 |
Description  |
|
|
BACKGROUND OF THE INVENTION
The present invention relates to a method of finding the image positions of three or more "fiducial markers", which are small objects that show up as small bright spots in images such as, for example, computed tomography (CT) and magnetic
resonance (MR) images. This process of finding positions of fiducial markers in an image is also referred to as image space localization.
The images discussed above are "volume" images, meaning that they are sets of contiguous slices that span a volume of space. They are acquired by placing the patient into a CT scanner or an MR scanner. One or both of these imaging modalities or
any other such imaging modality may be used for a given patient for a given surgery.
Different imaging modalities provide different types of information that can be combined to aid diagnosis and surgery. Bone, for example, is seen best on x-ray computed tomography (CT) images, while soft-tissue structures are best seen by
magnetic resonance imaging (MRI). Because of the complementary nature of the information in these two imaging modalities, the registration of MR images of the head with CT images is of growing importance for diagnosis and for surgical planning.
Furthermore, for the purpose of navigation during surgery it is helpful to be able to register images to the patient anatomy itself. Registration is defined herein as the determination of a one-to-one mapping between the coordinates in one space and
those of another, such that points in the two spaces that correspond to the same anatomic point are mapped to each other.
U.S. Pat. No. 4,945,914, U.S. Pat. No. 4,991,579, U.S. Pat. No. 5,142,930, and U.S. Pat. No. 5,230,338 disclose a method for utilizing fiducial markers to establish a coordinate system that facilitates the registration of image spaces and
physical spaces across time. The contents of U.S. Pat. No. 4,945,914, U.S. Pat. No. 4,991,579, U.S. Pat. No. 5,142,930 and U.S. Pat. No. 5,230,338 are incorporated herein by reference.
Briefly, these patents disclose using temporary or permanent markers that are imageable in the image space produced by a scanner. The markers may be attached to the skull via small posts that pierce the scalp and screw into the bone. The
markers may also be attached by first drilling a hole and then inserting via a self-tapping thread a base into the hole, to which a temporary marker is subsequently attached. In any case, since the posts or bases are physically attached to the skull,
the markers are referred to as "implanted" markers. Further, the markers are referred to as "external" markers, since the part of the marker that produces the bright spot in the image is outside the head.
In order to make full use of markers, they must be localized in the image space of the scan in question. Previous techniques have related to calling up successive images and manually locating spots whose brightness would appear to be indicative
of the presence of a marker. However, this is an error prone, time consuming and labor intensive process. Therefore, there remains a need for the further development of more automated techniques for localizing markers in images such as MR and CT volume
images.
SUMMARY OF THE INVENTION
Broadly speaking, the present invention uses a computerized program of operation to fully and automatically locate and localize attached markers in an image volume. The image volume is first searched for "seed points", which are candidate voxels
that lie within candidate (i.e., potential) markers. Next, the region around each candidate voxel (in the original image) is examined to discard false positives from the set of candidate voxels and to determine their centroids more precisely.
The present invention relates to an automatic technique for finding the centroids of fiducial markers attached to the head in, for example, magnetic resonance (MR) or X-ray computed tomography (CT) volume images, or images of any other imaging
modality. A marker's centroid is referred to as its "fiducial point". These fiducial points are used, for example, to register MR images with CT images.
The fiducial markers found using the technique of the present invention may be of any size or shape. However, in a preferred embodiment of the present invention, the fiducial markers are cylindrical fiducial markers.
The localization technique according to the present invention is fast, automatic, and knowledge-based, and includes two major portions. The first part of the method according to the present invention searches the entire image volume for
candidate voxels that lie within candidate markers. It is preferred that the dimensions of the marker (e.g., the inside height h and diameter d of a cylindrical fiducial marker) and the three dimensions of the voxels that make up the image are known,
that the intensity of the marker is higher than its surroundings, and that its image consists of one connected component. First, noise reduction via median filtering is applied if the image is an MR image, and the volume is subsampled to reduce
subsequent computation. A binary image is then formed in which voxels whose intensities are similar to air are set to background. For MR, in order to remove spurious detail from within the head, any background pixels recursively not connected within
the slice to the edge of the image are set to foreground.
