Method and device for the in-system calibration of gamma radiation detectors of a positron emission tomograph

Abstract

The invention relates to a method for the in-system calibration of gamma radiation detectors (15) of a positron emission tomograph (10), wherein the tomograph (10) has a receiving space (11) for introducing objects to be measured, around which the radiation detectors (15) are arranged, preferably in pairs opposite one another, with a defined position and orientation. According to the invention, the method comprises: introducing (103) an emission device (5) with a radioactive radiation source (7) into a reference position (R) within the receiving space (11), recording (105) coincidences between one or more pairs (15.1, 15.2) of the detectors (15), and - carrying out (107) virtual collimation, wherein determining (117) interaction positions (PI) comprises determining (119) an interaction depth (PI, T) within the detector (15.1) to be calibrated.

Description

[0001] Method and device for in-system calibration of gamma radiation detectors of a positron emission tomograph

[0002] The invention relates to a method for the in-system calibration of gamma radiation detectors of a positron emission tomograph, wherein the tomograph has a recording space for the introduction of objects to be measured, around which the radiation detectors are arranged, preferably opposite one another in pairs, with a defined position and defined orientation.

[0003] Positron emission tomography is a medical imaging and examination technique in which positron-emitting substances, such as modified sugar, are injected into a sample to visualize metabolic processes. Positron emission tomography uses radiation detectors designed to register pairs of gamma radiation photons, essentially simultaneously, emitted during the ß+ decay of a corresponding radionuclide. If a positron interacts with an electron, a process known as annihilation, for example, inside the sample, two high-energy photons, known as gamma photons, are emitted from the site of interaction at an angle of 180° (+ / - 0.25°). This radiation is also referred to as annihilation radiation. The goal of positron emission tomography is to localize this interaction site as precisely as possible.

[0004] Positron emission tomography has applications in research and clinical practice. Preclinical and clinical systems are known. Clinical systems have a large number of detectors (up to several hundred) and place particular demands on the efficiency of calibration and data processing of the positron emission tomography system. The same considerations also apply to systems with other application goals and a limited number of detectors, such as preclinical systems, which are often characterized by very high spatial resolution.

[0005] For detection, radiation detectors are used, which first convert the gamma photons into light in the optical spectrum in a scintillation crystal and then register this light on photosensors, for example silicon photomultipliers (SiPMs).

[0006] The invention primarily relates to the application field of gamma radiation detectors, which typically consist of an array of segmented scintillator elements (predominantly in clinical systems), a monolithic scintillator block (in some preclinical systems), or combinations thereof (mainly in the field of research) and are optically coupled to the photosensors. The radiation detectors each have a three-dimensional structure such that the photosensors define a detector surface on which a planar interaction position is typically detected in two dimensions, the so-called planar dimensions (e.g., marked by the x- and y-axes of a coordinate system). Due to their design, these scintillator elements in the radiation detectors have a certain height above the photosensors, e.g., in the direction of the z-axis of the aforementioned coordinate system.This means that the conversion of the gamma photon into optical photons, i.e., photons detectable by the photosensors, does not necessarily occur directly on the detector surface, i.e., in the plane of the photosensors, but rather at a distance or volume above it. With a scintillator element thickness of 10 mm, an offset of up to 10 mm can occur. This distance of interaction between the gamma radiation and the scintillator element and the detector surface is referred to below as the depth of interaction (DOI).

[0007] The interaction depth is typically not known as an element of the interaction position, but a suitable description based on the measured signals of the detector must be determined in calibrations, if technically possible, because in cases where the annihilation takes place off-center in the recording space, the gamma photons increasingly hit the detectors not perpendicularly, but at an angle, whereby the offsets reduce the localization accuracy of the interaction site.

[0008] So-called light-sharing detectors are one of the well-known concepts for determining interaction depth, although this also includes segmented architectures whose signal is split among multiple photosensors. From this light distribution, the interaction position of the gamma photon can be determined in all three dimensions for many architectures, such as a monolithic scintillator block, using statistical methods or machine learning algorithms. The best positioning results are typically achieved by individually calibrating the detectors followed by supervised machine learning. Generating the training data is crucial and represents a major challenge in practice.

[0009] Current calibration procedures often take place in a setup outside the system, making recalibration and quality control of the detectors within the system time-consuming and costly, if technically feasible at all. This would particularly affect clinical systems with several hundred detectors.

[0010] In addition, collimators are often used to precisely irradiate the detectors, which restrict the solid angle of the radiation from a calibration source located therein, resulting in longer calibration times.

[0011] Some known methods rely on planar irradiation along the sensor surface and aim to determine the DOI information based on, for example, parametric function variables. These methods are not functional on all known detectors, and their potential cannot be optimally exploited when using light-sharing detectors.

[0012] DE 102019215437 A1 discloses a device for calibrating a positron emission tomography system, which has a plurality of detectors and a recording chamber into which an element to be measured can be inserted. It proposes enclosing a radiation source in a collimator so that radiation from the radiation source exits the collimator in a directed manner. The radiation source is moved centrally through the recording chamber in a plurality of measurements so that the detectors of the tomograph can be successively calibrated. Although the apparatus design is fundamentally suitable for calibrating the planar dimensions, the high moving mass limits the precision, the range of motion, the travel speed, and thus the scalability of the system, with the additional challenge of the expanding radiation paths outside the collimator.Furthermore, no training data can be generated for calibrating the depth information. The invention was therefore based on the object of providing a method that overcomes the disadvantages described above as far as possible. In particular, the object of the invention was to provide a method with which the calibration of the detectors can be performed in a shorter time without compromising the precision of the calibration.

[0013] The invention solves the underlying problem by the method of the type described at the outset comprising the following steps:

[0014] - introducing an emission device with a radioactive radiation source into a reference position within the receiving space, wherein the emission device is designed to emit the radiation of the radiation source undirected within the receiving space,

[0015] - recording coincidences between one or more pairs of detectors, wherein the radioactive radiation source lies in a radiation path between the detectors of these pairs, and wherein each pair comprises a first detector to be calibrated and a second coincident detector, and

[0016] - performing a virtual collimation, wherein the virtual collimation preferably comprises:

[0017] - Determination of approximate gamma interaction positions in the coincident detectors,

[0018] - generating one or more virtual line paths, starting from the respective interaction positions of the coincident detectors, passing through the reference position of the radiation source, and preferably

[0019] - Determining interaction positions for the detectors to be calibrated based on the generated line paths, wherein determining the interaction position comprises determining a depth of interaction (DOI) within the detector to be calibrated.

