WO2018083464A1

WO2018083464A1 - Pet imaging system

Info

Publication number: WO2018083464A1
Application number: PCT/GB2017/053286
Authority: WO
Inventors: Ottavia BERTOLLI; Kris Thielemans
Original assignee: UCL Business Ltd
Current assignee: UCL Business Ltd
Priority date: 2016-11-02
Filing date: 2017-11-01
Publication date: 2018-05-11
Anticipated expiration: 2019-05-02
Also published as: GB201618487D0

Abstract

A method for determining a motion signal from positron emission tomography (PET) image data, the method comprising: providing PET image data obtained from a PET scan of a patient, wherein the PET image data comprises spatial coordinates, time-of-flight (TOF) information and acquisition timing information for each line of response (LOR) of the PET scan; allocating the PET image data to different TOF bins based on the TOF information; assigning weighting factors to the PET image data based on at least the TOF bin to which the PET image data is allocated so as to give greater weighting to PET image data corresponding to a region of interest in the field of view of the PET scan; and applying a motion detection method to the PET image data so as to generate a motion signal indicative of physiological movement of the patient, wherein the PET image data are used as an input for the motion detection method in accordance with the weighting factors assigned to the PET image data such that the motion signal is indicative of physiological movement in the region of interest.

Description

PET IMAGING SYSTEM

The present invention relates to the field of positron emission tomography (PET). More particularly, the present invention relates to a PET imaging system, imaging software for a PET imaging system and a method for determining a motion signal from PET image data. The motion signal can be used to improve the quality of images reconstructed from the PET image data.

Movement of a patient during a PET scan can negatively affect image quality. Typical physiological movements of the patient during a PET scan include head motion, respiratory motion, arm movement, cardiac contraction and bulk motion, for example.

Methods have been proposed to detect motion from the PET image data, with the greatest emphasis on extracting a signal related to respiratory motion. However, with the known techniques it can be difficult to distinguish between different types of physiological motion, and different types of motion can give rise to mixed signals. For example, a calculated motion signal that is intended to represent respiratory motion of the patient can be confounded by bulk movement of the patient, or arm movement of the patient. This is because the PET image data are projections of an electron-positron annihilation event along a line of response (LOR). As a result, the total count of annihilation events on an LOR can be affected by motion of any part of the patient along that LOR. For example, if the LOR passes through the heart of the patient and also through an arm of the patient, then arm movement would affect the determination of a motion signal that is intended to correspond to cardiac contraction.

Accordingly, it is desirable to improve the quality of a motion signal determined from PET image data. The motion signal can be used to improve the quality of images reconstructed from the PET image data.

According to an aspect of the present invention, there is provided a method for determining a motion signal from PET image data, the method comprising:

providing PET image data obtained from a PET scan of a patient, wherein the PET image data comprises spatial coordinates, time-of-flight (TOF) information and acquisition timing information for each LOR of the PET scan;

allocating the PET image data to different TOF bins based on the TOF

information;

assigning weighting factors to the PET image data based on at least the TOF bin to which the PET image data is allocated so as to give greater weighting to PET image data corresponding to a region of interest in the field of view of the PET scan; and

applying a motion detection method to the PET image data so as to generate a motion signal indicative of physiological movement of the patient, wherein the PET image data are used as an input for the motion detection method in accordance with the weighting factors assigned to the PET image data such that the motion signal is indicative of physiological movement in the region of interest.

According to a further aspect of the present invention, there is provided a PET imaging system comprising:

a PET scanner configured to obtain PET image data by performing a PET scan of a patient, wherein the PET image data comprises spatial coordinates, TOF information and acquisition timing information for each LOR of the PET scan; and

a processor configured to:

allocate the PET image data to different TOF bins based on the TOF information;

assign weighting factors to the PET image data based on at least the TOF bin to which the PET image data is allocated so as to give greater weighting to PET image data corresponding to a region of interest in the field of view of the PET scan; and

apply a motion detection method to the PET image data so as to generate a motion signal indicative of physiological movement of the patient, wherein the PET image data are used as an input for the motion detection method in accordance with the weighting factors assigned to the PET image data such that the motion signal is indicative of physiological movement in the region of interest.

According to a further aspect of the present invention, there is provided imaging software for a PET imaging system, the imaging software configured to process PET image data obtained from a PET scan of a patient, wherein the PET image data comprises spatial coordinates, TOF information and acquisition timing information for each LOR of the PET scan, the imaging software configured to: allocate the PET image data to different TOF bins based on the TOF information; assign weighting factors to the PET image data based on at least the TOF bin to which the PET image data is allocated so as to give greater weighting to PET image data corresponding to a region of interest in the field of view of the PET scan; and

Embodiments of the present invention will now be described, by way of example only, with reference to the accompanying drawings in which:

Figure 1 schematically depicts a PET imaging system according to an embodiment of the present invention;

Figure 2 shows an example of PET image data used in an embodiment of the present invention; Figure 3 schematically depicts PET image data aggregated into predefined time frames and allocated to different TOF bins according to an embodiment of the present invention;

Figure 4 shows an example of a table of weighting factors assigned to the PET image data allocated to each of the TOF bins according to an embodiment of the present invention;

Figure 5 schematically depicts different regions of a field of view (FOV) of a PET scan;

Figure 6 schematically depicts part of a motion detection method applied to PET image data according to an embodiment of the present invention;

Figure 7 schematically depicts a motion signal determined according to an embodiment of the present invention; and Figure 8 is a flow chart of a method for determining a motion signal from PET image data according to an embodiment of the present invention.

