Method and apparatus for area-specific MR-based attenuation correction of time-domain filtered PET studies

Abstract

A method (200) for attenuation correction (210) of a plurality of time-domain filtered PET images (222) of a moving structure, comprising: Creating (240) an ACF matrix (242) based on a base MR image (232) of the structure; Developing a plurality of image transformations (262), each of the image transformations (262) registering the base MR image (232) to a corresponding cine MR image (252) of the moving structure; applying each of the image transformations (262) to the ACF matrix (242) to generate a corresponding film ACF matrix (272); Time-correlating each of the cine MR images (252) and its associated cine ACF matrix (272) with one of a plurality of time-domain filtered PET data sets (222); and Generating an attenuation-corrected time-domain filtered PET image (212) from each of the time-domain filtered PET data sets (222).

Description

TECHNICAL BACKGROUND AREA

[0001] Embodiments of the invention generally relate to improving the quality of images obtained from PET-MR scanners. Certain embodiments relate to attenuation correction of PET images. DISCUSSION OF THE STATE OF THE ART

[0002] Positron emission tomography ("PET") instruments use one or more rings of scintillators or other detectors to generate electrical signals from gamma rays (photon pairs) produced by the recombination of electrons within a target material and positrons emitted from the decay of the radionuclide packaged in a tracer. Typically, recombination events occur within about 1 mm of the radionuclide decay event, and the recombination photons are emitted in generally opposite directions to arrive at different detectors.Paired photon arrivals occurring within a detection window (usually separated by less than a few nanoseconds) are counted as indicating a recombination event, and on this basis, computed tomography algorithms are applied to the scintillator position and detection data to localize the different recombination events, thereby generating three-dimensional images of the tracer distribution within the target material.

[0003] Typically, the target material is a body tissue, the tracer is a liquid analogous to a biological fluid, and the radionuclide is distributed primarily in body tissues that utilize the biological fluid. For example, a common form of PET uses fluorodeoxyglucose ( 18 F), which is analogous to glucose, with one of the hydroxyl groups that usually compose glucose replaced by the 13F radionuclide. Brain mass, kidneys, and growing cells (e.g., metastatic cancer cells) preferentially absorb both glucose and fluorodeoxyglucose. Therefore, PET is quite useful in oncological studies, for localizing specific organs, and for studying metabolic processes.

[0004] One challenge in achieving the desired PET image quality is that gamma rays, in the energy spectrum produced by positron-electron interactions, are easily attenuated by typical body tissues and attenuated differently by different tissues. Different attenuation will alter the probability of detecting recombination events from the patient, thereby confounding the process of generating an image from PET data. Therefore, it is highly desirable to provide means for attenuation correction ("AC").

[0005] For example, PET is often combined with computed tomography ("CT"), which uses a moving X-ray source and detectors to obtain images of internal structures. X-rays are photons and are attenuated much like the higher-energy photons produced by positron-electron recombination events. However, unlike PET photons, the source strength of CT photons is known. Therefore, CT image data are a direct measurement of photon attenuation between the source and detectors. Therefore, CT image data provide a useful basis for AC of concurrent PET imaging.

[0006] Combined PET-CT scans have proven particularly useful for breath-domain-filtered studies, such as studies performed during normal breathing to diagnose or evaluate lung cancer. Breath-domain-filtered images can reduce motion blur that would otherwise result from acquiring the PET image across a single respiratory cycle. However, the correct reconstruction of breath-domain-filtered PET studies requires domain-filter-specific attenuation correction information. Simultaneous acquisition of CT and PET data supports domain-filter-specific AC.

[0007] For example, consider two elements of a PET detector, with the coincidence line between the detectors passing through the patient's lower chest. During maximum inspiration, this line may pass through the lower lobes of the lungs, which attenuate 511 keV photons by a relatively small amount. As the patient exhales, the liver moves upward into the chest, so that at the end of exhalation, the same coincidence line now passes through soft tissue, which attenuates the 511 keV photons much more than the lungs. If the PET data is region-filtered so that there are different images acquired at maximum inspiration and end exhalation, each would benefit from an attenuation correction corresponding to the actual distribution of attenuating tissue during that phase of the respiratory cycle.