[0020] The method is based on the realization that it is possible to introduce the radioactive radiation source into an existing tomography system, starting from a reference position, to calibrate the system's radiation detectors without having to dismantle them. The invention combines several approaches to utilize a synergy that was unexpectedly successful. The omnidirectional emission of radioactive radiation from the radiation source enables the simultaneous detection of interaction events in a large number of detector pairs, thereby significantly accelerating the overall calibration. Furthermore, the invention takes advantage of the fact that a physical collimator is not required in the acquisition room; instead, only virtual collimation is performed starting from the omnidirectional radiation source. The method according to the invention is very versatile and can be used for a wide variety of detector types.For example, the interaction depth can also be determined for detection concepts where the light is read from only one channel. In this case, the temporal progression of the signal intensity can then be analyzed and calibrated. The developed method is therefore very universally applicable to all detectors that benefit from individual calibration, with at least one dimension being the interaction depth (DOI).

[0021] Determining the interaction positions, including the interaction depth within the detector to be calibrated, results in a significant increase in the accuracy of position determination, particularly for off-center interaction events within the tomography system, because the so-called z-dimension, in which the interaction depth is determined, can cause measurement distortions to the greater extent the flatter the angle of incidence of the line path onto the radiation detector, especially when the line path impinges on the radiation detector at an angle.

[0022] By using virtual collimation to determine the interaction depth, the invention simultaneously increases the flexibility in the choice of geometry and the number of radiation sources, and thus the scalability of the calibration method. This determination of the interaction depth during in-system measurement makes it possible to use data from a wide portion of the irradiated solid angle range, for example, to generate data on planar interaction positions from non-perpendicular angles using iterative methods, as described below for preferred embodiments. No space is required for the collimator in the recording chamber, thus increasing the freedom of movement of the radioactive radiation source. The uncollimated radiation achieves a maximum irradiated solid angle range, which can be used for calibration and thus leads to significantly accelerated calibration data acquisition.In other words, the virtual collimation of the omnidirectional radiation in the tomograph's acquisition chamber and the analysis of the numerous irradiated detectors allow significantly more coincidences to be recorded and evaluated in a shorter time than previously possible. Furthermore, incompatibility of the collimator or associated structures with other components of the imaging system is avoided, especially in hybrid imaging systems, especially hybrids of PET and magnetic resonance imaging (MRI).

[0023] It is possible to know the reference position within the recording space only approximately in advance. For example, the calibration process can be started with an initial reference position that is only approximately known, and an improved reference position for the radiation source can then be determined from the determined interaction positions in order to start an iteration, as will be shown in the preferred developments of the invention.

[0024] The virtual collimation for a purely planar calibration is preferably carried out analogously to the virtual collimation described in: Peter Bruyndonckx, Cedric LemaTtre, Dennis Schaart, Marnix Maas, DJ van der Laan, Magalie Krieguer, Olivier Devroede, Stefaan Tavernier, Investigation of an in situ position calibration method for continuous crystal-based PET detectors, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, Volume 571 , Issues 1-2, 2007, Pages 304-307, ISSN 0168-9002, https: / / doi.Org / 10.1016 / j.nima.2006.10.088.

[0025] Virtual collimation for a purely planar calibration can also be performed analogously as described in: Andrea Gonzalez-Montoro, Larry A. Pierce II, William CJ Hunter, Antonio J. Gonzalez, and Robert S. Miyaoka, Validation of photon collimation techniques for monolithic PET detector calibration, Transactions on Radiation and Plasma Medical Sciences, 10.1109 / TRPMS.2020.3043397, IEEE.

[0026] By additionally determining the interaction depth from the obtuse- or acute-angled line paths, this information can be used as a corrected approximate interaction position in the purely planar calibration, both for the coincident detector and the detector to be calibrated.

[0027] In preferred embodiments, the determination of the interaction depth and optionally the planar interaction position is carried out analogously to: Yannick Kuhl, Stephan Naunheim, David Schug, Volkmar Schulz, and Florian Mueller, Angular Irradiation Methods for DOI Calibration of Light-Sharing Detectors — A Perspective for PET In-System Calibration, IEEE TRANSACTIONS ON RADIATION AND PLASMA MEDICAL SCIENCES, VOL. 7, NO. 7, SEPTEMBER 2023, pages 673 to 683, 10.1109TRPMS.2023.3272015.The procedure is preferably modified in that, in the above publication, the interaction depth is determined either from the collimated radiation path (the collimator generates a beam of defined width and not a line as in virtual collimation) and a known or previously determined / calibrated planar interaction position, or from two intersecting collimated radiation paths (DOI and planar simultaneously), whereas according to the invention, the generated line paths are used for this purpose. In other words, according to the invention, the interaction depth is preferably determined from the generated line path and a known or previously determined / calibrated planar interaction position, and / or from several intersecting line paths (DOI and planar simultaneously).According to the invention, several line paths of different directions can be used quite easily, since the solid angle is not restricted, which enables a higher precision of the interaction position determination.

[0028] Alternatively or additionally, the interaction depth can be determined analogously to the explanations in the following publication: Xin Li, Li Tao, Craig S. Levin and Lars R. Furenlid, Virtual Point Source Synthesis method for 3D Scintillation Detector Characterization, 10.13140 RG.2.2.19981 .46567. Here, too, the interaction depth is described as a function of several generated, intersecting radiation paths with the aid of a collimator.

[0029] As already indicated above, the method is advantageously further developed by comprising the step of providing or determining an initial reference position as the reference position. The initial reference position is, in a sense, the most probable annihilation position of the coincidence. This initial reference position can be determined from a first data set, e.g., a first image acquired with the tomograph in the uncalibrated state. If known, a predetermined coordinate can alternatively be specified as the initial reference position. In a preferred embodiment of the invention, determining the interaction positions further comprises determining a planar interaction position (x, y) on a detector surface of the respective detector of these pairs to be calibrated.The previously determined interaction depth, if available, can preferably be included in the determination of the planar interaction positions on the detector surface.

