311, and the third cold rod is substantially parallel to the travel vector 311. The cold rods may be fixedly embedded in third, fourth and fifth holes (302a, 302b, 302c) in the phantom housing 307 such that they are stable and do not move, particularly during the multi-modality image acquisition process. In some implementations, third hole 302a is disposed along a top edge of first plate 310a, fourth hole 302b is disposed on second plate 310b and fifth hole 302c is disposed on third plate 310c. First cold rod embedded in third hole 302a may be oriented nominally vertically, while second cold rod embedded in fourth hole 302b may be oriented nominally left-right horizontally and third cold rod embedded in fifth hole 302c may be oriented nominally axial horizontally. First and second cold rods are in a plane at the rear of the phantom 206 (i.e., away from gantry 202). The cold rods may be used to measure the pitch and yaw angles of one of the cameras (e.g., CT camera) with respect to the table travel vector 311.

FIG. 3b shows a top view of an exemplary phantom 206. The top view shows a square opening 320 formed by the upper edges 322a-d of first, second, third and fourth side plates (310a, 310b, 310c, 310d). In some implementations, there are two vertical rods (i.e., first removable hot rod 305a and first cold embedded rod 330a). A first hole 304a is disposed in the center of upper edge 322c of third side plate 310c to receive the first vertical hot rod 305a. A third hole 302a is disposed in the center of upper edge 322a of first side plate 310a to receive the first vertical cold rod 330a.

FIG. 3c shows a front view of third plate 310c disposed at the front of the phantom 206, while FIG. 3d shows a cross-sectional view along line A-A in FIG. 3c. As shown, phantom 206 has a housing 307 that forms a shell. The housing 307 is supported by three legs 208a-c. Second cold rod 330b may be inserted into fourth hole 302b in a left-right horizontal direction, while third cold rod 330c may be inserted into fifth hole 302c in an axial horizontal direction below second cold rod 330b.

Third cold rod 330c may be placed low (e.g., ⅓ the height of phantom 206) along the left side (when the hot rods are facing gantry 202). In some implementations, third cold rod 330c is shorter (e.g., 10 cm) than both first and second cold rods 330a-b (e.g., 15 cm). Third cold rod 330c may extend axially from the rear of the phantom 206 less than all the way (e.g., about ⅔ of the way) to the front of the phantom 206. This placement of the third cold axial horizontal rod 330c provides asymmetry, which advantageously facilitates robust recognition of the location and orientation of the phantom 206 in the image volume acquired by the first camera (e.g., CT camera), such that the placement of the phantom 206 by the user may be validated prior to the longer image acquisition by the second camera (e.g., PET camera) of the multi-modality imaging system 120.

FIG. 3e shows a front view of fourth plate 310d disposed at the right side of the phantom 206. There are two left-to-right horizontal rods (second removable hot rod 305b and second cold embedded rod 330b) shown. Second cold rod 330b may be inserted into fourth hole 302b in a left-right horizontal direction, while second hot rod 305b may be inserted into second hole 304b in a left-right horizontal direction. Both second hole 304b and fourth hole 302b may be at the same height from the base of the housing 307.

FIG. 3f shows a cross-sectional view along line B-B in FIG. 3e. Second cold rod 330b may be inserted into fourth hole 302b in a left-right horizontal direction, while third cold rod 330c may be inserted into fifth hole 302c in an axial horizontal direction below second cold rod 330b. In some implementations, third cold rod 330c extends from the rear of the device (away from phantom 206) all the way to the front of the phantom 206. The ends of the second cold rod 330b may be capped with, for example, epoxy or other suitable material.

FIG. 4 shows an exemplary method 400 of image alignment. It should be understood that the steps of the method 400 may be performed in the order shown or a different order. Additional, different, or fewer steps may also be provided. Further, the method 400 may be implemented with computer system 101 of FIG. 1, multi-modality imaging system 120 of FIG. 2, phantom 206 of FIGS. 3a-f, a different system, phantom or device, or a combination thereof.