A morphological opening, which is a nonlinear operation on binary images that causes changes in the shape of the foreground objects, is performed on the brighter components using an element slightly larger than a marker to remove small objects
(including markers). The morphological opening is followed by a morphological dilation using a smallest possible element that consists of, for example, one voxel. Morphological dilation includes complementing the image by setting all foreground
information to background and all background information to foreground, rotating the image 180 degrees, performing a morphological erosion, and complementing the resulting image. A morphological erosion includes placing a structuring element
successively at all possible positions within an image, noting the center pixel of the element when the entire element lies within the foreground, and generating a new image in which each noted position is set to foreground and other positions are set to
background. Alternatively, an erosion may be performed with an element approximately equal in size to a marker followed by a dilation with an element slightly bigger than the erosion element. A three dimensional connected-components labeling is
executed on the objects that were removed by the morphological opening. Finally, the centroid of each connected component is determined. Each such centroid serves as a candidate voxel.
The second part of the method according to the present invention examines the region around each candidate voxel (in the original image) to discard false positives from the first part of the method and to determine their centroids more precisely. First, a local threshold is determined to segment the voxels within a spherical region of radius R equal to the greatest straight-line distance between any pair of points on the marker and centered on the candidate voxel. For example, for a cylindrical
marker of height h and diameter d, the radius R=(h.sup.2 +d.sup.2).sup.1/2. The determination of the threshold is accomplished through a knowledge-based search as follows. S(t) is defined as the set of voxels in this region whose intensity i is greater
than the threshold t and which are connected recursively (in 3-D) to the starting point, V(t) is defined as the volume of the set of voxels S(t), and V.sub.h is the maximum volume and V.sub.t is the minimum volume that are allowed for a marker image.
The method according to the present invention first searches for the smallest t such that no voxel included in the set of voxels S(t) is farther than R from the starting point. If V(t)>V.sub.h, t is increased until V(t)=V.sub.h. If V(t)<V.sub.t
the segmentation fails and the candidate voxel is discarded. If the segmentation succeeds, the fiducial point f is calculated on the basis of the final threshold t.sub.f as f=.SIGMA.(i-i.sub.0)r/.SIGMA.(i-i.sub.0), where the sum is taken over all voxels
in S(t.sub.f), i.sub.0 is the intensity of an empty voxel, and r is the three dimensional position vector of a voxel.
After these two major parts of the method of the present invention are implemented, the fiducial points localized in the image volume are ranked according to the average intensity of the marker voxels from brightest to darkest. The n brightest
marker voxels, where n is the number of markers known to be present in the volume, are declared to be fiducial points. Thus the number of false positives is less than or equal to the number of false negatives.
BRIEF DESCRIPTION OF THE DRAWINGS
Other features and advantages of the present invention will become apparent from the following description taken in conjunction with the attached drawings, wherein:
FIG. 1 illustrates a temporary fiducial marker assembly which may be used in implementing an embodiment of the present invention;
FIG. 2A and FIG. 2B illustrate an example of the placement of fiducial implants in the anatomy which may be localized using the method of the present invention;
FIG. 3 illustrates an imaging system which may be used in implementing an embodiment of the present invention;
FIG. 4 is a flowchart illustrating a first portion of an embodiment of the present invention;
FIG. 5 is a flowchart illustrating a second portion of an embodiment of the present invention; and
FIG. 6 is a flowchart illustrating an embodiment of the localization technique of the present invention.
DETAILED DESCRIPTION
Referring now specifically to the drawings, FIG. 1 illustrates a temporary fiducial marker assembly which may be used in implementing an embodiment of the present invention. The temporary fiducial marker assembly illustrated in FIG. 1 includes a
base 10 and an imaging marker 20.