[0008] Increasingly, and for a variety of reasons, including lifetime radiation dose reduction goals, PET / CT scans are being replaced by combining PET scans with magnetic resonance imaging (“MR”). This new combination (“PET-MR”) presents new technical challenges. For example, while CT forms an image based on the detection of X-rays emitted from a source through the target, MR forms an image based on the detection of rotating “relaxation” magnetic fields generated within a target by nuclei that have odd atomic numbers—i.e., the total number of neutrons and protons is not divisible by two—in response to fluctuations in an imposed magnetic field. Therefore, MR measures a phenomenon fundamentally different from the photons detected by CT and PET.

[0009] One advantage of MR is that magnetic fields do not attenuate in body tissues, so nuclear locations can be determined (using Fourier analysis) based solely on frequency shifts between the applied magnetic field and the response field. Another advantage is that, by carefully selecting pulse sequences, different tissues or materials can be highlighted. Accordingly, MR is often used to distinguish tissue types within a patient and is also used to identify fine detail structures. Typically, different pulse sequences are used for tissue differentiation. For example, a T1 pulse sequence can be used to obtain images with water appearing darker and fat appearing brighter. Conversely, a T2 pulse sequence can be used to obtain an image with fat darker and water brighter.

[0010] Therefore, a single device combining PET and MR (a "PET-MR scanner") can provide fine details, tissue differentiation, and metabolic data. However, because MR signals do not attenuate in the same way as PET or CT signals, and because MR signal recovery is highly dependent on the type of pulse sequence used (with each pulse sequence emphasizing a different material), whereas the PET signal is attenuated by each layer of material intervening between a recombination event and a pair of detectors, single-scan MR imaging data does not necessarily provide a reliable basis for AC from degraded PET images.

[0011] MR imaging has the ability to acquire a volume of images over a single respiratory cycle in a manner known as retrospectively region-filtered cine MRI. However, the types of MR images that can be acquired in a cine-like manner have not been considered suitable for generating PET attenuation correction. The MR protocols conventionally considered effective for generating PET attenuation correction values, such as 2-point Dixon water-fat scans, typically require more than 10 seconds to acquire and are therefore not suitable for acquiring many images during a single respiratory cycle. Instead, these MR imaging data were typically acquired during a long breath-hold to minimize motion blur and other respiratory artifacts.However, an MR image in which the breath was held does not show the intermediate lung positions required for accurate AC in a PET scan. Patent application DE 10 2012 218 289 A1 describes a method for generating a motion-corrected PET image of an examination region in a combined MR-PET system, as well as a corresponding MR-PET system. The article "Investigation of MR-Based Attenuation Correction and Motion Compensation for Hybrid PET / MR" calculates an MR-based 4-dimensional attenuation map by combining an ultrashort echo time (UTE) image with a person-specific motion model, with the motion model being derived from a second, near-real-time 3D MR image acquisition (Buerger et al., IEEE Transactions on Nuclear Science, Vol. 59, No. 5, Oct. 2012, pp. 1967). SHORT DESCRIPTION

[0012] The invention is characterized by the features of the independent claims. Specific embodiments are characterized by the features of the dependent claims. Embodiments of the invention provide a method for attenuation correction of a plurality of time-domain filtered PET images of a moving structure.The method includes creating an ACF matrix based on a baseline MR image of the structure; developing a plurality of image transforms, each of the image transforms registering the baseline MR image to a corresponding cine MR image of the moving structure; applying each of the image transforms to the ACF matrix to generate a corresponding cine ACF matrix; time-correlating each of the cine MR images and its corresponding cine ACF matrices with one of the plurality of time-domain filtered PET datasets; and generating an attenuation-corrected time-domain filtered PET image from each of the time-domain filtered PET datasets.

[0013] Aspects of the invention provide an apparatus for generating a plurality of attenuation-corrected, time-domain filtered PET images. The apparatus includes a PET-MR scanner; a controller communicatively connected to the PET-MR scanner; and an imaging processor communicatively connected to the controller. The controller is configured to acquire a baseline MR image of a structure, acquire a sequence of cine MR images of the structure, and acquire a plurality of time-domain filtered PET data sets of the structure.The imaging processor is configured to generate a baseline ACF matrix based on the baseline MR image, register the baseline MR image with at least some of the cine MR images, generate cine ACF matrices based on the registration of the baseline MR image with the cine MR images, and attenuation correct the time-domain filtered PET datasets based on the cine ACF matrices.