Figure 1 schematically depicts a PET imaging system 10 according to an embodiment of the present invention. The PET imaging system 10 comprises a PET scanner 11. The PET scanner 11 is configured to perform a PET scan on a patient.

During the PET scan, radioisotopes in the patient undergo positron emission decay. The emitted positron and an electron are annihilated in annihilation event, thereby producing a pair of annihilation photons moving in opposite directions along an LOR. These photons are detected by detectors of the PET scanner. Pairs of photons that are not detected within a set timing-window of each other are ignored. The timing-window may be a few nanoseconds, for example. The PET scanner 11 is configured to obtain PET image data by performing the PET scan of the patient. The PET image data comprises spatial coordinates, TOF information and acquisition timing information for each LOR of the PET scan. Figure 2 shows an example of PET image data obtained by the PET scanner 11. The spatial coordinates of the PET image data are spatial coordinates of the LOR. For example, the spatial coordinates may be spatial coordinates in polar space. For example, one of the spatial coordinates may represent the shortest distance between the LOR and the origin. The origin is the central point of the FOV of the PET scanner 11. The other two spatial coordinates may be angles describing the orientation of the LOR in polar space. Alternatively the spatial coordinates might correspond to the locations of the pair of detectors that detected the two annihilation photons. The pair of detectors are at either end of the LOR. Accordingly, the location of the LOR can be calculated from the location of the pair of detectors. The TOF information is information about the time difference between the arrival times of one of the photons being detected and the other photon (travelling in the opposite direction along the LOR) being detected. For example, the TOF information may be a TOF time difference value representing the time difference between when the two photons were detected. A TOF time difference value of zero would correspond to a position on the LOR that is halfway between the two detectors of the PET scanner 11 that detected the two photons. A positive time difference would correspond to a position along the LOR closer to one of the detectors than the other. A negative time difference would indicate a position in the opposition direction along the LOR closer to the other detector.

Hence, by providing the TOF information, there is additional information available for each annihilation event. The arrival time difference (i.e. the TOF time difference value) of the two photons gives approximate location information along the LOR. The accuracy of this location is determined by the TOF timing resolution of the PET scanner 11. For example, the PET scanner 11 may have a TOF timing resolution in the region of from 400ps to about 600ps. A TOF timing resolution of 400ps corresponds to a location accuracy in the LOR of about 60mm. A TOF timing resolution of 600ps corresponds to a location accuracy along the LOR of about 90mm.

The acquisition timing information indicates the timing at which the photons were detected. This timing information can be used to later divide the PET image data into predefined time frames. Each time frame corresponds to a different period of time of the PET scan.

As depicted in Figure 1, the PET imaging system 10 may comprise a processor 13. The processor 13 is configured to process the PET image data. For example, the processor 13 may process the PET image data so as to reconstruct an image of the FOV in image space. The processor 13 is configured to run imaging software. The imaging software instructs the processor 13 to determine a motion signal from the PET image data.

Optionally, the processor 13 is configured to use the motion signal to provide an improved reconstructed image of the FOV, or a portion of the FOV, based on the PET image data. For example, the determined motion signal can be used to reduce undesirable artefacts in the reconstructed image that are due to physiological motion of the patient during the PET scan.

As depicted in Figure 1, optionally the PET imaging system 10 comprises a memory 14. The memory 14 is configured to store the PET image data at various stages of the processing. The memory 14 can be used to store other information such as the weighting factors assigned to each of the TOF ends, e.g. as depicted in Figure 4. The memory 14 can be used to store imaging software that comprises instructions for the processor 13.

As depicted in Figure 1, optionally the PET imaging system 10 comprises an input/output device 12. The input/output device 12 is configured to allow a user to input information to the PET imaging system 10. For example, the input/output device 12 may comprise a keyboard, a mouse or a touchscreen, for example, to allow the user to input instructions and selections to the PET imaging system 10. The input/output device 12 may optionally comprise a display to output information to the user. For example, the input/output device 12 may display reconstructed images formed from the PET image data, or sinograms (which are data constructed by histogramming the PET image data according to the above mentioned LOR coordinates) of the PET image data.