At 402, a first image of imaging phantom 206 is acquired using a first camera of first modality (e.g., CT) of multi-modality imaging system 120. The phantom 206 may include two removable hot rods 305a-b and three embedded cold rods 330a-c, as previously described. The phantom 206 is positioned on patient table 204. For instance, a user may place the phantom 206 on the patient table 204, then move the patient table 204 into the gantry 202 and turn on the laser alignment system of the multi-modality imaging system 120. The user may align the orientation markings 340 on the phantom 206 in all six degrees of freedom with the laser markings from the laser alignment system, adjusting the height of the patient table 204 as needed. A scan protocol may be loaded and initiated in multi-modality imaging system 120 to acquire the first image of the phantom 206. The first image is, for instance, a three-dimensional (3D) image volume. Analysis module 107 may reconstruct the first image using, for example, maximum intensity projection (MIP) or other suitable reconstruction algorithm. The reconstructed image may then be preprocessed to remove unwanted features, such as the patient table 204 and legs 208a-c.

FIG. 5 shows exemplary slices 502-506 of an exemplary reconstructed and preprocessed first image. Slice 502 shows the plan view, slice 504 shows a side elevation view and slice 506 shows a front elevation view from the gantry 202. Segments 510a-b are the only segments that will visible in the second image.

Returning to FIG. 4, at 404, analysis module 107 determines tip centroids near top points and bottom points (or tips) of the two vertical rods in the first image. The two vertical rods may be a vertical hot rod 305a and a vertical cold rod 330a. It is not known at this point which rod is the hot rod 305a and which rod is the cold rod 330a.

Analysis module 107 may locate the top points of the two vertical rods by examining horizontal image slices from the top down. Even in the worst case of the phantom being placed 10 degrees away from the orthogonal line, there is a range of slices in which the tops of the two vertical rods 305a and 330a are the only features in the slice. Top points of the two vertical rods 305a and 330a are located at two spots with intensity above a pre-determined threshold value. At the worst-case angle, the tops of the two vertical rods 305a and 330a may be separated by a vertical distance of, for example, 38 mm. The bottom points of the vertical rods 305a and 330a may similarly be located by examining horizontal slices from the bottom up. For each vertical rod (305a, 330a), the 3D distance between the top and bottom points may be, for example, 150 mm±10 mm.

After the top and bottom points (or tips) are found, each rod may be traced slice-by-slice, moving vertically, as the peak location in each cross-sectional slice should be near (e.g., within a few millimeters of) the peak location in the adjacent slice. Following the peak locations, slice-by-slice, a range of slices located with a predetermined range of distance (e.g., between 10 mm to 20 mm) from the end of each tip may be identified for each vertical rod. For each of these ranges of slices, a Gaussian fit may be used to identify the centroid of each slice. The average of all the centroids of all of the slices may be determined in each of these 10 mm segments to create a single centroid near the top point and a single centroid near the bottom point of each rod.

At 406, analysis module 107 determines the next location of the phantom 206 for acquiring a second image using a second camera of second modality (e.g., PET) in the multi-modality imaging system 120. In some implementations, the next location of the phantom 206 may be determined based on the previously determined first tip centroids of the vertical rods. If the x position of the single cross-section of a rod in the first image is to the right of the average of the four x values for the first tip centroids (as viewed from the gantry 202), then the phantom 206 is placed with the “hot” rods 305a-b towards the gantry 202. In some implementations, the next position may be determined by calculating the average z location of the vertical rod closer to the gantry 202 plus half the field-of-view (FOV) length of the second camera minus the default Z offset. The tips of this rod that is closer to the gantry 202 is identified as the “hot” vertical rod 305a, and the tips of the other rod is identified as the “cold” vertical rod 330a.

Else, if the x position of the single cross-section of a rod is to the left of the average of the four x values for the centroids of the tips of the vertical rods (as viewed from the gantry 202), the phantom 206 is rotated 180° (in the horizontal plane), and the “hot” rods are away from the gantry 202. The next position for acquisition of the second image may be determined by calculating the average z location of the vertical rod further from the gantry plus half the FOV length of the second camera minus the default Z offset. The tips of this rod are identified as the “hot” vertical rod 305a, and the tips of the other rod is identified as the “cold” vertical rod 330a.

