Improvement of the attenuation map generated in the LSO background

Abstract

Various systems and computer-implemented methods for background radiation-based attenuation correction are disclosed. Nuclear scan data, including scan data associated with a first imaging modality and background radiation data, are received. An initial background radiation attenuation map is generated and provided to a trained model configured to generate a final background radiation-based attenuation map from the initial background radiation attenuation map. Attenuation correction is performed on the scan data associated with the first imaging modality based on the background radiation-based attenuation map, and a nuclear image is reconstructed from the attenuation-corrected scan data associated with the first imaging modality.

Description

Technical Field

[0001] This application generally relates to attenuation correction in nuclear imaging diagnosis, and more particularly to attenuation correction in nuclear imaging diagnosis obtained using lutetium oxy orthosilicate (LSO) or lutetium yttrium oxyorthosilicate (LYSO) scintillation crystals.

Background Art

[0002] During nuclear medicine imaging, a patient is positioned on a table and data is acquired using one or more scanning modalities such as computed tomography (CT), positron emission tomography (PET), single photon emission computed tomography (SPECT), magnetic resonance (MR), etc. Multiple data sets can be collected for one patient. Various types of image reconstruction methods have been developed to control and / or eliminate artifacts. However, although different parameters are used in each reconstruction method, the underlying patient is the same in each reconstruction method.

[0003] Attenuation correction is performed to provide a quantitatively accurate distribution of radioactive isotopes from various imaging methods. For example, a PET-CT scanner acquires attenuation map information by performing an additional CT scan and uses this to perform attenuation correction. However, this additional acquisition increases the radiation dose to the patient. Current PET-MR scanners utilize an attenuation map derived from a Dixon sequence. The map derived from Dixon is inaccurate.

Summary of the Invention

Problems to be Solved by the Invention

[0004] In various embodiments, a computer-implemented method for attenuation correction in PET is disclosed. Nuclear scan data including scan data (scanning data) related to a first imaging modality and background radiation data is acquired. An initial background radiation attenuation map is generated and provided to a trained model configured to generate a final background radiation-based attenuation map from the initial background radiation attenuation map. Attenuation correction of the scan data related to the first imaging modality is performed based in part on the background radiation-based attenuation map. An image is reconstructed from the attenuation-corrected scan data related to the first imaging modality.

[0005] In various embodiments, a system is disclosed that includes a first imaging modality configured to generate a first set of scan data and a plurality of detectors configured to generate background radiation data. The system further includes a non-transitory memory having instructions stored therein and a processing device configured to read the instructions for acquiring the first set of scan data and the background radiation data. An initial background radiation attenuation map is generated and provided to a trained model configured to generate a final background radiation-based attenuation map from the initial background radiation attenuation map. Attenuation correction of the first set of scan data is performed based in part on the final background radiation-based attenuation map, and an image is reconstructed from the attenuation-corrected first set of scan data.

[0006] In various embodiments, a computer-implemented method is disclosed for training a model to generate a background radiation-based attenuation map. A set of training data is obtained that includes one or more initial background radiation attenuation maps and one or more ground truth attenuation maps. Each of the one or more ground truth attenuation maps is associated with one of the one or more initial background radiation attenuation maps. Based on this set of training data, an untrained model is iteratively trained to output a trained model configured to generate a background radiation-based attenuation map. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] The following will become apparent from the elements of the figures provided for purposes of illustration and not necessarily drawn to scale.

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

DETAILED DESCRIPTION OF THE INVENTION

[0008] This description of the exemplary embodiments is intended to be read in connection with the accompanying drawings, which are considered to be a part of the entire written description.

[0009] In the following, various embodiments are also described with respect to the system recited in the claims, in the same manner as the method recited in the claims. Features, advantages, or alternative embodiments herein can be assigned to the subject matter recited in other claims, and vice versa. In other words, the claims regarding the system can be described in relation to, or improved by, the features described or claimed in this method. In this case, the functional features of this method are implemented by the target units of the system.