[0014] Other aspects of the invention provide a method for attenuation correction of PET images. The method includes obtaining, in a first MR acquisition, a baseline MR image suitable for deriving PET attenuation correction factors; and, during a second MR acquisition, obtaining a cine sequence of MR images during a respiratory cycle concurrent with the time-domain filtered PET study. The method further comprises non-rigidly (elastically) registering the baseline MR image to each of the cine MR images to generate a group of registered MR images corresponding to PET datasets from the respiratory-domain filtered PET study and usable for deriving attenuation correction factors. The registered MR images are then used to form attenuation correction factors for their corresponding time-domain filtered PET datasets. DRAWINGS

[0015] The present invention will be better understood by reading the following description of non-limiting embodiments with reference to the attached drawings, in which: Fig. 1 shows a conventional method for attenuation correction of PET images based on simultaneous CT images. Fig. 2 shows an inventive method for attenuation correction of PET images based on contemporaneous and baseline MR images in accordance with a first embodiment of the invention. Fig. 3 shows an exemplary process for registering a base MR image to a cine MR image in accordance with an embodiment of the invention. Fig. Figure 4 shows an equation for mutual information of a baseline MR image and a cine MR image. Fig. 5 shows another method for attenuation correction of PET images based on contemporaneous and baseline MR images in accordance with a second embodiment of the invention. Fig. 6 shows an apparatus for implementing inventive methods for attenuation correction of PET images based on contemporaneous and baseline MR images in accordance with embodiments of the invention. DETAILED DESCRIPTION

[0016] Reference will be made in detail below to exemplary embodiments of the invention, examples of which are illustrated in the accompanying drawings. Where possible, the same reference numerals used in the drawings refer to the same or similar parts without repetition of description. Exemplary embodiments of the present invention will be described with reference to combined PET-MR scanners, although embodiments may be adapted for use with other imaging systems.

[0017] Aspects of the invention relate to improving the simultaneous collection of PET and MR images by providing attenuation correction (AC) of PET images based on the MR imaging data.

[0018] In PET / CT studies, a conventional procedure excludes 100% of respiratory area filtered attenuation (as in Fig. 1) involves obtaining 110 a sequence of cine CT images 112 from the same anatomical region and during the same time period as acquiring 120 a sequence of time-domain filtered PET data sets 122. The sequence of cine CT images 112 may be acquired at rates on the order of one image per second, allowing the acquisition of six to ten images during a typical respiratory cycle, so that the set of CT images can then be correlated with the filtered set of raw PET data files 120.

[0019] The sequence of CT images 110 is used to derive 130 a sequence of attenuation correction factor matrices (“ACF matrices”) 132, and then each ACF matrix is ​​time-correlated 140 with one of the time-domain filtered PET data sets 122 to reconstruct 150 a set of attenuation-corrected (“AC”) PET images 152.

[0020] Fig. 2 shows a method 200 according to a first embodiment of the invention for MR-based attenuation correction 210 of a sequence of respiratory-domain-filtered PET data sets 222 to generate attenuation-corrected PET images 212. In addition to acquiring 220 the PET data set 222, the method 200 also includes acquiring 230 a tissue-discriminating or "baseline" image 232 suitable for later generating 240 a three-dimensional matrix 242 of photon attenuation correction factors ("baseline ACF matrix"). In one embodiment, the baseline MR image 232 is obtained as a substantially static image using a patient breath-hold protocol and a scanning protocol suitable for tissue discrimination studies—e.g., a 2-point Dixon scan that clearly highlights water and fat and requires more than ten seconds to complete.Although the baseline MR image 232 may, in certain embodiments, be obtained as an average image using tissue discrimination study during normal respiration, or as a composite image of several different single tissue scans during respiration or during a breath hold, it can be expected that an average image or a composite image will not be optimal for the subsequent generation of the static ACF matrix 242.

[0021] Based on the tissues identified in the baseline MR image 232, standard photon attenuation calculations for the typical positron recombination energies are undertaken to generate the baseline ACF matrix 242.

[0022] In addition to generating 240 the base ACF matrix 242, the method 200 also includes acquiring 250 a sequence of cine MR images 252 and registering 260 each of the cine MR images 252 back to the static MR image 232. In the process of registering 260, a plurality of image transformations 262 are generated, each of the image transformations registering a corresponding one of the cine MR images 252 to the static MR image 232. In one embodiment, the image transformations are non-rigid (elastic). For example, Vemuri, et al. Med. Image Anal. 7:1-20 (2003) describes a "level set" algorithm for elastic image registration. Other mutual information-based algorithms (e.g., as taught by Woods (1992) or by Viola and Wells (1994)) can also be used.In most embodiments, the cine MR images 252 are obtained using different (faster) scanning protocols than those used to obtain the baseline MR image 232—e.g., zero-TE or ultrashort-TE scanning protocols. Also, in certain embodiments, different scanning protocols are used for different cine MR images 252. Accordingly, registration 260 may also include pixel intensity scaling or shifting to match tissue types between different scanning protocols.