The present invention may be embodied as a method for determining a motion signal from PET image data. Figure 8 is a flow chart of an exemplary method for determining a motion signal from PET image data according to an embodiment of the present invention.

The method comprises a step of providing PET image data. Optionally, the PET image data is PET image data that has previously been obtained from a PET scan of the patient. For example, a PET scan of the patient may have been performed and the PET image data then saved in a memory. The memory is then read so as to provide the PET image data. Alternatively, the PET image data may be obtained by performing a PET scan of the patient. Hence, the PET image data may be obtained and provided in real time. As shown in step 81 of Figure 8, the PET image data that is provided may be list mode PET image data. List mode PET image data is raw PET image data that has not undergone processing. For example, the list mode PET image data may comprise spatial coordinates, TOF information and acquisition timing information for each detected annihilation event.

As shown in Figure 8, optionally the method comprises a step 82 of aggregating the list mode PET image data into predefined time frames. The aggregation of PET image data into predefined time frames is sometimes called unlisting. For example, the list mode PET image data can be unlisted by using histogramming into predefined time frames. The aggregation of the PET image data into the predefined time frames may be performed based on the acquisition timing information. For example, each time frame may have a set length, such as 500ms. 500ms is merely an example and other lengths of time frame may be used, such as 300ms. The length of the time frame is not particularly limited. However, it is desirable for all of the time frames to have the same length.

Based on the acquisition timing information included in the list mode PET image data, the data for each annihilation event is aggregated into the appropriate time frame. Optionally the unlisted PET image data can be spatially down-sampled so as to reduce the overall amount of data that is being handled. This saves on memory and speeds up the image processing. The different time frames can be compared so as to find information about physiological movement of the patient during the PET scan. Hence, the length of the predefined time frames may be selected with a view to the expected speed of

physiological movement. For rapid physiological movements, it may be desirable for the lengths of the predefined time frames to be shorter.

As shown in Figure 8, the method comprises the step 83 of allocating the PET image data to different TOF bins. The different TOF bins correspond to different timing differences between the two photons being detected. Hence, the magnitude of the TOF information represents an approximation of how far away from the centre of the LOR the annihilation event occurred.

For example, one of the TOF bins, which may be called the central TOF bin, may correspond to the smallest TOF time difference values. For example, the central TOF bin may comprise PET image data for which the TOF time difference value is between about -45ps and about +45ps.

Another TOF bin may correspond to TOF time difference values from about 45ps to about 135ps. A further TOF bin may correspond to PET image data which has a TOF time difference value of from -135ps to about -45ps. Optionally, these two TOF bins may be combined to form a single TOF bin corresponding to PET image data for which the TOF time difference value has a magnitude of between 45ps and 135ps (i.e. without regard to whether the TOF time difference value are positive or negative). However, it may be desirable to have separate TOF bins for positive and negative TOF time difference values. This is because by having separate TOF bins it may be possible to focus on physiological motion in a more specific region of the field of view. For example, the negative TOF time difference values may generally include PET image data of annihilation events that occurred generally more on the left side of the patient than on the right side of the patient. This may make it possible to determine a motion signal corresponding to movement of the left arm of the patient, for example. Alternatively, it may be that negative TOF time difference values correspond more generally to annihilation events at positions in the vertically lower half of the PET scanner 11.

The spatial position associated with the sign of the TOF information (i.e. the TOF time difference values) depends on the convention used by the PET scanner 11 to calculate the TOF information. For example, one convention might be that the TOF time difference value is calculated by subtracting the detection time for the left-most detector from the acquisition time for the right-most detector out of the two detectors at either end of the LOR. With such a convention, annihilation events that occur in the left side of the PET scanner 11 are more likely to be allocated to the TOF bin that corresponds to positive TOF time difference values. The convention used by the PET scanner 11 may be selected so as to suit how the user wants to divide up the FOV so as to focus on physiological movement in a particular region of the FOV.

The number of TOF bins provided by the scanner is usually determined by the hardware. However the number of TOF bins considered for the extraction of the motion signal can be smaller. The number of TOF bins may be selected by the user depending on which regions in the FOV they want to focus on.

Figure 3 schematically depicts the PET image data after it has been aggregated into predefined time frames and allocated to difference TOF bins. In Figure 3, three TOF bins 31, 32, 33 are shown. Within each TOF bin, there is a plurality of PET image data sets, each set corresponding to a different time frame. For example, PET image data set 31-1 may correspond to the first time frame within TOF bin 31 while PET image data set 31-n corresponds to the PET image data in the nth time frame of the TOF bin 31. Similarly, the PET image data set 32-1 corresponds to the PET image data aggregated into the first time frame and allocated to the TOF bin 32, while the PET image data set 32-n corresponds to the PET image data aggregated into the nth time frame and allocated to the TOF bin 32. Each PET image data set shown in Figure 3 corresponds to sinogram data. This means that each PET image data set could be used to produce a sinogram. However, it may not be necessary to actually produce these sinograms in order to carry out the method of the present invention. Instead, the division of the PET image data as shown in Figure 3 may be an intermediate step in the method of an embodiment of the invention.