[0010] Furthermore, the following describes various embodiments regarding a method for performing attenuation correction using an attenuation map obtained from LSO (lutetium oxyorthosilicate)-based or LYSO (lutetium yttrium oxyorthosilicate)-based background radiation data, and a system, as well as a method for training a neural network for generating an attenuation map from LSO / LYSO background radiation data, and a system. Features, advantages, or alternative embodiments herein can be assigned to the subject matter recited in other claims, and vice versa. In other words, claims for a method and system for training a neural network for generating an attenuation map using LSO / LYSO background radiation data are described in relation to, or can be improved by, the features claimed for a method and system for performing attenuation correction using LSO / LYSO background radiation data, and vice versa.

[0011] Generally, a learned function mimics the cognitive function by which a human associates the mind of another human. In particular, through learning based on training data, the learned function can adapt to new situations and detect and estimate patterns.

[0012] Generally, through learning, the parameters of a learned function can be adapted. In particular, combinations of supervised learning, semi-supervised learning, unsupervised learning, reinforcement learning, and / or active learning can be used. Furthermore, representation learning (also referred to as "feature learning" alternatively) can be used. In particular, the parameters of the learned function can be adaptively repeated through several steps of training.

[0013] In particular, the learned function can include a neural network, a support vector machine, a decision tree, and / or a Bayesian network, and / or the learned function can be based on k-means clustering, Q-learning, genetic algorithms, and / or association rules (correlation rules). In particular, the neural network can be a deep neural network, a convolutional neural network, or a convolutional deep neural network. Further, the neural network can be an adversarial network, a deep adversarial network, and / or an adversarial generative network.

[0014] FIG. 1 illustrates one embodiment of a nuclear imaging system 2 according to some embodiments. The nuclear imaging system 2 includes a scanner for at least a first modality 12 provided within a first gantry 16a. The first modality 12 can include any suitable imaging modality such as a positron emission tomography (PET) modality. The patient 17 is lying horizontally on a patient bed 18 that is movable between the gantries. In some embodiments, the nuclear imaging system 2 includes a scanner for a second imaging modality 14 provided within a second gantry 16b. The second imaging modality 14 can be any suitable imaging modality, such as, for example, a PET modality, a SPECT modality, a CT modality, a magnetic resonance (MR) modality, and / or any other suitable imaging modality. Each of the first modality 12 and / or the second modality 14 can include one or more detectors 50 configured to detect annihilation photons, gamma rays, and / or other nuclear imaging events. In some embodiments, one or more of the detectors 50 generate background radiation data during the scan.

[0015] Scan data from the first modality 12 and / or the second modality 14 is stored in one or more computer databases 40 and processed by one or more computer processing devices 60 of the computer system 30. The graphical depiction of the computer system 30 in FIG. 1 is provided for illustrative purposes only, and the computer system 30 can include one or more separate computing devices. The nuclear image data set can be provided by the first modality 12, the second modality 14, and / or, for example, from a memory connected to the computer system 30 as a separate data set. The computer system 30 can include one or more processing electronics for processing signals received from one of the plurality of detectors 50. In some embodiments, the scan data includes background radiation-based attenuation. The computer system 30 can use one or more background radiation-based attenuation maps during image reconstruction to correct for background radiation attenuation.

[0016] In some embodiments, the computer system 30 is configured to generate at least one initial background radiation-based attenuation map for use in image reconstruction of data obtained by the first modality 12 and / or the second modality 14. The background radiation-based attenuation map can be generated using any suitable algorithm, any suitable parameters such as noise values, event counts, etc. The attenuation map can be generated and / or refined by a learned neural network (or function). In an embodiment, the initial background radiation-based attenuation map is generated using the maximum likelihood transmission (MLTR) algorithm, although of course other algorithms can be applied to generate the initial background radiation-based attenuation map.

