error term, and produce at least one of a tilt error term, an angle error term, or a twist error term.

COPYRIGHT

User acknowledges that Fairview Research and its third party providers retain all right, title and interest in and to this xml under applicable copyright laws. User acquires no ownership rights to this xml including but not limited to its format. User hereby accepts the terms and conditions of the License Agreement.