At 408, analysis module 107 determines second tip centroids of horizontal rods (305b and 330b-c) in the first image. In some implementations, to determine the second tip centroids, the shorter axial cold rod 330c are eliminated from the first image by deleting everything lower than a pre-determined distance (e.g., around 75 mm) below the top tip of the vertical cold rod 330a. The same way the vertical rods (305a, 330a) were found earlier, but working in sagittal slices inward from the left and the right, analysis module 107 may identify the left and right tips of the hot and cold horizontal rods (305b, 330b). Analysis module 107 may identify which horizontal rod is “hot” and which is “cold” by using the locations of the “hot” and “cold” vertical rods (305a, 330a). The distance between the tips of each rod may be, for example, 150 mm±10 mm. Analysis module 107 may determine a centroid for a segment from 10 mm to 20 mm from each tip, the same as was done for the vertical rods.

Returning to the original image containing the axial cold rod 330c, a cross-section of this rod may be located prior to acquiring the second image. From that location, analysis module 107 may work in coronal slices both forward and backward until the two tips of the axial cold rod 330c are found. The 3D distance between these two tips should be, for example, 100 mm±10 mm. Excluding the first slice at each tip meeting the threshold criteria, analysis module 107 may determine a centroid for a 10 mm segment near each tip, in the same manner as already done for the other four rods in the first image. The output may be six (x, y, z) points, two for each “cold” rod.

At 410, a second image of the phantom 206 at the determined location is acquired using a second camera of second modality (e.g., PET) of multi-modality imaging system 120. Prior to acquiring the second image, the phantom 206 may be moved to the determined location by using, for example, a patient-handling system (PHS) of the patient table 204. The second image may be a three-dimensional image volume. The second image may be acquired until sufficient image statistics are obtained.

At 412, analysis module 107 determines third tip centroids at or near the ends of each of the vertical and horizontal hot rods 304a-b. In some implementations, analysis module 107 examines the second image by horizontal slices from the top and from the bottom, and by sagittal slices from the left and from the right, until peak activity is found exceeding a pre-determined threshed value at each “tip”. Once those tips are located, analysis module 107 may continue processing slices until the last slice is found exceeding the pre-determined activity threshold value for each rod.

For each of the four segments representing the hot rods 304a-b, analysis module 107 may exclude the first slice in which activity above threshold value was found, and calculate a centroid using the remaining slices (e.g., 8 mm of length) with activity. For each cross-section, in each slice, the center may be calculated using a Gaussian fit. For the vertical rod, this should produce several (x, z) pairs, with the y coordinate given by the slice location. For the horizontal rod, this should produce several (y, z) pairs, with the x coordinate given by the slice location. For each of the four segments, analysis module 107 may calculate the average of all of its (x, y, z) points to determine the location (x, y, z) of a single centroid.

For the vertical rod, the 3D distance between the top centroid and the bottom may be, for example, 130 mm±10 mm. For the horizontal rod, the 3D distance between the left centroid and the right centroid may be, for example, 130 mm±10 mm. While calculating the centroids for each segment, analysis module 107 may determine the sum of all the activity concentrations of all the pixels of each cross-section used, then divide the sum by the number of slices used. This provides an average activity for each hot rod. If the lesser of these is less than 70% of the greater, the workflow may be failed for mismatched sources. The output may be four (x, y, z) points representing the tip centroids, two for each hot rod 304a-b.

At 414, analysis module 107 determines misalignment parameters of the first and second cameras based on the first, second and third tip centroids. The misalignment parameters may be used for compensation in future acquisitions by the multi-modality imaging system 120 as well as quality control. The misalignment parameters may include, but are not limited to, absolute angular offsets (e.g., pitch, yaw) of the first and/or second camera with respect to the table travel vector 311, as well as relative translational offsets (e.g., X, Y, Z) and rotational offsets (e.g., pitch, yaw, roll) between the two cameras.

FIG. 6 shows a table 602 of exemplary misalignment parameters that may be determined for a CT-PET imaging system using the present framework. Relative translational offsets (X, Y, Z) and relative rotational offsets (pitch, yaw, roll) from the PET camera to the CT camera may be determined based on data from the two hot rods (305a, 305b). Absolute pitch and yaw values from the CT camera to the PHS (or gantry) may be determined based on data from the cold rods (330a, 330b, 330c). Absolute pitch and yaw values from the PET camera to the PHS (or gantry) may be determined based on data from both hot rods (305a. 305b) and cold rods (330a, 330b, 330c). For example, {absolute PET pitch, absolute PET yaw}={absolute CT pitch, absolute CT yaw}±{relative pitch offset, relative yaw offset}.