As shown in Figure 8, the method comprises the step 84 of assigning a weighting factor to the PET image data in each of the TOF bins. The step 84 of assigning a weighting factor to the PET image data is performed so as to give weighting to PET image data corresponding to a region of interest in the FOV of the PET scan. The weighting factors are assigned based on at least the TOF bin to which the PET image data is allocated. For example, if the region of interest is in the centre of the FOV (for example so as to focus on respiratory motion), then greater weighting may be given to the PET image data allocated to TOF bins that include generally more PET image data of annihilation events that occurred in the centre portion of the FOV of the PET scan. For example, the TOF bin that corresponds to the smallest TOF time difference values comprises more PET image data of annihilation events in the centre of the FOV compared to the other TOF bins. For this reason, this TOF bin maybe called the central TOF bin. In Figure 3, TOF bin 31 represents the central TOF bin. Hence, depending on the region of interest in the FOV of the PET scan, different weighting factors may be assigned to the PET image data allocated to the TOF bins. If the user is interested in determining a motion signal for cardiac contraction, then they may give greater weighting (i.e. larger weighting factors) to the PET image data allocated to TOF bins that include generally more PET image data from annihilation events in the region of the heart of the patient.

Optionally, the weighting factors are assigned based on only the TOF bin to which the PET image data is allocated. This means that all of the PET image data allocated to a particular TOF bin are assigned the same weighting factor. This is therefore equivalent to assigning a weighting factor to each TOF bin.

Figure 4 shows such an example of a weighting factor being assigned to each TOF bin.

However, different PET image data within one TOF bin may be assigned different weighting factors. For example, the PET image data may be assigned different weighting factors at the level of their LOR. This means that all of the PET image data of a particular LOR are assigned the same weighting factor. As a further alternative, in an embodiment the PET image data are assigned different weighting factors at the level of annihilation events. This means that the PET image data corresponding to different annihilation events can be assigned different weighting factors (and hence PET image data within the same LOR can be assigned different weighting factors). Optionally, the assignment of the weighting factors is based not only on the TOF bin to which the PET image data is allocated, but also on additional information. For example, in an embodiment, the weighting factors are assigned based on the TOF bin and the spatial coordinates of the PET image data. For example, the weighting factors may be assigned based partly on the distance of the LOR from the origin. Merely as an example, each weighting factor may have a value of from 0 to 1. A weighting factor of zero means that the PET image data is not taken into account at all when determining the motion signal. This can help to exclude certain types of physiological movement so that those physiological movements do not undesirably affect the determination of the motion signal of the target physiological movement. For example, by selecting appropriate TOF bins (and excluding other TOF bins) it is possible to exclude arm movements from otherwise distorting the determination of the motion signal corresponding to respiratory motion.

Figure 5 schematically shows the FOV 50 of a PET scan divided into different regions 51 to 55. In particular, Figure 5 shows the FOV 50 divided into a central region 51, two intermediate regions 52 and 53 and two outer regions 54 and 55. PET image data for annihilation events occurring in the central region 51 are more likely to be allocated to the central TOF bin 31 (which is for PET image data with small TOF time difference values). This is because the annihilation event in the central region 51 must be reasonably close to the central of the LOR, thereby providing a small TOF time difference value. Similarly, annihilation events that occur in the outer regions 54 and 55 are likely to have large TOF time difference values and therefore be allocated to TOF bins for large TOF time difference values.

Of course, there may not be an exact one-to-one correspondence between the region in which the annihilation event occurred and the TOF bin to which the image data is allocated. For example, one can imagine an LOR that extends only within the outer region 54 because the detectors at either end of the LOR are relatively close to each other in the PET scanner. Such an LOR does not pass through the central region 51. If an annihilation event occurred in the centre of that LOR, then it would have a zero (or near- zero) TOF time difference value and hence be allocated to the central TOF bin 31 even though the annihilation event occurred in the outer region 54. Furthermore, there is some uncertainty about the location of each annihilation event due to the TOF timing resolution issue mentioned above. This leads to a slight blurring between the different TOF bins.

Nevertheless, the division of the PET image data into different TOF bins has been found through experiment to be particularly effective at separating the PET image data into sets that can be used to focus on physiological movement in a particular region of interest of the FOV 50. This is because although the correspondence is not exact, there is a high level of correspondence between the spatial region in which an annihilation event occurred and the TOF bin to which the image data is allocated.

In step 84, each weighting factor may be assigned depending on the extent to which data from annihilation events in the region of interest are expected to be allocated to the different TOF bins.