[0030] The interaction depth is preferably a function of one of the generated line paths and a planar interaction position, and / or a function of several of the generated line paths, in particular one or more actual or approximate intersection points of the generated line path.

[0031] The planar interaction position for the former variant can be taken from a previous calibration, e.g., from an in-system measurement with a line path perpendicular to the detector to be calibrated, or an oblique line path if a previous interaction depth is already known. In other words, non-perpendicular line paths can also be used after determining the DOI in the detector to be calibrated.

[0032] In the second of the aforementioned variants, the gamma interactions in the intersection points can be found by determining the similarity of their signals, e.g., using a k-nearest neighbor method, as described for collimators in: Yannick Kuhl, Stephan Naunheim, David Schug, Volkmar Schulz, and Florian Mueller, Angular Irradiation Methods for DOI Calibration of Light-Sharing Detectors — A Perspective for PET In-System Calibration, IEEE TRANSACTIONS ON RADIATION AND PLASMA MEDICAL SCIENCES, VOL. 7, NO. 7, SEPTEMBER 2023, pages 673 to 683, 10.1 109TRPMS.2023.3272015. This directly results in an interaction depth and a planar interaction position for building or training a model. A prior calibration can be integrated into this method.One advantage of the invention is seen in the fact that many different line paths can be used, so that the individual line paths of the individual gamma interactions are more diverse, which simplifies the search for similar signals. Furthermore, the use of uncollimated radiation results in a large number of intersection points being obtained within the detector volume of the detector to be calibrated, with a significantly reduced number of individual measurements compared to collimator methods. Another advantage of the invention is that a collimator needs to be moved from position to position, and in the case of two intersecting radiation paths, only a single intersection point is generated, whereas the omnidirectional radiation source can, in principle, generate an infinite number of intersection points even if the radiation source has only two distinguishable source positions (e.g., point sources).As a result, a static setup (without a motor to move the source position) for a collimator is significantly more difficult than for the uncollimated case.

[0033] In a further preferred embodiment of the invention, the detector to be calibrated within a pair is located closer to the radiation source than the coincident detector. This is particularly advantageous when both detectors, the detector to be calibrated and the coincident detector, have the same spatial resolution, or the coincident detector has a reduced spatial resolution. In other words, the reference position of the radiation source is closer to the detector to be calibrated than to the coincident detector. The coincident detector is the one used to generate the line path for virtual collimation. It can also be referred to as the collimation detector.The positional inaccuracy that is unavoidable in the evaluation of the sensor values ​​of the coincident detector has a less pronounced impact on the interaction position of the detector to be calibrated, which is then determined by the line path, due to the greater distance of this coincident detector from the reference position. If an iterative optimization is to be performed, this asymmetrical arrangement of the radiation source between the detectors allows the number of necessary iteration steps to be reduced even more, the closer the radiation source is arranged to the detector to be calibrated. In a further preferred embodiment, the recording of the coincidences comprises pairs of detectors that are opposite one another, preferably parallel, and / or that are aligned at a right angle relative to the generated line paths.

[0034] In a further preferred embodiment, the recording of the coincidences comprises pairs of detectors that are angularly opposed to one another and / or in which at least the detector to be calibrated is aligned at an acute or obtuse angle relative to the generated line paths. The detectors of a pair can be arranged relative to one another in various ways. Detectors within a ring perpendicular to the longitudinal axis of the tomograph can form a pair, but detectors that are spaced apart from one another in the axial direction of the tomograph can also form a pair. In all these cases, the line path does not impinge on the detector surfaces of the detectors perpendicularly, but rather at an acute or obtuse angle.

[0035] In a further preferred embodiment, determining the approximate gamma interaction positions comprises determining a planar position approximation and / or an interaction depth approximation.

[0036] The method preferably further comprises adjusting the approximate interaction positions in the coincident detectors by means of an iteration based on the determined interaction positions of the detectors to be calibrated. According to the invention, this means that after determining the interaction position on the detector to be calibrated in combination with, for example, the reference position of the radiation source, an optimized interaction position can be calculated in the coincident detector, which then, for example, in a further iteration step, again with the reference position of the radiation source, produces a further optimized interaction position on the detector to be calibrated.

[0037] In a further preferred embodiment, the method comprises the step of selecting the detected coincidences as a function of a predetermined threshold. In this case, coincidences that do not reach a predetermined threshold with regard to the signal intensities or other parameters on the photosensors or their combined measured values ​​are discarded. Depending on the definition of the threshold, certain thresholds must be reached, exceeded, or undercut in order to be able to be selected successfully. In a further preferred embodiment, the invention further comprises creating orTraining one or more models using training data, with the determined interaction positions as training output data and training input data that are indicative of a light intensity distribution or a temporal light intensity profile measured for the determined interaction positions.

[0038] The model to be created or trained is preferably a statistical data model or a K-model, more preferably a machine learning model, particularly preferably a Gradient Tree Boosting (GTB) model or a neural network (NN), or a combination of several of these models. The training input data is the so-called labeled data for the model to be created or trained.

[0039] In preferred embodiments, the training input data comprises one, several or all of the following:

[0040] - Signal values ​​of the photosensors of the detectors, especially photon counts, raw,

[0041] - Light distribution of the photosensors of the detectors, first moment,

[0042] - Light distribution of the photosensors of the detectors, second moment,

[0043] - Readout channel ID of the detectors that have the highest optical photon count,

[0044] - Photon sum projections,

[0045] - total photon sum,

[0046] - squared channel intensities on a predefined channel range, and / or

[0047] - Channel with the highest signal intensity, normalized to the sum of the channel intensities, of a predefined channel range.