[0017] FIG. 2 shows an embodiment of an artificial neural network 100. Alternative terms for "artificial neural network" are "neural network", "artificial neural net", "neural net", or "trained function". The artificial neural network 100 includes nodes 120-132 and edges 140-142, where each edge 140-142 is a connection leading from a first node 120-132 to a second node 120-132. Generally, the first node 120-132 and the second node 120-132 are different nodes 120-132, but the first node 120-132 and the second node 120-132 may be the same. For example, in FIG. 2, edge 140 is a directed connection from node 120 to node 123, and edge 142 is a directed connection from node 130 to node 132. The edges 140-142 leading from the first node 120-132 to the second node 120-132 can be represented as "input edges" for the second node 120-132 and "output edges" for the first node 120-132.

[0018] In this embodiment, the nodes 120-132 of the artificial neural network 100 can be arranged in layers 110-113, and the layers can include a unique order introduced by the edges 140-142 between the nodes 120-132. In particular, the edges 140-142 can exist only between adjacent layers of nodes. In the illustrated embodiment, there is an input layer 110 that has no input edges and includes only nodes 120-122, an output layer 113 that has no output edges and includes only nodes 131, 132, and hidden layers 111, 112 between the input layer 110 and the output layer 113. Generally, the number of hidden layers 111, 112 can be arbitrarily selected. The number of nodes 120-122 in the input layer 110 is usually related to the number of input values of the neural network, and the number of nodes 131, 132 in the output layer 113 is usually related to the number of output values of the neural network.

[0019] In particular, a number (real number) can be assigned as a value to all the nodes 120 to 132 of the neural network 100. Here, x (n) i represents the numerical value of the i-th node 120 to 132 in the n-th layer 110 to 113. The values of the nodes 120 to 122 in the input layer 110 correspond to the input values of the neural network 100, and the values of the nodes 131, 132 in the output layer 113 correspond to the output values of the neural network 100. Further, each edge 140 to 142 can include a weight that is a real number. In particular, the weight is a real number within the interval [-1, 1] or within the interval [0, 1]. Here, w (m,n) i,j represents the weight of the edge between the i-th node 120 to 132 in the m-th layer 110 to 113 and the j-th node 120 to 132 in the n-th layer 110 to 113. Further, for the weight w (n,n+1) i,j an abbreviated form w (n) i,j is defined.

[0020] In particular, in order to calculate the output value of the neural network 100, the input value propagates through the neural network. In particular, the value of the node 120 to 132 in the (n + 1)-th layer 110 to 113 can be calculated by the following formula based on the value of the node 120 to 132 in the n-th layer 110 to 113.

Equation

[0021] In particular, the values are propagated layer by layer through the neural network. Here, the values of the input layer 110 are given by the input to the neural network 100, the values of the first hidden layer 111 are calculated based on the values of the input layer 110 of the neural network, the values of the second hidden layer 112 can be calculated based on the values of the first hidden layer 111, and so on.

[0022] Value w regarding the edge (m,n) i,j To set the value w, the neural network 100 needs to be trained using training data. In particular, the training data (denoted as t i as shown) includes training input data and training output data. For the training step, the training input data is applied to the neural network 100 to generate the calculated output data. In particular, the training data and the calculated output data consist of a number of values, and that number is equal to the number of nodes in the output layer.

[0023] In particular, to recursively adapt the weights within the neural network 100, a comparison between the calculated output data and the training data is used (error backpropagation algorithm). In particular, the weights are changed according to the following formula.

Equation

Equation

Equation

[0024] In some embodiments, the neural network 100 is configured or trained to generate a background radiation-based attenuation map. For example, in some embodiments, the neural network 100 is configured to receive background radiation data collected by one or more detectors during a scan of a first patient. The neural network 100 can receive the background radiation data in any suitable form, such as, for example, list mode or sinogram data, raw data, etc. The neural network 100 is trained to generate an attenuation map (e.g., a mu map).

[0025] FIG. 3 is a flowchart 200 showing a method of attenuation correction using LSO / LYSO background radiation data according to some embodiments. FIG. 4 is a process flow 250 for performing attenuation correction using LSO / LYSO background radiation data according to the method shown in FIG. 3 according to some embodiments. At step 202, a first set of scan data 252 and a set of background radiation data 254 are acquired. The first set of scan data 252 is associated with a first imaging modality. The background radiation data 254 is associated with a PET imaging modality. The background radiation data can include LSO (lutetium oxyorthosilicate)-based or LYSO (lutetium yttrium oxyorthosilicate)-based background radiation data. In some embodiments, a second set of scan data (not shown) associated with a second imaging modality is also acquired. Although specific embodiments are discussed herein, it can be understood that the disclosed systems and methods are applicable to any scan data and / or scan modality that includes background radiation.