As shown in Figure 8, the method comprises the step 85 of applying a motion detection method to the PET image data in accordance with the weighting factors. In particular, the motion detection method is applied so as to generate a motion signal indicative of physiological movement of the patient. The PET image data are used as an input for the motion detection method in accordance with the weighting factor of the PET image data. For example, if the weighting factor of the PET image data allocated to a particular TOF bin is zero, then the PET image data allocated to that TOF bin are not used at all as an input for the motion detection method. Alternatively, if the weighting factor is one, then the PET image data allocated to that TOF bin are used as an input for the motion detection method. Weighting factors can take any value between 0 and 1 depending on the extent to which the PET image are wanted to be used as an input for the motion detection method. This allows the motion signal that is determined to be indicative of physiological movement in the region of interest. For example, it is possible to determine a motion signal that is indicative of respiratory motion, while filtering out arm movement, for example.

Various motion detection methods can be used in accordance with the present invention so as to generate the motion signal from the PET image data of the selected TOF bins. Merely as an example, in the case where the assigned weighting factors of the PET image data in the TOF bins are 1 for one or more central TOF bins and 0 for the other TOF bins, the motion detection method comprises a principal component analysis (PCA) to generate the motion signal. A further embodiment with a different choice of weighting factors will be described below. In addition other methods such as other techniques for

dimensionality reduction or a spatial displacement method could be used in conjunction with the present invention. It should be noted that many motion detection techniques obtain a signal with undefined scale and sign and therefore need to be complemented by a sign-determination method. Those skilled in the art will realise that such a sign- determination method will benefit from the weighting factors determined in the present invention. For completeness, a broad outline of a PCA method that can be used in the present invention is provided below.

Figure 6 schematically depicts part of the PCA method. The left hand side of Figure 6 shows the central TOF bin 31, with separate PET image data sets 31-1 to 31-n for the first to nth time frames. Optionally, the PET image data sets 31-1 to 31-n may be processed to reduce noise, for example. However, this is not necessary in all embodiments of the invention.

Optionally, the PCA process comprises generating a time-averaged PET image data set 61. The time-averaged PET image data set 61 is produced from the PET image data sets 31-1 to 31-n. In particular, if the PET image data sets 31-1 to 31-n are considered as sinogram data sets then average information is generated for every element in the sinogram data sets so as to compute the average of the corresponding elements. The data of the time-averaged PET image data set 62 is then subtracted from each of the PET image data sets 31-1 to 31-n in turn so as to generate a corresponding zero mean data set 62-1 to 62-n.

The zero mean data sets 62-1 to 62-n of the zero mean information 62 are then used in the PCA method. In general, a similar process would be used for the other PET image data which have been assigned a non-zero weight. The PCA method is used to find the dominant principal components (sometimes called eigenvectors) of the estimated covariance matrix of the data. The PCA method outputs a number of principal components and their corresponding coefficients (sometimes called weights). Each principal component is a data set that has the same size as each of the zero mean data sets 62-1 to 62-n.

It is usually only necessary to compute the first few principal components in order to describe well the physiological motion represented by the variation over time of the zero mean data sets 62-1 to 62-n.

Figure 7 schematically depicts the relationship between time and the coefficient for the first principal component. The principal components are labelled in order of decreasing eigenvalue, that reflects the amount of variation in the data that is described by the principal component. Hence, the first principal component is the principal component that has generally the largest eigenvalue, i.e. it is the principle component that is most important for describing the physiological movement represented by the zero mean data sets 62-1 to 62-n. In Figure 7, each motion signal data point 71-1 to 71-n is the coefficient for the first principal component for the corresponding zero mean data set 62-1 to 62-n. The coefficient can be generated by doing an element-wise multiplication of the values of the first principal component with each of the zero mean data sets 62-1 to 62-n, and summing over the zero mean data set, thus obtaining a single number for each zero mean data set. As shown in Figure 7, the coefficients 71-1 to 71-n form the data points for a motion signal. The motion signal corresponds to physiological motion in the central region 51 of the FOV 50 of the PET scan. This is because the data used to form the motion signal data points shown in Figure 7 are the data of the central TOF bin 31. The motion signal represented in Figure 7 therefore corresponds to respiratory motion (because the patient is centered in the FOV 50). Accordingly, the movement of the arms of the patient has been excluded. This improves the quality of the motion signal relating to respiration.

In some instances it might be advantageous to select a group of TOF bins and not only the central one. In this context, selection of a TOF bin means assigning a non-zero weighting factor to the PET image data allocated to the TOF bin. In this case PCA would be applied to all of the selected TOF bins and not only to the central one.

In a further embodiment, the weighting factors assigned to the PET image data could also have values included in the range between 0 and 1, so as to more precisely distinguish between different areas of the FOV. In this case, the above mentioned method for motion extraction (i.e. PCA) can be modified as follows: the values of the elements of the PET image data sets are multiplied by their corresponding weighting factor before the application of PCA. The resulting effect of this process is that the PET image data with low weighting factors will contribute less to the motion signal.