[0048] The creation or training of the model in the case of the application of a neural network can, for example, be carried out as described in Peter Bruyndonckx, Cedric LemaTtre, Dennis Schaart, Marnix Maas, DJ van der Laan, Magalie Krieguer, Olivier Devroede, Stefaan Tavernier, Investigation of an in situ position calibration method for continuous crystalbased PET detectors, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, Volume 571 , Issues 1-2, 2007, Pages 304-307, ISSN 0168-9002, https: / / doi.Org / 10.1016 / j.nima.2006.10.088.

[0049] The creation or training of the model in the case of the application of a GTB model can, for example, be carried out as described in F. Müller, D. Schug, P. Hallen, J. Grahe and V. Schulz, "Gradient Tree Boosting-Based Positioning Method for Monolithic Scintillator Crystals in Positron Emission Tomography," in IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 2, no. 5, pp. 411-421 , Sept. 2018, doi: 10.1109 / TRPMS.2018.2837738

[0050] The creation or training of the model can alternatively also be carried out as described in Yannick Kuhl, Florian Mueller, Stephan Naunheim, Matthias Bovelett, Janko Lambertus, David Schug, Bjoern Weissler, Eike Gegenmantel, Pierre Gebhardt, Volkmar Schulz, A finely segmented semi-monolithic detector tailored for high-resolution PET, Med Phys. 2024;1-16. DOI: 10.1002 / mp.16928

[0051] In a further preferred embodiment, the method is further developed by the step: Determining, as a function of the created or trained model:

[0052] - an optimized interaction position; and / or

[0053] - an optimized reference position of the radiation source, whereby the optimized

[0054] Reference position is preferably provided as a new initial reference position for a repeat execution of the procedure; and / or

[0055] - an optimized position and / or orientation of one or more detectors, wherein the optimized position and / or orientation is preferably provided as a new position and / or orientation of the detectors for a renewed execution of the method.

[0056] Thanks to the determination of the interaction depth during virtual collimation, the method also allows, and preferably includes, in further preferred embodiments, the use of the generated data, i.e. in particular the data of the generated interaction positions, and preferably of the created or trained model, to evaluate or test a previous model, wherein the evaluation preferably comprises quantifying differences in the behavior of the detectors. This includes, for example, determining the positioning error distribution (PSF, "point-spread function"), on the basis of which further metrics can be determined, such as the mean absolute error (MAE), the bias vector, and the spatial resolution (SR) of each detector. These metrics provide information about how well the previous model is applicable to the current behavior of the detector.Detector behavior can change, for example, due to changing measurement conditions (e.g., temperature, power supply) or due to component aging. Such quantification may be necessary as part of clinical quality control routines.

[0057] Alternatively or additionally, the invention enables, and preferably includes, performing a time calibration using the interaction depth; and / or performing an energy calibration using the interaction depth. The time calibration is carried out in preferred embodiments as described by Stephan Naunheim et al. 2023 Phys. Med. Biol. 68 025013, Analysis of a convex time skew calibration for light sharing-based PET detectors, DOI 10.1088 / 1361-6560 / ACA872.

[0058] Energy calibration, and optionally time calibration, is preferably performed as described in Mueller F, Naunheim S, Kuhl Y, Schug D, Solf T, Schulz V. A semi-monolithic detector providing intrinsic DOI encoding and sub-200 ps CRTTOF capabilities for clinical PET applications. Med Phys. 2022;49:7469–7488. https: / / doi.org / 10.1002 / mp.16015

[0059] In a further preferred embodiment, the spatial shape of the radiation source is selected from the following:

[0060] Sphere, polyhedron, cylinder, hollow cylinder, disc, ring, toroid, helix, or a combination of several of the above.

[0061] In preferred embodiments, the radiation source has as radioactive component a ß + Emitter on.

[0062] In a further preferred embodiment, the radiation source is a first radiation source, the emitting device has one or more further radiation sources, and the step of introducing further comprises: introducing one or more further radioactive radiation sources into a reference position within the receiving space.

[0063] In a further preferred embodiment, the method comprises the step of: adjusting the reference position starting from the initial reference position, wherein the adjustment is preferably carried out in an iterative reconstruction process.

[0064] The radiation sources are preferably positioned on the emission device in relation to one another in a specific geometric arrangement, wherein the geometric arrangement is preferably selected from the following:

[0065] Sphere, polyhedron, cylinder, hollow cylinder, disc, ring, toroid, helix, or a combination of several of the above.

[0066] In a further preferred embodiment, the method comprises the step:

[0067] Inserting additional detectors at known reference positions inside or outside the recording chamber. Using the additional detectors, additional solid angles can be covered inside or outside the recording chamber. Detectors similar to the radiation detectors installed in the system, or detectors with alternative designs, can be used. For example, detectors with higher spatial resolution can also be used outside the recording chamber. High gamma positioning accuracy can be achieved, in particular, if the coincident detector is irradiated along the spatial dimensions for which it achieves the best spatial resolution. Due to the less severe space restrictions outside the recording chamber, planning detectors with a different design, for example, with very flat scintillator elements, can also be used.The additional detector(s) could, for example, be detectors that do not require dedicated calibration effort. External detectors with collimators or alternative scintillator materials, such as bismuth germanate (BGO) scintillator elements, could also be used. External detectors with particularly high detection efficiency, such as large volumes of scintillation material, could also be used.

[0068] In a further preferred embodiment, the reference position(s) of the radiation source is / are a first reference position, and the method further comprises:

[0069] - Moving the radiation source to one or more additional reference positions, and

[0070] - Repeating one, several or all steps of the method of one of the preferred embodiments described above for the further reference position(s).

[0071] The invention has been described above with reference to a first aspect and the method according to the invention. In a further aspect, the invention also relates to a device for in-system calibration of radiation detectors of a positron emission tomograph, in particular in a method according to one of the preferred embodiments described above. The tomograph has a receiving space for introducing objects to be measured, around which the radiation detectors are arranged, preferably in pairs opposite one another, with a known position and known orientation, and has a mounting interface for attaching the device.

