Methods and apparatus for attenuation correction of medical images for image reconstruction

Abstract

Systems and methods for generating synthetic attenuation correction maps (e.g., µ-maps), and for reconstructing images based on emission data and the synthetically generated attenuation correction maps, are disclosed. In some examples, a nuclear imaging reconstruction system receives emission data, such as positron emission tomography data, from an image scanning system. The nuclear imaging reconstruction system applies a segmentation process to the emission data and, based on the segmentation process, generates segmentation data characterizing a plurality of segments. Further, the nuclear imaging reconstruction system generates attenuation correction data that includes attenuation correction values for each of the plurality of segments. For example, based on the type of a particular segment (e.g., lung, heart, liver segments), the nuclear imaging reconstruction system assigns corresponding attenuation correction values. The nuclear imaging reconstruction system may reconstruct an image based on the emission data and the attenuation correction data.

Description

METHODS AND APPARATUS FOR ATTENUATION CORRECTION OF MEDICAL IMAGES FOR IMAGE RECONSTRUCTIONFIELD

[0001] Aspects of the present disclosure relate in general to medical diagnostic systems and, more particularly, to attenuation correction of images for image reconstruction in nuclear imaging systems.BACKGROUND

[0002] Nuclear imaging systems can employ various technologies to capture images. For example, some nuclear imaging systems employ positron emission tomography (PET) to capture images. PET is a nuclear medicine imaging technique that produces tomographic images representing the distribution of positron emitting isotopes within a body. Some nuclear imaging systems combine images from PET and a co-modality, such as computed tomography (CT) or Magnetic Resonance Imaging (MRI). CT is an imaging technique that uses x-rays to produce anatomical images. Magnetic Resonance Imaging (MRI) is an imaging technique that uses magnetic fields and radio waves to generate anatomical and functional images, and may also be used as a co-modality. These nuclear imaging systems can combine images from PET and comodality scanners during an image fusion process to produce images that show information from both the PET scan and the co-modality scan (e.g, PET / CT systems).

[0003] Moreover, the nuclear imaging systems may generate an attenuation map that can be used to correct the PET measurement data during image reconstruction. For example, PET images may be corrected for attenuation to provide more accurate reconstructed images. In some PET / CT imaging systems, a measured CT image volume is directly converted to attenuation coefficient (e.g, p-map) values, and a corresponding PET image is corrected based on the attenuation coefficient values. In some PET / MR imaging systems, a segmentation-based p-map is generated from a multi-point MR Dixon sequence, and a corresponding PET image is corrected based on the p-map.

[0004] These and other methodologies have several drawbacks. For instance, the generated p-maps can be problematic due to tissue misclassification, truncation, and incomplete or incorrect correction values, thereby resulting in less accurate reconstructed images.Moreover, these methodologies require a patient to be scanned with the co-modality merely to obtain the attenuation correction values. For these and other reasons, there are opportunities to address deficiencies in nuclear imaging systems.SUMMARY

[0005] Systems and methods for generating synthetic attenuation correction maps (e.g., p-maps), and for reconstructing images based on emission data (e.g., positron emission tomography (PET) data, single-photon emission computed tomography (SPECT) data) and the synthetically generated attenuation correction maps, are disclosed.

[0006] In some embodiments, a computer-implemented method includes receiving emission data from an image scanning system. The method also includes applying a segmentation process to the emission data and, based on the segmentation process, generating segmentation data characterizing a plurality of segments. Further, the method includes generating attenuation correction data comprising at least one attenuation correction value for each of the plurality of segments. The method also includes reconstructing an image based on the emission data and the attenuation correction data.

[0007] In some embodiments, a system includes a memory device storing instructions, and at least one processor communicatively coupled the memory device. The at least one processor is configured to execute the instructions to receive emission data from an image scanning system. The at least one processor is also configured to execute the instructions to apply a segmentation process to the emission data and, based on the segmentation process, generate segmentation data characterizing a plurality of segments. Further, the at least one processor is configured to execute the instructions to generate attenuation correction data comprising at least one attenuation correction value for each of the plurality of segments. The at least one processor is also configured to execute the instructions to reconstruct an image based on the emission data and the attenuation correction data.

[0008] In some embodiments, a non-transitory computer readable medium stores instructions that, when executed by at least one processor, cause the at least one processor to perform operations including receiving emission data from an image scanning system. The operations also include applying a segmentation process to the emission data and, based on thesegmentation process, generating segmentation data characterizing a plurality of segments. Further, the operations include generating attenuation correction data comprising at least one attenuation correction value for each of the plurality of segments. The operations also include reconstructing an image based on the emission data and the attenuation correction data.BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The following will be apparent from elements of the figures, which are provided for illustrative purposes and are not necessarily drawn to scale.

[0010] FIG. 1 illustrates a nuclear imaging system, in accordance with some embodiments.

[0011] FIG. 2 illustrates a block diagram of an example computing device that can perform one or more of the functions described herein, in accordance with some embodiments.

[0012] FIG. 3 illustrates the generation of segmentation data using a trained neural network, in accordance with some embodiments.

[0013] FIG. 4A illustrates a nuclear image, in accordance with some embodiments.

[0014] FIG. 4B illustrates a segmented image, in accordance with some embodiments.

[0015] FIG. 4C illustrates an attenuation corrected image, in accordance with some embodiments.

[0016] FIG. 5 illustrates a flowchart of an exemplary method to reconstruct an image, in accordance with some embodiments.

[0017] FIG. 6 illustrates a flowchart of an exemplary method to train a neural network, in accordance with some embodimentsDETAILED DESCRIPTION

[0018] This description of the exemplary embodiments is intended to be read in connection with the accompanying drawings, which are to be considered part of the entirewritten description. Independent of the grammatical term usage, individuals with male, female, or other gender identities are included within the term.