Optionally, the method comprises generating an image of the FOV 50, or a portion therefore, based on the PET image data and the motion signal. For example, the motion signal shown in Figure 7 can be used to improve the quality of the generated image. The motion signal is used so as to compensate for motion of the patient in the image generated.

As shown in Figure 7, the motion signal represents a roughly sinusoidal motion. It is possible to group together time frames that correspond to similar phases of the physiological movement. For example, in the motion signal shown in Figure 7, the third and ninth time frames corresponding to the motion signal data points 71-3 and 71-9 could be grouped together because they appear to correspond to roughly the same phase of the sinusoidal respiratory motion. Similarly, the first and fourth time frames could be grouped together because they represent a similar phase in the respiratory motion. Optionally, the method comprises reconstructing the PET image data from the grouped time frames into images representing emission activity distribution. Optionally, the different images corresponding to different groups of times frames may be registered to each other. Hence, the motion signal can be used to determine different physiological positions of the patient. An image can then be generated for each physiological position of the patient. These images can then be registered to each other, if desired.

The quality of the motion signal determined using the present invention is better than motion signals developed using known techniques. Furthermore, dividing the PET image data into different TOF bins can be performed particularly quickly with very little computation from the raw PET image data. In particular, it is not necessary to perform any back projection (i.e. attributing the annihilation events to the voxels in image space) in order to filter out PET image data that originates from certain regions of the FOV 50. Hence, the present invention provides a very quick and very computationally efficient way of focusing on a particular region of interest so as to provide a good quality motion signal of physiological motion in that region of interest. The method can be performed on the raw PET image data without complicated back projection processing. The better quality motion signal then allows a better quality image to be reconstructed from the PET image data.

Optionally, the PET image data allocated to a subset of the TOF bins are assigned a weighting factor of zero such that the PET image data assigned to the subset of TOF bins are excluded from the motion detection method. For example, if it is desired to exclude respiratory motion from the motion signal detection, the PET image data allocated to the central TOF bin 31 could be assigned a weighting factor of zero. This would mean that the PET image data corresponding to annihilation events in the central region 51 of the FOV 50 would largely be excluded from the motion detection method. This would make it possible to determine a better quality motion signal for physiological movements apart from the respiratory motion.

Optionally, one or more TOF bins correspond to TOF information that indicates that the time period between detection of the two photons is greater than a threshold time period. The data in these TOF bins may generally include more PET image data from annihilation events at the periphery of the FOV 50. Optionally, the PET image data allocated to these TOF bins are assigned a zero weighting factor so as to exclude physiological movement outside the central portion of the FOV 50 that comprises the region of interest. Hence, this makes it possible to exclude the effect of arm movement, for example, on the detection of the motion signal for respiratory motion or cardiac contraction, for example. For example, it may be that the PET image data allocated to all of the TOF bins except for the central TOF bin 31 (or possibly a certain number of the most central TOF bins) are assigned zero weighting factors (or near-zero weighting factors) so as to be excluded from the motion signal detection.

Optionally, one or more TOF bins correspond to TOF information that indicates that the time period between detection of the two photons is less than a threshold time period.

Optionally, the method comprises truncating the PET image data by excluding any PET image data for which the spatial coordinates indicate that the LOR is greater than a threshold distance from the origin. Merely as an example, according to an exemplary embodiment the PET image data allocated to all TOF bins except for the central TOF bin 31 are assigned a zero weighting factor and are therefore excluded from the motion signal determination. The truncation is then a further processing step. The spatial coordinates of the PET image data allocated to the central TOF bin 31 can be used to compute the distance of the LOR from the origin. It is possible therefore to use this information from the spatial coordinates to exclude PET image data from annihilation events that occurred more than a threshold distance from the origin. This can be used to exclude annihilation events that occurred outside the central region 51 of the FOV.

The motion detection method can then be applied to the truncated PET image data, so as to exclude physiological movement beyond the threshold distance from the origin. For example, the threshold distance could be set as the radius of the central region 51 of the FOV. However, the threshold distance is not particularly limited. For example, the region of interest is in one of the intermediate regions 52, 53 of the FOV 50, then the threshold distance may be set to the distance from the origin of the FOV 50 to the outer perimeter of the intermediate regions 52, 53. In a further embodiment, different weighting factors are assigned for different LORs or different annihilation events within each TOF bin. As an example, as opposed to truncating the PET image data for which the spatial coordinates indicate that the LOR is greater than a threshold distance from the origin, a weighting factor could be assigned that depends on this distance. In this context, if a piece of PET image data is assigned a weighting factor of zero, then this means that PET image data is excluded (i.e. the total PET image data is truncated by excluding any PET image data assigned a weighting factor of zero). For example, the PET image data may be assigned the weighting factor based partly on the distance of the LOR from the origin. Alternatively, the PET image data may be assigned the weighting factor based partly on the estimated location of the annihilation event, where the TOF information is used in order to make this estimate. Hence, the weighting factor may be assigned to the PET image data partly on the basis of the LOR location, or on the basis of the estimated annihilation event location, in which case both the TOF bin and the spatial coordinates are taken into account when assigning the weighting factors.