[0026] In step 204, an initial background radiation attenuation map 264 is generated from the LSO / LYSO background radiation data 254 by the background attenuation map generation process 262. The initial background radiation attenuation map 264 can be generated using any suitable generation process or algorithm, such as, for example, the MLTR process. In step 206, the initial background radiation attenuation map 264 is provided to the trained attenuation model 260 configured to generate a final background radiation-based attenuation map 266. The trained model 260 includes a machine learning model trained using a training dataset, as discussed in more detail below. In some embodiments, the trained attenuation model 260 includes a neural network. The trained attenuation model 260 improves and / or enhances the initial background radiation attenuation map 264 to generate the final background radiation-based attenuation map 266. The final background radiation-based attenuation map 266 is used to correct for attenuation in the first set of scan data 252. The trained model 260 can include one or more iterative processes for generating the final background radiation-based attenuation map 266, including, but not limited to, applying one or more conventional mu-map generation algorithms. As discussed in more detail below, the trained attenuation model 260 can be trained using CT scan data and / or long scan LSO / LYSO data.

[0027] In step 208, attenuation correction is applied to the first set of scan data 252, and in step 210, one or more clinical images are generated from the first set of attenuated corrected scan data 252. Although steps 208 and 210 are illustrated as separate steps, it is understood that these steps can be performed as part of a single image reconstruction process 268. The attenuation correction is performed by the image reconstruction process 268 using any suitable attenuation correction process and at least partially based on the final background radiation-based attenuation map 266.

[0028] Clinical image 270 can include, for example, diagnostic images, planning images, and / or any other suitable clinical images. The clinical image 270 can be stored on a non-transitory medium and / or provided to a clinician for use in diagnosis, planning, and / or other purposes. One or more clinical images 270 can be stored as an image file, as attenuation correction data, and / or using any other suitable storage method. In some embodiments, the first set of scan data is a PET data set, but it will be understood that the attenuation correction can also be applied to a second set of scan data including other imaging modalities such as SPECT. As will be discussed in more detail below, the learned attenuation model 260 can be trained using CT scan data and / or long scan LYSO data.

[0029] The method of image reconstruction using the background radiation attenuation map discussed in connection with FIG. 3 provides distinct advantages over current systems. For example, current systems rely primarily on CT scans for the generation of attenuation maps. The use of attenuation correction maps created from LSO / LYSO background radiation enables the use of imaging systems without CT components, reducing the cost of the system, the operating costs (i.e., not requiring a certain level of radioactive tracer), and the radiation exposure of the patient. Systems without CT components can be made more compact and thus can be incorporated, for example, in spaces that cannot currently support PET / CT systems. In systems that include CT components, the LSO / LYSO background radiation attenuation map can be more accurate when MLAA is used to generate the attenuation map with the attenuation map from the background LSO / LYSO as input. The attenuation map output from MLAA better matches the emission data, thereby reducing motion artifacts. While specific advantages are considered here, it will be recognized that the method of attenuation correction using LSO / LYSO background radiation data discussed here provides additional advantages beyond those discussed.

[0030] FIG. 5 is a flowchart 200a illustrating a method of image reconstruction including attenuation correction using a background radiation-based attenuation map, according to some embodiments. FIG. 6 is a process flow 250a for performing image reconstruction including attenuation correction using a background radiation attenuation map according to the method shown in FIG. 5, according to some embodiments. Flowchart 200a and process flow 250a are similar to flowchart 200 and process flow 250 described above, and the same description will not be repeated here. In some embodiments, after generating an initial background radiation attenuation map 264 at step 204, a maximum likelihood estimation of activity and attenuation (MLAA) process 280 is used to generate an initial background radiation-containing attenuation map 282. For example, in some embodiments, a first set of scan data 252 and an initial background radiation attenuation map 264 are provided to an MLAA process 270 configured to generate an initial background radiation-containing attenuation map 282. In some embodiments, the MLAA process 270 further generates activity information 284 that can be used in reconstruction and / or other processes.