A CT gantry angle may be determined. The CT gantry angle is the angle, in the pitch and/or yaw direction, of the CT gantry with respect to the table travel vector. Determination of the CT gantry angle is based on the skew of the image which occurs when the CT axis is not parallel to the travel vector of the table (or phantom) in the gantry. FIG. 7a illustrates a perfect alignment between the CT axis 704 and the travel vector 701 of the phantom in the gantry. In this illustration, the angle 702 between the two features (or rods) is 90° in the image acquired by the CT camera. FIG. 7b illustrates an imperfect alignment between the CT axis 704 and the travel vector 701 of the phantom in the gantry.

When the CT axis is not aligned with the travel vector 701, the angle 704 between the two features in the image acquired by the CT camera is changed by the same angle as the CT misalignment. It should be appreciated that the feature 706 parallel to the travel vector 701 is unaffected. It is not critical that the phantom itself be perfectly aligned with the travel vector (e.g., within 10° is fine). The features in the phantom need not intersect. The angle between the features of the phantom in the orthogonal projections should be known to serve as a reference for analysis of the acquired image. In an exemplary phantom, the angle between the features is 90°. Other known angles may also be used. This illustration applies equally for a side view (pitch angle) or a plan view (yaw angle).

In some implementations, the pitch angle of the CT camera with respect to the table travel vector 311 is determined by measuring the angle between the vertical and axial cold rods 330a and 330c in the first image. The difference from 90 degrees is the pitch angle of the CT camera with respect to the table travel vector 311.

FIG. 8 shows an elevation view of exemplary vertical and axial cold rods 330a and 330c. Point A represents the tip centroid of the axial cold rod 330c further from the gantry, while point B represents the tip centroid of the axial cold rod 330c nearer the gantry. Point C represents the tip centroid near the bottom of the vertical cold rod 330a, while point D represents the tip centroid near the top of the vertical cold rod 330a. There are two possible configurations (802, 804) of the lines AB and CD, wherein the vertical cold rod 330a represented by line CD is located near or away from the gantry.

To best visualize the angle between lines AB and CD, the lines may be translated (without changing their angles) so that they intersect at the origin (0, 0, 0), as shown in configuration 806. Line AB may be translated to line A′B′, wherein A′=(0,0,0) and B′=B−A. Line CD may be translated to C′D′, wherein C′=(0, 0, 0) and D′=D−C. The pitch angle of the CT camera with respect to the gantry may be determined by determining the difference of angle W from 90 degrees, where angle W is the angle between the two lines A′C′ and C′D′. If the CT camera is leaning towards the PHS, angle W is less than 90 degrees. If the CT camera is leaning away the PHS, angle W is greater than 90°. The difference of angle W from 90° is the estimate of the CT-gantry pitch angle.

In some implementations, the yaw angle of the CT camera with respect to the travel vector is determined by measuring the angle between the left-right horizontal and axial cold rods (330b, 330c) in the first image acquired by the CT camera, and determining the difference of this angle from 90°.

FIG. 9 shows a plan view of exemplary left-right horizontal and axial cold rods 330b and 330c. Point A represents the tip centroid of the axial cold rod 330c further from the gantry, while point B represents the tip centroid of the axial cold rod 330c nearer the gantry. Point C represents the tip centroid near the left end of the horizontal cold rod 330b, while point D represents the tip centroid near the right end of the horizontal cold rod 330b. There are two possible configurations (902, 904) of the lines AB and CD, wherein the horizontal cold rod 330b represented by line CD is located near or away from the gantry.

To best visualize the angle between lines AB and CD, the lines may be translated (without changing their angles) so that they intersect at the origin (0, 0, 0), as shown in configuration 906. Line AB may be translated to line A′B′, wherein A′=(0,0,0) and B′=B−A. Line CD may be translated to C′D′, wherein C′=(0, 0, 0) and D′=D−C. The yaw angle of the CT camera with respect to the travel vector may be determined by determining the difference of angle V from 90 degrees, wherein angle V is the angle between the two lines A′C′ and C′D′. If the CT camera is rotated counterclockwise, angle V less than 90 degrees. If the CT camera is rotated clockwise, angle V is greater than 90°. The difference of angle V from 90° is the estimate of the CT-gantry yaw angle.

In some implementations, the relative translational and rotational offset parameters between the PET camera with respect to the CT camera may be determined based on the hot rods