[0023] For example, a locally perturbed (non-rigid) rigid body registration 260 based on mutual information as in Fig. 3, in which one of the film images 252 is represented as a differentiable cubic spline function f T(x), which maps pixel intensities to a linear pixel array that can be interpolated at any intermediate pixel position within the set of sampled values ​​that compose the image. Similarly, the base image 232 is represented as another differentiable cubic spline function f R (x). A smoothed composite histogram of µ = arg min S(f R , f T og(•|µ)), which relates the film image 252 back to the base image 232, is defined as a cross product of the two spline functions. The smoothed composite histogram is obtained by applying a set of transformation parameters µ = {γ, θ, ϕ, t x , t y , t z ; δ j} are transformed to the base image 232, which are chosen to optimize the mutual information S(µ) according to a hierarchical multi-resolution optimization scheme. In the set of transformation parameters µ, {y, θ, ϕ}, the roll-pitch-yaw Euler angles of the whole, [t x , t y , t z ] is a translation vector concerning the whole and δ j is a set of deformation coefficients, each corresponding to a plurality of control points defined by a sparse grid superimposed on the base image 232. Typically, the sparse grid is a regular grid. However, in other embodiments, the sparse grid may be weighted by pixel intensity or other image metrics. For any given set of non-stiff transformation parameters, the mutual information S(µ) is calculated (as in Fig. 4) and mutual information gradients 7 S(µ) are calculated based on probability distributions p, pT and pR, which correspond respectively to: the smoothed composite histogram; a slightly smoothed histogram for the film image 252; and a slightly smoothed histogram for the base image 232. The probability distributions are assumed to be independent of the transformation parameters. The parameters τ and κ, shown in Fig. 4, are pixel intensity histogram location indices for the film image 252 and for the base image 232, respectively. Varying the histogram location sizes can adjust the quality of the image registration. The registration is described as a locally perturbed rigid body, since it has local deformation coefficients ∂j with rigid body transformations {γ, θ, ϕ, [t x , t y , t z]}. Evaluating the quality of the match, the registration process 260 either iterates on the modified µ or outputs µ (the image transformation 262 that registers the base image 232 with the particular film image 252) for a satisfactory value of S(µ).

[0024] Then, the method 200 includes applying each of the image transforms 262 to the static ACF matrix 242 to generate 270 a plurality of cine ACF matrices 272. Each of the image ACF matrices 272 is time-correlated 280 with one of the plurality of PET data sets 222 within the breath-domain filtered PET sequence 220 and is applied to the corresponding PET data set 222 to generate a reconstructed and attenuation-corrected PET image 212. In certain embodiments, each of the PET data sets 222 is time-correlated to more than one of the movie ACF matrices, and one of a simple average, a time-weighted average, a velocity-weighted average, or a time-velocity-weighted average is used in the application 210 of the multiple movie ACF matrices 272 to the PET data set 222.

[0025] Although static MR imaging acquisition 230 is shown preceding cine MR sequence acquisition 250, these steps can be performed in the opposite order as well. In certain embodiments, the cine MR sequence can be performed repeatedly (e.g., before and after the static breath-hold image), or the static breath-hold image can be repeated (e.g., before and after the cine sequence). Similarly, registration 260 can be performed concurrently with acquisition 250 or after all images have been acquired.

[0026] Although the plurality of registered ACF matrices 272 are described as being obtained by applying image transformations 262 to the base ACF matrix 242, registered ACF matrices 272 may also be calculated from the base MR image 232 as transformed according to the plurality of image transformations 262. Therefore, an alternative method 500, as described in accordance with other embodiments of the invention in Fig. 5, a first MR acquisition 230 is used to obtain a static (breath-held) MR image 232 suitable for deriving 240 baseline PET attenuation correction factors 242; and a second MR acquisition 250 is used to obtain a sequence of cine MR images 252 acquired during a respiratory cycle concurrent with a respiratory domain-filtered PET study 220. The static MR image 232 is non-rigidly registered to each of the cine MR images 252 to produce a set of transformed MR images 562 that correspond 560 to PET data sets 222 from the respiratory domain-filtered PET study and that are useful for deriving 570 attenuation correction factors 572. The registered MR images are then used to form 570 the attenuation correction factors 572, which are time-correlated 280 for generating 210 attenuation-corrected PET images 212 from their corresponding time-domain filtered PET data sets 222.