As exemplified above, the PCA method may be performed on sinogram data, namely data from which the sinogram can be produced directly. The sinogram data are projections of the annihilation events along the LORs. The sinogram data are not in image space. By performing that PCA method on the sinogram data, it is not necessary to process the sinogram data to provide image space data in order to determine the motion signal.

Instead, the motion signal can be determined much more directly from the raw PET image data. As mentioned above, the weighting factors may be determined based on the extent to which each TOF bin and/or LOR is expected to be allocated PET image data

corresponding to annihilation events in the region of interest. However, the

correspondence between the regions of the FOV 50 and the TOF bins or LORs is not exactly one-to-one. Optionally, the weighting factors are determined based on a forward projection of the region of interest. The forward projection of the region of interest indicates to what extent each TOF bin and/or LOR is expected to be allocated PET image corresponding to the region of interest. In particular, by performing the forward projection, it can be seen which TOF bins the PET image data of annihilation events within the region of interest are allocated to. This information can be used to determine the weighting factors. Furthermore, also the weighting factors for the different LORs of the sinogram data of each TOF bin can be determined from the projection of the region of interest. As an example, the weighting factors could be proportional to the value of the forward projection.

It may be possible to perform such a forward projection once, and then store the appropriate weighting factor information that can be used on data obtained from many different PET scans. For example, the forward projection may be performed to find out which TOF bins and/or LORs and/or sinogram data points (i.e. an individual element of sinogram data) it would be appropriate to choose if it was desired to focus on movement of the right arm of the patient. Similarly, forward projection could be used to determine which TOF bins are appropriate to focus on the myocardium of the patient. These weighting factors could then be stored and used by the user depending on which part of the patient the user wants to focus on.

The appropriate weighting factors for the different parts of the patient may depend on factors such as the size of the patient. Hence, a forward projection may be performed for different sizes of patient so as to build up a database of the appropriate weighting factors to focus on particular parts of patients of different sizes.

Optionally, the region on interest is determined based on an image of the FOV generated from the PET image data. For example, the location of the arms can be determined based on an image (e.g. MR or CT, or a non-attenuated PET reconstruction, or a simple TOF- back projection of the PET image data). This would allow determination of a region of interest (which would be somewhat larger than the segmented arms to accommodate full motion).

Alternatively, detecting arm movement may be based on knowing the position of the arms (normally sideways of the patient). This means that data originating from above or below the patient can be ignored.

Once the region of interest is determined, which TOF bins and/or LORs actually have PET image data from that region of interest can be determined based on purely geometrical grounds. Alternatively, as explained above, the region of interest can be forward projected to see which TOF bins contain information from the arms. Other TOF bins can then be masked out (by assigning the PET image data allocated to them a zero weighting factor). It is also possible to use a weighted mask (using weighting values between zero and one), allowing the motion detection method to assign more weight to the PET image data allocated to the most relevant TOF bins.

The methods of the present invention can be generalised to any region of interest. For example, for cardiac applications, it would advantageous to keep only PET image data from a region of interest from around the myocardium. Other types of physiological motion could then be ignored by the motion detection method.

Similarly, for respiratory signal extraction, it might be advantageous to eliminate bins with PET image data from the myocardium so as to reduce the effects of cardiac contraction on the determined motion signal.

Detecting a motion signal related to arm movement may be advantageous. For instance, in PET imaging, attenuation correction and scatter correction can be performed provided that a density map is known. The density map is usually derived from CT or MR. This density may need to be aligned with the PET image itself to avoid artefacts and quantitative errors. Movement of the patient during the PET scan can cause misalignment between the density map and the PET image. The present invention makes it possible to detect when the arm movement happens. Once this is known, the data can be split into different time periods according to different arm positions and the position of the arms can be estimated using techniques that estimate attenuation from the PET image data itself or by back projecting the PET image data related to the different time periods.

Optionally, the method comprises using two different regions of interest corresponding to different types of physiological movement. For example, in the cardiac application, a first region of interest excluding the heart might be used to extract a respiratory signal. A second region of interest on the myocardium could be used to extract a signal related to cardiac contraction, potentially after correcting that data for respiratory movement.