[0072] The invention also achieves the object described above in this device, in that the device has a mounting interface corresponding to the mounting interface of the tomograph for the reversibly detachable attachment of the device in a calibration position on the tomograph, and an emission device with a radioactive radiation source, wherein the emission device is configured to arrange the radiation source in a reference position within the recording space when the device is attached to the mounting interface, and wherein the emission device is configured to emit the radiation of the radiation source undirected within the recording space, wherein the device has a data processing device configured to provide coincidences recorded by the detectors between one or more pairs of the detectors,wherein the radioactive radiation source lies in a radiation path between the detectors of these pairs, and wherein each pair has a first detector to be calibrated and a second, coincident detector, and wherein the data processing device is configured to perform virtual collimation, which preferably comprises determining approximate gamma interaction positions in the coincident detectors for each pair, generating one or more virtual line paths based on the respective interaction positions of the coincident detectors that pass through the reference position of the radiation source, and / or determining interaction positions for the detectors to be calibrated based on the generated line paths in at least one dimension, wherein the at least one dimension comprises an interaction depth.

[0073] The device according to the invention utilizes the same findings as the method according to the invention. Preferred embodiments of the method according to the invention of the first aspect are also preferred embodiments of the device, and vice versa, which is why, to avoid repetition, reference is made to the above explanations in this regard.

[0074] The device according to the invention is advantageously further developed in that the device, in particular the data processing device, is designed to carry out the steps of the method according to one of the preferred embodiments described above.

[0075] The invention is described in more detail below using preferred embodiments with reference to the accompanying figures.

[0076] Fig. 1 shows the schematic structure of a positron emission tomograph including an associated calibration device according to an embodiment,

[0077] Fig. 2 shows a schematic sequence of the method according to the invention according to an embodiment,

[0078] Fig. 3 shows the schematic view of a radiation detector arrangement of the device according to Fig. 1 in the method according to Fig. 2, Figs. 4a, b schematic illustrations of the virtual collimation from the method according to Fig. 2 in a first variant,

[0079] Figs. 5a, 5b are schematic illustrations of the virtual collimation from the method according to Fig. 2 in a second variant, and

[0080] Figs. 6a, b, c schematic spatial views of radiation sources for use in the method and device according to the invention.

[0081] Fig. 1 shows a positron emission tomography scanner 10. The scanner 10 has a recording chamber 11 configured to accommodate the object to be measured, for example, on a patient bed 13. The patient bed 13 can be moved uniaxially or multiaxially within the recording chamber 11 and out of it or into it.

[0082] The tomograph 10 further comprises an array of gamma radiation detectors 15. The gamma radiation detectors 15 each comprise a photosensor 17 with a detector surface 18, above which a scintillator element 19 is mounted. The gamma radiation detectors 15, and with them the detector surfaces 18 and the scintillator elements 19, are each arranged at defined locations and orientations within the tomograph 10. According to the invention, the term "defined" means that the position and orientation of the detectors are known with sufficient precision, with an accuracy in the single-digit millimeter range being sufficient. Their position is preferably known at least approximately. The gamma radiation detectors 15 surround the recording chamber 11 and are preferably arranged opposite one another in pairs.

[0083] A first mounting interface 21 is provided on the tomograph 10. The first mounting interface 21 is reversibly detachably connected to a correspondingly designed second mounting interface 3 of a device 1 for calibrating the tomograph 10.

[0084] The device 1 is configured for in-system calibration of the radiation detectors 15 of the tomograph 10 and is intended to be introduced into the recording space 11 of the tomograph 10 when it is connected to the tomograph 10 in the calibration position shown. The device 1 has, in addition to the second mounting interface 3, an emission device 5. The emission device 5 is fixedly arranged on the device 1 relative to the second mounting interface 3 and carries one (or more) radiation source(s) 7. The radiation source 7 is typically accommodated in a container and enclosed by it. The enclosure can reduce physical limitations (“positron ranges”, “non-colinearity”), but does not represent a shield that would restrict the radiation source to a directionally directed emission. The radiation source is configured to emit radiation non-directionally. +-radiation. The local position of the radiation source 7 is thus also defined at any time via the local position of the mounting interfaces 21, 3, which is known by means of the control system of the tomograph 10 (see above), at least within the scope of the available coordinate information of the control system of the tomograph 10 and the assembly and manufacturing tolerances to be accepted.

[0085] The device 1 further comprises a data processing device 9, which is equipped with a processor (not shown), memory, and other data processing means, and is configured to receive, process, and provide data signals from the radiation detectors 15 of the tomograph 10. In preferred embodiments, the data processing device 9 can also be configured to simultaneously control the tomograph 10 and, for this purpose, be connected to the tomograph in a signal-conducting manner.

[0086] The device 1 is configured by means of the data processing device 9 to calibrate the tomograph 10 in the calibration position shown, wherein the method of Fig. 2 described below is preferably used for this purpose.

[0087] The method according to the invention begins, by way of example, in a first step 101, in which an initial reference position Ri is provided or determined as the reference position R for the radioactive radiation source. If it is not to be specified as a predefined reference position, the reference position R can also be determined in advance using a first, uncalibrated image after the radiation source has been introduced into the recording chamber of the tomograph 10. For the illustration of the method, it is assumed in the present exemplary embodiment that the reference position R is initially provided. With the reference position R thus known, the emission device 5 with the radioactive radiation source 7 is introduced into the reference position R within the recording chamber 11 in the next method step 103.The emission device 5 is constructed in such a way that it emits the radiation of the radioactive radiation source 7 unhindered and undirected within the receiving space 11.

[0088] In a next step 105, coincidences are recorded between one or more pairs of detectors 15, wherein the radioactive radiation source 7 is located in a radiation path between the detectors 15 of these pairs, and wherein each pair of detectors 15 has a first detector 15.1 to be calibrated and a second coincident detector 15.2 opposite along the radiation path.

[0089] In the next step 107, a selection of the recorded coincidences is optionally selected, with only those coincidences whose signal intensity at the detector surfaces 18, where photosensors are arranged, fulfills predefined conditions being further processed. The selection step could also be preceded by signal filtering before the recording of the coincidences from step 105.