[0027] Fig.6 schematically shows an apparatus 600 according to an embodiment of the invention for carrying out the method 200 as described above. The apparatus 600 comprises a combining PET-MR scanner 610 communicatively connected to a controller 620, which in turn is communicatively connected to an image processor 630. The controller and the image processor can both be implemented as separate software in a single computer or separately in computers adjacent to the PET-MR scanner 610; or one or both of the controller 620 and the image processor 630 can be arranged outside (remotely) from the PET-MR scanner and / or from each other. The controller 620 is configured to carry out at least the acquisition 220 of the PET data set 222, the acquisition 230 of the base MR image 232, and the acquisition 250 of the sequence of cine MR images 252.The image processor 630 is configured to at least perform the following operations: generating 240 the base ACF matrix 242, registering 260 the base MR image 232 with each of the sequence of cine MR images 252, generating 270 the plurality of cine ACF matrices 272, and attenuation correcting 210 the time-domain filtered PET data sets 222 based on the plurality of cine ACF matrices 272. Either the controller 620 or the image processor 630 may be configured to perform the time-correlating 280 of the cine MR images 252 with the time-domain filtered PET data sets 222. Alternatively, the controllers 620 and the image processor 630 may be integrated into a single software / processor or may be implemented in a cloud (SaaS) paradigm.

[0028] Therefore, embodiments of the invention provide region-specific attenuation correction factors for time-domain filtered PET data based on attenuation correction data that is not acquired during the duration of a single region.

[0029] For example, some embodiments of the invention provide a method for attenuation correction of a plurality of time-domain filtered PET images of a moving structure. The method comprises creating an ACF matrix based on a base MR image of the structure; developing a plurality of image transforms, each image transform registering the base MR image to a corresponding cine MR image of the moving structure; applying each of the image transforms to the ACF matrix to generate a corresponding cine ACF matrix; time-correlating each of the cine MR images and its corresponding cine ACF matrix to one of the plurality of time-domain filtered PET data sets; and generating an attenuation-corrected time-domain filtered PET image from each of the time-domain filtered PET data sets. In certain embodiments, the base MR image is a composite image.In other embodiments, the baseline MR image is an averaged image. In some embodiments, the baseline MR image is obtained while the structure is substantially static. In some embodiments, the baseline MR image is obtained using a tissue-discriminating scanning protocol. In some embodiments, each of the cine MR images uses a same scanning protocol as each of the other cine MR images. In certain embodiments, developing a plurality of image transforms comprises scaling pixel intensity or shifting pixel intensity. In certain embodiments, developing the plurality of image transforms comprises iteratively calculating mutual information of a joint pixel intensity histogram constructed from the baseline MR image and from an intermediate transform of the image MR image.In certain embodiments, developing the plurality of image transformations comprises generating an image transformation µ for each cine MR image having rigid transformation parameters {γ, θ, φ, [t. x , t y , t z]} and deformation coefficients {δj} for registering the base MR image to the cine MR image. In certain embodiments, generating an attenuation-corrected, time-domain filtered PET image comprises applying at least one cine ACF matrix that is time-correlated to the PET data set to one of the plurality of time-domain filtered PET data sets. In certain embodiments, generating an attenuation-corrected, time-domain filtered PET image comprises applying a simple average, a time-weighted average, a velocity-weighted average, or a time-velocity-weighted average of many cine ACF matrices that are time-correlated to the PET data set to one of the plurality of time-domain filtered PET data sets.