[0031] The initial background radiation-containing attenuation map 282 is inferior in SNR (signal-to-noise ratio). The initial background radiation-containing attenuation map 282 can be provided to a trained attenuation model 260a. This trained attenuation model 260a is configured to improve the SNR of the initial background radiation-containing attenuation map 282 and generate a final background radiation-based attenuation map 266 suitable for use in attenuation correction of scan data 252 generated by the first imaging modality. As will be discussed in more detail below, the learned attenuation model 260a can be learned using CT scan data and / or long scan LYSO data.

[0032] FIG. 7 is a flowchart 300 showing a method of training a machine learning model for generating a final background radiation based attenuation map according to some embodiments. FIG. 8 is a process flow 350 for training a machine learning model according to the method shown in FIG. 7 according to some embodiments. At step 302, a set of training data 352 is received. The set of training data includes labeled data configured to repeatedly train an untrained machine learning model 358 to generate a final background radiation based attenuation map. The set of training data 352 can include a set of initial background radiation attenuation maps 354, a set of MLAA generated background radiation containing attenuation maps 356, and / or a set of associated ground truth attenuation maps 358. The set of ground truth attenuation maps 358 can be generated by mapping LSO / LYSO background radiation data to image data from a second imaging modality such as, for example, a CT imaging modality, or based on long scan LSO / LYSO background radiation data, or using any other suitable attenuation map generation process, and / or combinations thereof.

[0033] In some embodiments, the set of training data 352 can include raw background radiation data and / or TOF-PET data, and can generate respective initial background radiation attenuation maps 354 and / or MLAA generated background radiation containing attenuation maps 356 from the raw data and provide them to the untrained model 358. For example, in some embodiments, a set of initial background radiation attenuation maps 354 can be generated from LSO / LYSO background radiation data using an MLTR process. As another example, in some embodiments, a set of MLAA generated background radiation containing attenuation maps 356 can be generated from a set of initial background radiation attenuation maps 354 and raw TOF-PET data using an MLAA process.

[0034] In step 304, the set of learning data 352 is provided to the untrained machine learning model 360, and in step 306, the untrained machine learning model 360 executes an iterative training process. In some embodiments, the iterative learning process includes, for example, comparing an initial background radiation attenuation map with a corresponding one of the ground truth attenuation maps 356 and adjusting the untrained machine learning model 360 based on the identified differences to train a set of first embedding (or hidden) layers to refine the initial background radiation attenuation map. In embodiments that include the background radiation-containing attenuation map 356 generated by MLAA, the machine learning model 360 can be iteratively trained by comparing the background radiation-containing attenuation map 356 generated by MLAA with the ground truth attenuation map 358 to refine the background radiation-containing attenuation map generated by MLAA, for example, to increase the SNR. In some embodiments, an intermediate machine learning model 362 is generated and used in subsequent iterative learning steps. The intermediate machine learning model 362 is further refined using the set of learning data 352 to generate a trained machine learning model 260. Although embodiments including the untrained machine learning model 360 are described herein, it can be understood that a pre-trained machine learning model can be used as the initial learning model 360 for use in the iterative learning process.

[0035] In step 308, the trained machine learning model 260 is output. The trained machine learning model 260 is configured to generate a final background radiation-based attenuation map for use in attenuation correction. The trained machine learning model 260 can be used to generate a final background radiation-based attenuation map for attenuation correction of scan data, for example, according to the methods discussed herein, as discussed in connection with FIG. 3.

[0036] The first embodiment includes a computer-implemented method for attenuation correction. The computer-implemented method includes receiving nuclear scan data including scan data related to a first imaging modality and background radiation data, applying a trained model configured to generate a background radiation-based attenuation map from the background radiation data, performing attenuation correction of the scan data related to the first imaging modality based on the background radiation-based attenuation map, and generating an image from the attenuation-corrected scan data related to the first imaging modality.