For the purpose of detecting bulk motion, the central TOF bins can be neglected (i.e. their allocated PET image data can be assigned a zero weighting factor or a near-zero weighting factor). This is because typically the major causes of motion in the centre are respiration and heart beating. Hence, only the outer TOF bins would be used for the purpose of detecting bulk motion. The length of the predefined time frames could be longer so as to decrease the noise. Furthermore, although spatial down-sampling can be performed so as to reduce the amount of data for detecting respiratory motion, for example, spatial down-sampling is less desirable for the purpose of detecting bulk motion. This is because more spatial information is needed in order to detect bulk motion. If bulk motion were detected, the PET image data could then be split into time periods to not include the duration when the bulk motion occurred. The frames could then be further analysed for detecting respiratory motion from the central TOF bins, where the effect of bulk motion would have been removed. The number of TOF bins is not particularly limited. For example, the number of TOF bins may be five, 11, 55, for example. It may not be necessary to have such a large number of TOF bins as 55 in order to focus on the desired regions of interest. A smaller number of TOF bins may speed up processing. The present invention may be embodied as a method for determining a motion signal from PET image data. The present invention may also be embodied as a method of generating an image based on PET image data. Furthermore, the present invention may be embodied as a PET imaging system or as imaging software for a PET imaging system. The imaging software may be installed on, and used by, the processor 13 of the PET imaging system 10.

The steps and the order of the steps shown in Figure 8 are merely exemplary of an embodiment of the invention. Other steps and other orders are possible. For example, the step 82 of aggregating the PET image data into predefined time frames can alternatively be performed after the step 84 of assigning weighting factors to the PET image data.

Additionally or alternatively, if the PET image data that is provided in step 81 is already separated into predefined time frames, then it may not be necessary to further aggregate the PET image data into predefined time frames.

Claims

1. A method for determining a motion signal from positron emission tomography (PET) image data, the method comprising:

providing PET image data obtained from a PET scan of a patient, wherein the PET image data comprises spatial coordinates, time-of-flight (TOF) information and acquisition timing information for each line of response (LOR) of the PET scan;

allocating the PET image data to different TOF bins based on the TOF information;

2. The method of claim 1, comprising:

generating an image of the field of view, or a portion thereof, based on the PET image data, compensating for physiological movement of the patient during the PET scan based on the generated motion signal.

3. The method of claim 2, comprising:

using the motion signal to determine different physiological positions of the patient; and

generating an image for each physiological position of the patient.

4. The method of claim 3, comprising:

registering to each other the images that correspond to the different physiological positions of the patient.

5. The method of any preceding claim, wherein the PET image data allocated to a subset of the TOF bins are assigned a weighting factor of zero such that the PET image data allocated to the subset of TOF bins are excluded from the motion detection method.

6. The method of claim 5, wherein the PET image data allocated to TOF bins corresponding to TOF information that indicates that the time period between detection of the two photons is greater than a threshold time period are assigned a weighting factor of zero, so as to exclude physiological movement outside a central portion of the field of view that comprises the region of interest.

7. The method of any preceding claim, wherein the weighting factors are assigned to the PET image data based on both the TOF bin to which the PET image data is allocated and the spatial coordinates of the PET image data so as to give greater weighting to PET image data corresponding to a region of interest in the field of view of the PET scan.

8. The method of any preceding claim, whereina subset of the PET image data are assigned a weighting factor of zero based on the spatial coordinates indicating that the LOR is greater than a threshold distance from the origin, such that the motion detection method excludes physiological movement beyond the threshold distance from the origin.

9. The method of any preceding claim, wherein:

the PET image data are aggregated into predefined time frames such that the PET image data are sinogram data; and

the motion detection method is performed on the sinogram data.

10. The method of any preceding claim, comprising:

aggregating the PET image data into predefined time frames based on the acquisition timing information.

11. The method of any preceding claim, wherein the motion detection method comprises a dimensionality reduction technique such as principal component analysis to generate the motion signal.

12. The method of any preceding claim, wherein the weighting factors are determined based on a forward projection of the region of interest that indicates to what extent each TOF bin and/or LOR is expected to be allocated PET image data corresponding to the region of interest.

13. The method of claim, comprising:

forward projecting the region of interest to determine to what extent each TOF bin 5 and/or LOR is expected to be allocated PET image data corresponding to the region of interest.

14. The method of any preceding claim, wherein the region of interest is determined based on an image of the field of view generated from the PET image data.

10

15. The method of claim, comprising:

generating the image of the field of view from the PET image data so as to identify a region of interest corresponding to a part of the patient.

1516. A PET imaging system comprising:

a processor configured to:

20 allocate the PET image data to different TOF bins based on the TOF

information;

25 apply a motion detection method to the PET image data so as to generate a motion signal indicative of physiological movement of the patient, wherein the PET image data are used as an input for the motion detection method in accordance with the weighting factors assigned to the PET image data such that the motion signal is indicative of physiological movement in the region of interest.

30

17. Imaging software for a PET scanner, the imaging software configured to process PET image data obtained from a PET scan of a patient, wherein the PET image data comprises spatial coordinates, TOF information and acquisition timing information for each LOR of the PET scan, the imaging software configured to: allocate the PET image data to different TOF bins based on the TOF information; assign weighting factors to the PET image data based on at least the TOF bin to which the PET image data is allocated so as to give greater weighting to PET image data corresponding to a region of interest in the field of view of the PET scan; and