[0030] Other aspects of the invention provide an apparatus for generating a plurality of attenuation-corrected time-domain filtered PET images. The apparatus comprises a PET-MR scanner, a controller communicatively connected to the PET-MR scanner, and an image processor communicatively connected to the controller. The controller is configured to acquire a base MR image of a structure, acquire a sequence of cine MR images of the structure, and acquire a plurality of time-domain filtered PET data sets of the structure. The image processor is configured to generate a base ACF matrix based on the base MR image, register the base MR image to at least some of the cine MR images, generate a cine ACF matrix based on the registration of the base MR image to the cine MR image, and attenuation-corrected time-domain filtered PET data sets based on the cine ACF matrices.In certain aspects, the controller is configured to time-correlate each of the cine ACF matrices with one of the time-domain filtered PET datasets. In certain aspects, the image processor is configured to time-correlate each of the cine ACF matrices with one of the time-domain filtered PET datasets. In certain aspects, the image processor is configured to register the baseline MR image with each of the cine MR images using locally perturbed rigid registration. In certain aspects, the image processor is configured to register the baseline MR image with each of the cine MR images using a mutual information algorithm. In certain aspects, the controller is configured to obtain the baseline MR image while the structure is substantially static. In certain aspects, the controller is configured to obtain the MR images and the PET datasets simultaneously while the structure is moving.In certain aspects, the image processor is configured to attenuation correct at least one of the plurality of time-domain filtered PET data sets using one of a simple average, a time-weighted average, a velocity-weighted average, or a time-velocity-weighted average of a plurality of cine ACF matrices that are time-correlated to the PET data set.

[0031] Other aspects of the invention provide a method for attenuation correction of PET imaging. The method comprises obtaining, in a first MR acquisition, a base MR image suitable for deriving PET attenuation correction factors; and obtaining, in a second MR acquisition, a cine sequence of MR images during a respiratory cycle concurrent with the respiratory-domain-filtered PET study. The method further comprises non-rigidly registering the base MR image to each of the cine MR images to generate a group of registered MR images corresponding to PET datasets from the respiratory-domain-filtered PET study and usable for deriving attenuation correction factors. The registered MR images are then used to form attenuation correction factors for their corresponding time-domain-filtered PET datasets.

[0032] It is to be understood that the above description is intended to be illustrative and not restrictive. For example, the above-described embodiments (and / or aspects thereof) may be used in combination with one another. In addition, many modifications may be made to adapt a particular situation or material to the teachings of the invention without departing from its scope. While the dimensions and types of materials described herein are intended to describe the parameters of the invention, they are by no means limiting and are exemplary embodiments. Many other embodiments will become apparent to those skilled in the art upon review of the above description. The scope of the invention should, therefore, be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled.In the appended claims, the terms "including" and "in which" are used as the plain English equivalents of the eloquent terms "comprising" and "wherein." Furthermore, in the following claims, terms such as "first," "second," "third," "upper," "lower," "bottom," "above," etc., are used merely as labels and are not intended to impose any numerical or locational requirements on their objects.

[0033] This written description uses examples to disclose various embodiments of the invention, including the best mode contemplated, and also to enable one skilled in the art to practice embodiments of the invention, including making and using any devices or systems and methods encompassed. The patentable scope of the invention is determined by the claims and may include other examples that would occur to one skilled in the art. Such other examples are intended to be within the scope of the claims if they have structural elements that do not depart from the literal language of the claims or if they include equivalent structural elements with insubstantial differences from the literal language of the claims.

[0034] As used herein, an element or step referred to in the singular and followed by the word "a," should not be construed as excluding the plural of elements or steps unless such exclusion is explicitly stated. Further, references to "one embodiment" of the present invention are not intended to exclude the existence of additional embodiments that also include the stated features. Moreover, embodiments that "comprise," "include," or "have" an element or a plurality of elements having a particular characteristic may include other such elements that do not have that characteristic, unless expressly stated otherwise.

[0035] Since certain changes may be made in the above-described embodiments without departing from the spirit and scope of the invention contained herein, it is intended that all such items of the above description or shown in the accompanying drawings be understood merely as examples illustrating the inventive concept herein and should not be construed as limiting the invention.