The base 10 has a threaded portion 12 at a first end. The threads 12 enable a surgeon to securely attach the base 10 into the skull or other desired portion of bone tissue of a patient. Other connecting structure is provided to securely and
releasibly link the imaging marker 20 with the base 10. For example, in the illustrated embodiment, the end of the base 10 opposite the threaded portion 12 terminates in a socket head 14 which contains a socket-like recess 16. (It is anticipated that
the base would be implanted into bone with the aid of an insertion tool that twists the base into the bone or into a hole provided in the bone. The recess is non-circular so as to better transmit the torque provided by such an insertion tool.) Just
beneath the socket head 14 are a plurality of grooves 18 (i.e., three grooves around the circumference of the base portion 10). As shall be further explained below, the socket 14 and the grooves 18 provide for the secure and releasable attachment of the
imaging marker portion 20 with base portion 10.
The imaging marker portion 20 of the temporary fiducial marker assembly may consist of two principal portions, a cylinder 22 and a cap 24. The cylinder 22 contains a cavity 26 (shown by dotted lines in FIG. 1) for receiving a mixture of imaging
agents whose composition is determined by the imaging modalities to be employed. While in this version, the vessel containing the imaging agents is preferably cylindrical to simplify the process by which the centroid of the corresponding volume of
imaging agent is determined, other shapes (such as a box or sphere) could be employed as well. The cylinder 22 is closed at one end and open at the other end to allow for the introduction of the imaging agents. In one version of the device, a cap 24 is
used to seal off the open end of the cylinder 22 once the imaging agents have been added to the cylinder. In this version, the cap may be cemented or welded into place. The cap may be provided with a plug portion 28 that protrudes into and thereby
helps seal off the cavity 26 of the cylinder 22 against leakage of the imaging agents. Other conventional filling and sealing techniques, such as blow-molding, ultrasonic welding, or heat sealing, may be used.
Where a cap is employed, it may be provided with a protruding boss 30 and a plurality (e.g., three) of snap arms 32, which terminate with inwardly projecting portions 34. The shape and dimensions of the boss 30 are in direct correspondence with
the shape and size of the socket 16 provided in the base 10 to properly and securely center the imaging marker on the base. The snap arms 32 cooperate with the grooves 18 of the base 10 to detachably secure the imaging marker onto the base. While this
example shows the use of snap arms, other fastener structures may be provided for attaching the marker to the base (e.g., screw threads, clasps, hooks, etc.).
The dimensions of the temporary fiducial marker assembly will be somewhat dependent on the state of the art of imaging. The greater the sensitivity of the scanner employed, the lesser the quantity of imaging material necessary to provide a
suitable image, which in turn makes it possible to reduce the corresponding size of the marker that must be employed to contain the imaging material. The fiducial marker assembly including a base portion approximately 12 mm in length and 2 mm-3 mm in
diameter is sufficiently large to provide for the secure placement of the base into the bone beneath the skin. When the clinician prepares the patient for imaging, the base portion is exposed and an imaging marker approximately 6 mm in length is
attached to the base; the marker itself may protrude from the scalp and be exposed to air while a scan is performed on the patient. The base and the imaging marker housing are constructed of a bio-compatible organic polymer, such as polyether imide.
While an example of a fiducial marker which may be used in implementing the present invention has been described as set forth above, it is noted that the present invention is not limited to such a fiducial marker. For example, the fiducial
marker used in implementing the present invention need not necessarily be a cylindrical fiducial marker. Any fiducial marker may be used in implementing the present invention which is imageable in an image space produced by a scanner such as, for
example, a magnetic resonance (MR) or computer tomography (CT) scanner.
In order to practice the present invention, a three-dimensional internal coordinate system must be set up which is fixed within a human anatomy. The internal coordinate system is established within the anatomy by fixing three or more fiducial
implants (i.e., fiducial markers) to portions of the anatomy. The fiducial markers are fixed to points which will not change their spacial relationship to one another over a relatively long period of time, such as a few months or more.
An example of placement of fiducial implants in the anatomy is shown in FIG. 2A and FIG. 2B. In these figures, fiducial implants 40a, 40b and 40c are implanted in three separate, spaced locations within the skull 42 of a patient.