[0090] In a next method step 109, a virtual collimation is performed, preferably analogous to and expanded by the additional parameters as explained above in the publications incorporated by reference in the general section, wherein the virtual collimation preferably comprises several substeps. The virtual collimation comprises, in step 110, determining one or more approximate gamma interaction positions PI A in the coincident detectors 15.2, and, in a next step 115, generating one or more virtual line paths L, starting from the respective approximate interaction positions PI A of the coincident detectors 15.2, which extend through the reference position of the radiation source.If the radiation source is not a point source but a radiation source with spatial extension, the reference position is preferably represented by a most probable position determined from the radiation source arrangement and approximately determined interaction positions in the coincident detector 15.2 and the detector 15.1 to be calibrated, as well as the time difference of both interactions of a coincidence.

[0091] In a next step 117, an interaction position Pi is determined on the basis of the generated line paths L for the detectors 15.1 to be calibrated, wherein the determination of the interaction position Pi comprises at least the determination of an interaction depth PI T within the detector 15.1 to be calibrated, represented by a sub-step 119.

[0092] Preferably, determining the interaction position Pi in step 117 also comprises determining a planar interaction position PI,P, represented by sub-step 121, on a detector surface 18 of the detector 15.1 to be calibrated.

[0093] Preferably, the reference position R is selected such that the distance between the radiation source 7 and the detector 15.1 to be calibrated is smaller than the distance between the radiation source 7 and the coincident detector 15.2, see Figures 4a, b. The smaller the distance between the detector 15.1 to be calibrated and the radiation source 7 in relation to the aforementioned distance between the radiation source 7 and the coincident detector 15.2, the smaller the effect of the positioning inaccuracy of the approximate interaction position PI A on the determined interaction position Pi,p in the plane of the detector surface 18. The interaction depth PI T, also referred to as DOI, is preferably a function of one or more generated line paths L, or a function of a generated line path and a planar interaction position Pi,p, see Figures 4a, b.

[0094] A fairly accurate determination of the planar interaction position Pi,p is possible, see Fig. 4a, if coincidences are evaluated which hit a pair of detectors which are preferably parallel to each other and / or which the line path L hits at a right angle when no information on the interaction depth PI T is yet available.

[0095] An approximate interaction position preferably comprises determining a planar position approximation in a step 111 and / or determining an interaction depth approximation in a step 113. One of the two approximations, or both, can be performed. The approximations can be performed sequentially.

[0096] Once interaction positions PI,P; PI T have been determined, these can be used in a next method step 123 to adjust the approximate interaction positions in the coincident detectors 15.2 by means of an iteration based on this determined interaction position of the detectors 15.1 to be calibrated. The virtual collimation can then be iterated based on these determination results to improve the accuracy of the determination of the interaction positions.

[0097] Alternatively or additionally, represented by method step 125, after determining the interaction positions during virtual collimation, the reference position R can be adjusted in order to obtain an improved initial reference position Ri. Alternatively or additionally, to determine further interaction position data, the radiation source 7 can be moved to a new reference position R', R", and the method can be repeated for the new reference position. Alternatively or additionally, to determine further interaction position data, the position of the detector 15.1 to be calibrated can be corrected.

[0098] However, the determined interaction positions from step 117 can also be used to create or train one or more models, preferably Kl models, particularly preferably a GTB model or a neural network, using training data in a method step 127, wherein the determined interaction positions Pi,p and PI A are training output data, and the signal data obtained from the radiation detectors 15, which are indicative of a light intensity distribution at the determined interaction positions, are training input data.

[0099] In step 129, an optimized interaction position and / or an optimized reference position of the radiation source can then be determined as a function of the model thus created or trained, wherein the optimized reference position is preferably provided as a new reference position for a renewed execution of the method, and / or an optimized position and / or orientation of one or more detectors, wherein the optimized position and / or orientation is preferably provided as a new position and / or orientation of the detectors for a renewed execution of the method.

[0100] In a step 131, the method further comprises the step of adjusting the reference position starting from the initial reference position, wherein the adjustment is preferably carried out in an iterative reconstruction process.

[0101] Finally, it is also optionally provided at this point that in step 133 the radiation source 7 is moved by means of the emission device 5 to one or more further reference positions R', R", and then one, several or all steps of the method are repeated for the respective new reference position.

[0102] Fig. 3 shows an exemplary arrangement of radiation detectors 15 of the tomograph 10 according to Fig. 1 to illustrate the mode of operation of the device 1 and the method sequence according to Fig. 2 in a first variant.

[0103] A radiation source 7 has been introduced into the recording space 11 by means of the emission device 5 into a reference position R. In addition to the reference position R shown, the radiation source 7 can and should be guided successively through several further positions R', R". The radiation source 7 radiates undirected within the recording space 11, so that coincidences are registered between a plurality of detector pairs. For example, the radiation source 7 is arranged in a first beam path Si between a first detector 15.1 to be calibrated and a more distant, coincident second detector 15.2; at the same time, however, the radiation source 7 is also located in a second beam path S2 between the detector 15.1 to be calibrated and another coincident detector 15.2'.Similarly, a plurality of coincidences are registered between a plurality of detector pairs in the recording room 11, all of which are available for calibrating the detectors.

[0104] Figures 4a and 4b show various stages of virtual collimation. Fig. 4a depicts a gamma event in which the emitted gamma radiation impinges essentially at right angles on a detector 15.1 to be calibrated, which is located close to the radiation source 7, and on an opposite coincident detector 15.2, which is located relatively farther from the radiation source 7. By evaluating the signal at the coincident detector 15.2, an approximate interaction position Pi is determined within the framework of a positional predetermination. ais determined, and from this, for example, its signal intensity center of gravity, a line path L is constructed through the reference position R, which then intersects the detector surface 18 of the detector 15.1 to be calibrated. Due to the perpendicular incidence, this intersection point is, to a good approximation, the planar interaction position Pi p for the detector to be calibrated.

[0105] At the same time, however, for the same position R of the radiation source 7, a coincidence can also be recorded between the detector 15.1 to be calibrated and another coincident detector 15.2', in which the line path L, constructed in the same way, impinges on the detector 15.1 to be calibrated at an acute angle, or alternatively at an obtuse angle, but in any case at an angle other than 90°. This allows not only the planar interaction position but also the interaction depth PI T to be determined. The interaction depths PI,P; PI T can be used as training input data for training a Kl model in the manner described above.