Claims

[1] A method (200) for attenuation correction (210) of a plurality of time-domain filtered PET images (222) of a moving structure, comprising: Creating (240) an ACF matrix (242) based on a base MR image (232) of the structure; Developing a plurality of image transformations (262), each of the image transformations (262) registering the base MR image (232) to a corresponding cine MR image (252) of the moving structure; applying each of the image transformations (262) to the ACF matrix (242) to generate a corresponding film ACF matrix (272); Time-correlating each of the cine MR images (252) and its associated cine ACF matrix (272) with one of a plurality of time-domain filtered PET data sets (222); and Generating an attenuation-corrected time-domain filtered PET image (212) from each of the time-domain filtered PET data sets (222). [2] The method (200) of claim 1, wherein the base MR image (232) is a composite image. [3] The method (200) of claim 1, wherein the baseline MR image (232) is an averaged image. [4] The method (200) of claim 1, further comprising a step of obtaining the baseline MR image (232) while the structure is substantially static. [5] The method (200) of claim 1, further comprising a step of obtaining the baseline MR image (232) using a tissue-discriminating scanning protocol. [6] The method (200) of claim 1, wherein each of the film MR images (252) uses the same scanning protocol as each other of the film MR images (252). [7] The method (200) of claim 1, wherein developing the plurality of image transformations (262) comprises pixel intensity scaling or pixel intensity shifting. [8] The method (200) of claim 1, wherein generating the plurality of image transforms (262) comprises iteratively calculating mutual information of a composite pixel intensity histogram for each of the cine MR images (252) created from the base MR image (232) and from an intermediate transform of the cine MR image (252). [9] The method (200) of claim 1, wherein generating the plurality of image transformations (262) comprises creating an image transformation µ for each cine MR image (252) having rigid transformation parameters {γ, θ, ϕ, [tx, ty, tz]} and deformation coefficients {δj} for registering the base MR image (232) to that cine MR image (252). [10] The method (200) of claim 1, wherein creating an attenuation-corrected time-domain filtered PET image (212) comprises applying at least one cine ACF matrix (272) time-correlated with the PET data set to one of the plurality of time-domain filtered PET data sets (222). [11] The method (200) of claim 10, wherein generating an attenuation-corrected, time-domain filtered PET image (212) comprises applying a simple average, a time-weighted average, a velocity-weighted average, or a time-velocity-weighted average of a plurality of cine ACF matrices (272) time-correlated with that PET data set to one of a plurality of time-domain filtered PET data sets (222). [12] Apparatus (600) for generating a plurality of attenuation-corrected, time-domain filtered PET images (212), the apparatus comprising: a PET-MR scanner (610); a controller (620) communicatively connected to the PET-MR scanner (610), the controller (620) being configured to acquire a base MR image (232) of a structure, acquire a sequence of cine MR images (252) of the structure, and acquire a plurality of time-domain filtered PET data sets (222) of the structure; and an image processor (630) communicatively connected to the controller (620), wherein the image processor (630) is configured to generate a base ACF matrix (242) based on the base MR image (232), register the base MR image (232) with at least some of the cine MR images (252), generate cine ACF matrices (272) based on the registration of the base MR image (232) with the cine MR image (252), and attenuation correct the time-domain filtered PET data sets (222) based on the cine ACF matrices (272). [13] The apparatus (600) of claim 12, wherein the controller (620) is configured to correlate each of the film ACF matrices (272) with one of the time-domain filtered PET data sets (222). [14] The apparatus (600) of claim 12, wherein the image processor (630) is configured to time correlate each of the film ACF matrices (272) with one of the time-domain filtered PET data sets (222). [15] The apparatus (600) of claim 12, wherein the image processor (630) is configured to register the base MR image (232) with each of the cine MR images (252) using locally perturbed rigid registration. [16] The apparatus (600) of claim 12, wherein the image processor (630) is configured to register the base MR image (232) with each of the cine MR images (252) using a mutual information algorithm. [17] The apparatus (600) of claim 12, wherein the controller (620) is configured to obtain the baseline MR image (232) while the structure is substantially static. [18] The apparatus (600) of claim 12, wherein the controller (620) is configured to obtain the film MR image (252) and the PET data sets (222) simultaneously while the structure is moving. [19] The apparatus (600) of claim 12, wherein the image processor (630) is configured to attenuation correct at least one of the plurality of time-domain filtered PET data sets (222) using a simple average, a time-weighted average, a velocity-weighted average, or a time-velocity-weighted average of a plurality of film ACF matrices (272) time-correlated with that PET data set. [20] A method for attenuation correction of PET imaging in a breath-filtered PET study, comprising: Obtaining a baseline MR image (232) in a first MR acquisition suitable for deriving PET attenuation correction factors; Obtaining a cine sequence of MR images in a second MR acquisition during a respiratory cycle simultaneously with the respiratory zone-filtered PET study; and non-rigidly registering the base MR image (232) with each of the cine MR images (252) to generate a group of registered MR images corresponding to the PET data set from the respiratory domain-filtered PET study and usable for deriving attenuation correction factors, wherein the registered MR images are then used to form attenuation correction factors for their corresponding time-domain-filtered PET data sets (222).

Images