[0037] The computer-implemented method of the first embodiment may include background radiation data based on LSO (lutetium oxyorthosilicate) or LYSO (lutetium yttrium oxyorthosilicate). Attenuation correction of the scan data related to the first imaging modality can be based on an attenuation map based on a second imaging modality.

[0038] In the first embodiment, the trained model can be trained by mapping the background radiation data to computed tomography (CT) scan data and / or mapping the background radiation data to long scan background radiation data generated using a known radiation source.

[0039] The first embodiment can include generating an initial background radiation-based attenuation map and providing the initial background radiation-based attenuation map to a trained model. The trained model is configured to generate a background radiation-based attenuation map by refining the initial background radiation-based attenuation map to increase the signal-to-noise ratio. The initial background radiation-based attenuation map can be generated using a maximum likelihood estimation of radioactivity and attenuation (MLAA) process.

[0040] In the first embodiment, the first imaging modality may be a positron emission tomography (PET) modality, and the second imaging modality may be a magnetic resonance (MR) modality.

[0041] In the second embodiment, the system includes a first imaging modality configured to generate a first set of scan data, a plurality of detectors configured to generate background radiation data, a non-transitory memory having instructions stored therein, and a processing device configured to read the instructions. The processing device receives the first set of scan data and the background radiation data, applies a learned model configured to generate a background radiation-based attenuation map from the background radiation data, performs attenuation correction on the first set of scan data based on the background radiation-based attenuation map, and generates an image from the attenuated corrected first set of scan data.

[0042] The second embodiment can include a second imaging modality configured to generate a second set of scan data, and the plurality of detectors can be associated with the second imaging modality.

[0043] The detectors of the second embodiment can include LSO (lutetium oxyorthosilicate)-based or LYSO (lutetium yttrium oxyorthosilicate)-based detectors.

[0044] In the second embodiment, the learned model can be trained by mapping the background radiation data to computed tomography (CT) scan data and / or by mapping the background radiation data to long scan background radiation data generated using a known radiation source.

[0045] The processing device can further be configured to generate an initial background radiation attenuation map and provide the initial background radiation attenuation map to the trained model. The trained model is configured to generate a background radiation-based attenuation map by refining the initial background radiation-based attenuation map to increase the signal-to-noise ratio. The initial background radiation-based attenuation map can be generated using a maximum likelihood estimation of radioactivity and attenuation (MLAA) process.

[0046] In a third embodiment, a computer-implemented method is included for training a model for generating a background radiation-based attenuation map. The method includes receiving a set of training data including one or more subsets of background radiation data and one or more attenuation maps each associated with one of the one or more subsets of the background radiation data, repeatedly training an untrained model based on the set of training data, and outputting a trained model configured to generate a background radiation-based attenuation map.

[0047] In a third embodiment, the set of training data can include MLAA-based attenuation maps associated with each subset of the background radiation data. In such a case, the set of training data can include time-of-flight (TOF) positron emission tomography (PET) (TOF-PET) data. The MLAA-based attenuation maps associated with each subset of the background radiation data can be generated using the background radiation data and the TOF-PET data.

[0048] In a third embodiment, each of the one or more attenuation maps can be generated based on long scan background radiation data generated using a known radiation source and / or based on computed tomography (CT) scan data.

[0049] In the fourth embodiment, the learned model used in either the first or second embodiment can be generated by the computer implementation method of the third embodiment.

[0050] In the fifth embodiment, a non-transitory computer-readable medium contains instructions that, when executed by a processing device, cause the processing device to execute the method of the first, third, or fourth embodiment.