Since these three fiducial implants 40a, 40b and 40c are arranged in a noncollinear manner, a plane is formed which contains these fiducial markers 40a, 40b and 40c. Once a plane is defined, a three-dimensional coordinate system is defined. Any
point within the body will be within the internal coordinate system.
Although fiducial implants are shown, any three points that are affixed with respect to the region of interest can comprise the three points used to define the internal coordinate system. However, fiducial implants 40a, 40b and 40c that are
identifiable and measurable by different imaging systems, such as CT imagers and MR imagers are preferred. As described above, the fiducial markers 40a, 40b and 40c may be relatively small and unobtrusive so that no discomfort or self consciousness will
be experienced by the patient even though the patient may carry the implants for a relatively long period of time.
A number M of markers are attached to the patient's head before images of the head are acquired (M is equal to 3 in FIG. 2A and FIG. 2B). The markers are left in place during the imaging and are not removed until after the completion of the
surgical procedure that follows the imaging.
FIG. 3 illustrates an imaging system which may be used in implementing an embodiment of the present invention. Additional details of such an imaging system are included in U.S. patent application Ser. No. 08/162,986 filed on Dec. 8, 1993 and
which has been incorporated herein by reference. After imaging, but before the surgery, the images are loaded into the memory of a computer 52 illustrated in FIG. 3, which is referred to as the "planning workstation" and a method is executed by the
planning workstation computer 52 to find the position of each of the markers within each of the images. The term "planning workstation" relates to the fact that the computer 52 provides additional programs that the surgeon may use to plan the surgical
approach to be used on the patient. Because the volume image is three dimensional, the specification of a position requires three coordinates, typically referred to as x, y, and z. At this point, the method of the present invention is implemented to
provide image space localization of the markers in the image space.
The marker positions, which we call the "positions in image space" or "image space positions", are transferred from the planning workstation 52 to a computer 54 located in the operating room, which is referred to as the "intra-operative
computer". This transfer of the marker positions may be accomplished by means of a diskette, optical disk or tape, etc. which is written by the planning workstation and read by the intra-operative computer, or by means of a network connection 56 between
the planning workstation and the intra-operative computer. If N volume images are acquired, there will be M.times.N positions transferred from the planning workstation 52 to the intra-operative computer 54, one for each marker and each volume image, or
3.times.M.times.N coordinates, since three coordinates are required for each position. In addition, some or all of the N volume images of the patient may be transferred from the planning workstation 52 to the intra-operative computer 54.
In the operating room, after the patient's head has been securely fastened to the operating table in the standard manner and before the surgery begins, the physical positions of the markers are measured. These physical positions are referred to
as the "positions in physical space" or "physical space positions". The measurement of a physical position is referred to as "physical space localization". Physical space localization is accomplished with a precision instrument such as an
"intra-operative localization device" (or ILD) 58 that is connected to the intra-operative computer 54. The ILD 58 employs a hand-held wand (or pointer) 60 with which the surgeon touches each marker. The ILD is able to determine the position of the tip
of the pointer 60 at any time. When the surgeon has positioned the wand 60 so that it is touching a marker, the surgeon presses a button that causes the intra-operative computer 54 to write that position into its memory 62. The process of measuring the
physical positions of the markers is completed by touching each of the M markers with the wand 60 and recording its position in the memory 62 of the intra-operative computer 54.
The exact position of each of the markers may be established, for example, by providing a divot (not illustrated) formed in the top of the cylinder 12 of the imaging marker portion 10 of the fiducial marker assembly. The tip of the pointer 60 of
the ILD 58 may then be accommodated within the divot so that an exact position of the marker is established.
At this point two or more separate sets of coordinates are stored in the intra-operative computer: one set of 3.times.M physical space coordinates and N sets of 3.times.M image space coordinates, where M is the number of markers and N is the
number of images stored in the planning workstation computer 52. Two sets of coordinates will almost never be the same, because the orientation of the head relative to the image scanners will almost always differ from each other and from the ILD.
However, by comparing the set of physical space coordinates with one of the sets of image space coordinates a computer implemented method can | | |