[0106] The previous figures show a first method variant for determining the interaction depth. Fig. 5a and Fig. 5b show a second variant of the method according to Figure 2 for determining the interaction depth PI T ZU. In the variant according to Fig. 5a, the emission device has two radiation sources 7, i.e., a spatially extended radiation source arrangement, which, together with several coincident detectors 15.2, 15.2', 15.2", lead to the generation of line paths Li, L2, L3, L4, which each impinge at an angle on a detector 15.1 to be calibrated. Each of the line paths Li, L2, L3, L4 is to be understood symbolically for a multitude of coincidences, and not strictly for a single pair of gamma interactions. In the detector 15 to be calibrated.1, intersection points Ci, C2, C3, C4 are identified between the line paths, and at these intersection points Ci, C2, C3, C4, the planar interaction position PI,P, as well as the interaction depth PI T, can be directly determined from the angular relationships of the line paths Li, L2, L3, L4, indicated in Fig. 5a by the dashed lines for one of the intersection points Ci, C2, C3, C4. To avoid repetition, reference is made to the explanations in the general section (cf. k-Nearest-Neighbor method).

[0107] Fig. 5b illustrates the same principle, although for detector pairs in which the coincident detectors 15.2, 15.2' and 15.2" are axially adjacent and not in the circumferential direction as in Fig. 5a. Also in the constellation of Fig. 5b, several line paths Li, L2, L3, L4, L5, and Le are formed by their angular impact on the detector 15.1 to be calibrated, intersection points C nwhich each provide the determination of interaction positions as in Fig. 5a.

[0108] Figures 1, 3, 4a, b, and 5 each show one (Figure 5: two or three) essentially point-shaped or spherical radiation source(s) for carrying out the calibration method. However, it is alternatively possible and provided according to the invention to use other geometric shapes for emitting ß + -radiation in the recording room. Figures 6a, 6b, and 6c show a stylized recording room 11 as an example.

[0109] In Fig. 6a, a toroidal radiation source 7' is introduced into the receiving space 11, which, for example, can be designed as a closed hollow body with one or more ß + -emitter. For example, a fluid containing fluid-18 can be used as the radioactive medium. Such fluids are also injected into patients or test objects in general.

[0110] In the model shown in Fig. 6b, an array of several rings or toroidal bodies arranged parallel to one another, represented by reference numeral 7", is used as the radioactive radiation source. By means of a radioactive radiation source 7" spatially expanded in this way, even more detectors 15 can be reached simultaneously. The spacing of the volume bodies of the array 7" should be sufficiently large so that the signals on the detector surfaces 18 can be distinguished. Suitable distances can be determined experimentally.

[0111] In Fig. 6c, the arrangement of the radiation sources according to Fig. 5b is extended by several rods or hollow cylinders running in the axial direction, so that a body structure of the radioactive radiation source 7'" formed from several elements is created.

[0112] Other geometric shapes not shown individually here, which were referred to above in the general part, are conceivable in the same way and are provided for in the invention.

[0113] 1 Device 3 (second) mounting interface, device 5 Emission device 7 Radiation source(s) 7' Radiation source 7“ Radiation source 7'“ Radiation source

[0114] 9 Data processing device 10 Positron emission tomograph 11 Recording room

[0115] 13 Patient bed 15 Gamma radiation detectors 15.1 Detector to be calibrated 15.2 Coincident detector 15.2' Additional coincident detector 15.2“ Additional coincident detector 17 Photosensor 18 Detector surface 19 Scintillator element 21 (first) mounting interface, tomograph 101 Process step 103 Process step 105 Process step 107 Process step 109 Process step 110 Process step 111 Process step 113 Process step 115 Process step 117 Process step 119 Process step 121 Process step 123 Process step 125 Process step 129 Process step 131 Process step 133 Process step

[0116] Ci, C2, C3, C4, Cn

[0117] Intersection Li, L2, L3, l_4, Ls, Le,

[0118] Line path

[0119] Pi interaction position

[0120] PI A Interaction position, approximated

[0121] PI.P Interaction position, planar PI.T Interaction depth (DOI),

[0122] R Reference position

[0123] Ri initial reference position

[0124] Si, S2 ray path

Claims

Claims:

1. A method for in-system calibration of gamma radiation detectors (15) of a positron emission tomograph (10), wherein the tomograph (10) has a receiving space (11) for introducing objects to be measured, around which the radiation detectors (15) are arranged, preferably in pairs opposite one another, with a defined position and orientation, characterized in that the method comprises: - introducing (103) an emission device (5) with a radioactive radiation source (7) into a reference position (R) within the receiving space (11), wherein the emission device (5) is designed to emit the radiation of the radiation source (7) undirected within the receiving space (11), - recording (105) coincidences between one or more pairs (15.1, 15.2) of detectors (15), wherein the radioactive radiation source (7) lies in a radiation path (Si, S2) between the detectors (15.1, 15.2) of these pairs, and wherein each pair has a first detector (15.1) to be calibrated and a second coincident detector (15.2), and - performing (107) a virtual collimation, wherein the virtual collimation comprises: - Determining (110) approximate gamma interaction positions (PI A) in the coincident detectors (15.2), - generating (115) one or more virtual line paths (Li, L2) starting from the respective approximate gamma interaction positions (PI A) of the coincident detectors (15.2) passing through the reference position (R) of the radiation source, and - determining (117) interaction positions (Pi) for the detectors (15.1) to be calibrated on the basis of the generated line paths (Li, L2), wherein the determining (117) of the interaction positions (Pi) preferably comprises the determining (119) of an interaction depth (PI T) within the detector (15.1) to be calibrated.

2. Method according to claim 1, comprising the step of: Providing or determining an initial reference position (Ri) as a reference position (R).

3. The method according to claim 1 or 2, wherein determining (117) the interaction positions further comprises determining (121) a planar interaction position (PI.P) on a detector surface (18) of the respective detector (15.1) to be calibrated of these pairs.

4. Method according to one of the preceding claims, wherein the interaction depth (PI T) is a function of a generated line path (Li, L2) and a planar interaction position (PI, P), and / or a function of several of the generated line paths (Li, L2).