Claims

**Claim 1** A computer-implemented method for attenuation correction, comprising: obtaining nuclear scan data including scan data related to a first imaging modality and background radiation data; generating an initial background radiation attenuation map from the background radiation data by a maximum likelihood transmission (MLTR) process; applying a trained model configured to generate a final background radiation-based attenuation map from the initial background radiation attenuation map; performing attenuation correction of the scan data related to the first imaging modality based on the final background radiation-based attenuation map; and reconstructing an image from the attenuation-corrected scan data related to the first imaging modality, wherein the background radiation data is LSO (lutetium oxyorthosilicate) or LYSO (lutetium yttrium oxyorthosilicate) background radiation data. A method. **Claim 2** The attenuation correction of the scan data related to the first imaging modality is further based on an attenuation map from a second imaging modality. The method according to claim 1. **Claim 3** The trained model is trained by mapping one or more initial background radiation attenuation maps to a computed tomography (CT) attenuation map. The method according to claim 1. **Claim 4** The trained model is trained by mapping one or more initial background radiation attenuation maps to a long scan background radiation-based attenuation map generated using a known radiation source. The method according to claim 1. **Claim 5** A computer-implemented method for attenuation correction, comprising: obtaining nuclear scan data including scan data related to a first imaging modality and background radiation data; generating an initial background radiation attenuation map from the background radiation data by a maximum likelihood transmission (MLTR) process; applying a trained model configured to generate a final background radiation-based attenuation map from the initial background radiation attenuation map; Performing attenuation correction of the scan data associated with the first imaging modality based on the final background radiation-based attenuation map; Reconstructing an image from the attenuation-corrected scan data associated with the first imaging modality; and Generating a background radiation-containing attenuation map using a maximum likelihood estimation of radioactivity and attenuation (MLAA) process, comprising The background radiation data is LSO (lutetium oxyorthosilicate) or LYSO (lutetium yttrium oxyorthosilicate) background radiation data, The initial background radiation attenuation map and the scan data associated with the first imaging modality are provided as inputs to the MLAA process, The learned model is configured to generate the final background radiation-based attenuation map from the initial background radiation-containing attenuation map, Method. [

6. ] The first imaging modality is a positron emission tomography (PET) method, The method according to claim 1 or 5. [

7. ] A system comprising: A first imaging modality configured to generate a first set of scan data; A plurality of detectors configured to generate background radiation data; A non-transitory memory storing instructions; The following instructions: Obtaining the first set of scan data and background radiation data; Generating an initial background radiation attenuation map from the background radiation data by a maximum likelihood transmission (MLTR) process; Applying a learned model configured to generate a final background radiation-based attenuation map from the initial background radiation attenuation map; Performing attenuation correction of the first set of scan data based on the final background radiation-based attenuation map; Reconstructing an image from the attenuation correction of the first set of scan data; A processing device configured to read; comprising The detector is an LSO (lutetium oxyorthosilicate) or LYSO (lutetium yttrium oxyorthosilicate) detector, System. [

8. ] A system comprising: A first imaging modality configured to generate a first set of scan data; A plurality of detectors configured to generate background radiation data; A non-transitory memory storing instructions; Configured to generate a second set of scan data; The following instructions: Obtain the first set of scan data and background radiation data; Generate an initial background radiation-based attenuation map from the background radiation data by a maximum likelihood transmission (MLTR) process; Read the instructions for generating at least one additional attenuation map from the second set of scan data; Apply a trained model configured to generate a final background radiation-based attenuation map from the initial background radiation-based attenuation map; Perform attenuation correction on the first set of scan data based on the final background radiation-based attenuation map and the at least one additional attenuation map; Reconstruct an image from the attenuation correction of the first set of scan data; A processing device configured to read; Comprising; The detector is an LSO (lutetium oxyorthosilicate) or LYSO (lutetium yttrium oxyorthosilicate) detector, System.

9. The trained model is trained by mapping one or more initial background radiation attenuation maps to attenuation maps from computed tomography (CT). The system according to claim 7.

10. The trained model is trained by mapping one or more initial background radiation attenuation maps to long scan background radiation-based attenuation maps generated using a known radiation source. The system according to claim 7.

11. The processing device is configured to read the instructions for generating a background radiation-containing attenuation map using a maximum likelihood estimation of radioactivity and attenuation (MLAA) process, The initial background radiation attenuation map and the scan data related to the first imaging modality are provided as inputs to the MLAA process, The trained model is configured to generate the final background radiation-based attenuation map from the background radiation-containing attenuation map. The system according to claim 7.

Images