5. Method according to one of the preceding claims, wherein the detector (15.1) to be calibrated within a pair has a smaller distance to the radiation source (7) than the coincident detector (15.2).

6. Method according to one of the preceding claims, wherein the recording of the coincidences comprises pairs of detectors (15) which are opposite one another, preferably parallel, and / or which are aligned at a right angle relative to the generated line paths (Li).

7. Method according to one of the preceding claims, wherein the recording of the coincidences comprises pairs (15.1, 15.2) of detectors which are angularly opposite one another and / or in which at least the detector (15.1) to be calibrated is aligned at an acute or obtuse angle relative to the generated line paths (L2).

8. The method according to any one of the preceding claims, wherein determining (110) the approximate gamma interaction positions comprises determining - (111) a planar position approximation, and / or - (113) an interaction depth approximation.

9. Method according to claim 7 or 8, comprising adjusting (123) the approximate interaction positions (PI A) in the coincident detectors (15.2) by means of an iteration on the basis of the determined interaction positions (PI,P, PI T) of the detectors (15.1) to be calibrated.

10. Method according to one of the preceding claims, comprising the step of: selecting (107) the detected coincidences as a function of a predetermined threshold value. 11 . Method according to one of the preceding claims, comprehensive the step: - Creating or training (127) one or more models using training data, with the determined interaction positions (PI,P, PI T) as training output data, and training input data that are indicative of a light intensity distribution measured for the determined interaction positions (PI,P, IT).

12. The method of claim 11, wherein the training input data comprises one, several or all of the following: - Signal values ​​of the photosensors of the detectors, - Light distribution of the photosensors of the detectors, first moment, - Light distribution of the photosensors of the detectors, second moment, - Readout channel ID of the detectors that have the highest optical photon count, - Photon sum projections, - total photon sum, - squared channel intensities on a predefined channel range, and / or - Channel with the highest signal intensity, normalized to the sum of the channel intensities, of a predefined channel range.

13. A method according to any one of claims 11 or 12, comprising the step: Determine (129), as a function of the created or trained model: - an optimized interaction position; and / or - an optimized reference position of the radiation source, whereby the optimized Reference position is preferably provided as a new initial reference position for a repeat execution of the procedure; and / or - an optimized position and / or orientation of one or more detectors, wherein the optimized position and / or orientation is preferably provided as a new position and / or orientation of the detectors for a renewed execution of the method.

14. Method according to one of the preceding claims, comprising one, several or all of the steps: - Using the data of the generated interaction positions, and preferably the created or trained model, to evaluate or test a previous model, wherein the evaluation preferably involves quantifying differences in behavior the detectors (15), for example by determining the positioning error distribution (PSF, “point-spread-function”); - Performing a time calibration using the interaction depth; and / or - Perform energy calibration using interaction depth.

15. Method according to one of the preceding claims, wherein the spatial shape of the radiation source is selected from the following: Sphere, polyhedron, cylinder, hollow cylinder, disc, ring, toroid, helix, or a combination of several of the above.

16. Method according to one of the preceding claims, wherein the radiation source (7) contains as radioactive component a ß + emitter.

17. Method according to one of the preceding claims, wherein the radiation source (7) is a first radiation source, the emitting device comprises one or more further radiation sources, and wherein the step of introducing further comprises: introducing one or more further radioactive radiation sources into a reference position within the receiving space.

18. Method according to one of claims 2 to 17, comprising the step of: adapting (131) the reference position starting from the initial reference position, wherein the adaptation is preferably carried out in an iterative reconstruction process.

19. The method according to claim 17 or 18, wherein the radiation sources on the emission device are positioned relative to each other in a specific geometric arrangement, wherein the geometric arrangement is preferably selected from the following: Sphere, polyhedron, cylinder, hollow cylinder, disc, ring, toroid, helix, or a combination of several of the above.

20. A method according to any one of the preceding claims, comprising the step: - Introduction of additional detectors at known reference positions inside or outside the recording room 21 . Method according to one of the preceding claims, wherein the reference position(s) of the radiation source is / are a first reference position, and the method further comprises: - moving (133) the radiation source (7) to one or more further reference positions (R', R"), and - Repeating one, several or all steps of the method according to one of the preceding claims for the further reference position(s) (R', R").

22. Device (1) for in-system calibration of radiation detectors (15) of a positron emission tomograph (10), in particular in a method according to one of the preceding claims, wherein the tomograph (10) has a receiving space (11) for introducing objects to be measured, around which the radiation detectors (15) are arranged, preferably in pairs opposite one another, with a defined position and orientation, and has a mounting interface (21) for attaching the device (1), characterized in that the device (1) has a mounting interface (3) corresponding to the mounting interface (21) of the tomograph (10) for reversibly detachably attaching the device (1) in a calibration position on the tomograph (10), and - an emission device (5) with a radioactive radiation source (7), wherein the emission device (5) is configured to arrange the radiation source (7) in a reference position (R) within the receiving space (11) when the device (1) is attached to the mounting interface (21), and wherein the emission device (5) is configured to emit the radiation of the radiation source (7) undirected within the receiving space (11), wherein the device (1) has a data processing device (9) which is configured to - to provide coincidences recorded by the detectors (15) between one or more pairs (15.2, 15.2) of detectors, wherein the radioactive radiation source (7) lies in a radiation path (Si, S2) between the detectors of these pairs (15.2, 15.2), and wherein each pair (15.2, 15.2) has a first detector (15.2) to be calibrated and a second coincident detector (15.2), and wherein the data processing device (9) is configured to carry out a virtual collimation, which preferably comprises - to determine for each pair approximate gamma interaction positions (PI A) in the coincident detectors (15.2), - to generate one or more virtual line paths (Li, L2) starting from the respective approximate interaction positions (PI A) of the coincident detectors (15.2) passing through the reference position (R) of the radiation source, and / or - to determine interaction positions (Pi) for the detectors (15.1) to be calibrated based on the generated line paths (Li, L2) in at least one dimension, wherein the at least one dimension includes an interaction depth (PI T).

23. Device according to claim 22, characterized in that the device (1), in particular the data processing device (9), is configured to carry out one, several, or all steps of the method according to one of claims 2 to 21.

Images