[0019] The exemplary embodiments are described with respect to the claimed systems as well as with respect to the claimed methods. Furthermore, the exemplary embodiments are described with respect to methods and systems for image reconstruction, as well as with respect to methods and systems for training functions used for image reconstruction. Features, advantages, or alternative embodiments herein can be assigned to the other claimed objects and vice versa. For example, claims for the providing systems can be improved with features described or claimed in the context of the methods, and vice versa. In addition, the functional features of described or claimed methods are embodied by objective units of a providing system. Similarly, claims for methods and systems for training image reconstruction functions can be improved with features described or claimed in context of the methods and systems for image reconstruction, and vice versa.

[0020] Various embodiments of the present disclosure can employ machine learning methods or processes to provide clinical information from nuclear imaging systems. For example, the embodiments can employ machine learning methods or processes to reconstruct images based on captured measurement data, and provide the reconstructed images for clinical diagnosis. In some embodiments, machine learning methods or processes are trained, to improve the reconstruction of images.

[0021] The embodiments described herein can generate synthetic attenuation correction maps (e.g., p-maps) based on segmentation processes applied to emission data (e.g., PET data), and can further reconstruct images based on the emission data and the synthetically generated attenuation correction maps. For instance, the processes described herein may generate semantic information characterizing tissue classifications, such as organ segmentations generated by a trained neural network (e.g., a trained convolutional neural network (CNN), and may assign attenuation correction values (e.g., p-map values) to the various segmented regions to generate a synthetic attenuation correction map (e.g., p-map). The attenuation correction values assigned to the various segmented regions may be generated based on a-priori information (e.g., average historical values for each region), by a second neural network, or by any other suitable method asdescribed herein. A reconstruction process may then correct the emission data based on the generated synthetic attenuation correction map to generate a reconstructed image.

[0022] Referring now to the figures, FIG. 1 illustrates a nuclear imaging system 100 that includes an image scanning system 102 and an image reconstruction system 104. Nuclear imaging system 100 may be, for example, a PET nuclear imaging system. In some examples, nuclear imaging system 100 can capture PET images as well as images from a co-modality, such as computed tomography (CT) or Magnetic Resonance Imaging (MRI), and can produce images that show information from both PET scans and co-modality scans. For instance, nuclear imaging system 100 may be a PET / CT nuclear imaging system or PET / MR nuclear imaging system. The nuclear imaging system 100 also includes a data repository 120 that can be accessed by the image reconstruction system 104 and, in some examples, the image scanning system 102.

[0023] As illustrated, image reconstruction system 104 includes a segmentation engine 114, an attenuation map generation engine 116, and an image volume reconstruction engine 118. In some examples, all or parts of image reconstruction system 104 are implemented in hardware, such as in one or more Field-Programmable Gate Arrays (FPGAs), one or more System-on- Chips (SoCs), one or more application-specific integrated circuits (ASICs), one or more state machines, one or more computing devices, digital circuitry, or any other suitable circuitry. For example, all or parts of segmentation engine 114, attenuation map generation engine 116, and image volume reconstruction engine 118 may be implemented within one or more FPGAs and / or SoCs. In some examples, at least parts of image reconstruction system 104 can be implemented in software as executable instructions such that, when executed by one or more processors, cause the one or more processors to perform respective functions as described herein. The instructions can be stored in a non-transitory, computer-readable storage medium, and can be read and executed by the one or more processors. For instance, in some examples, one or more portions of any of segmentation engine 114, attenuation map generation engine 116, and image volume reconstruction engine 118 can be implemented by one or more processors executing instructions. The one or more processors may include, for instance, a microcontroller, a graphics processing unit (GPU), a central processing unit (CPU), a digital signal processor (DSP), a soft core, or any other suitable processing device.

[0024] Referring back to FIG. 1, segmentation engine 114 may receive PET measurement data 111 (e.g., an uncorrected PET image) from image scanning system 102. The PET measurement data 111 may characterize, for example, detected energy levels (e.g., energy depositions), first dimension positions (e.g., X-axis position), and second dimension positions (e.g., Y-axis position) of a scanner’s detecting crystals for a corresponding PET scan, as well as corresponding detection times. The segmentation engine 114 may apply a segmentation process (e.g., a trained segmentation process) to the PET measurement data 111 and, based on the segmentation processes, may generate segmentation data 115 characterizing one or more anatomical regions (e.g., tissues and organs) of the PET measurement data 111. For instance, the segmentation data 115 may identify organ segments, such as lung segments, a heart segment, and a kidney segment of the PET measurement data 111. Each of the identified anatomical regions may be referred to as a “segment” herein. The segmentation process may include, for example, application of a clustering algorithm (e.g., a K-means clustering algorithm), or a trained neural network (e.g., a CNN)), to the PET measurement data 111 to generate the segmentation data 115. In some instances, the segmentation process is configured to identify one or more predefined segments (e.g., predefined region classes).

[0025] Further, attenuation map generation engine 116 may receive the segmentation data 115, and may generate an attenuation correction map 117 (e.g., a p-map) that includes at least one attenuation correction value for each of the segments identified by the segmentation data 115. For example, the attenuation map generation engine 116 may associate attenuation correction values (e.g., attenuation correction coefficients) with each of the segments, and may generate the generate the attenuation correction map 117 characterizing the associated attenuation correction values. For instance, the attenuation map generation engine 116 may assign a first attenuation correction value to a first segment, and may generate the attenuation correction map 117 such that the first attenuation correction value is associated with a first portion of the PET measurement data 111 that corresponds to the first segment. Similarly, the attenuation map generation engine 116 may assign a second attenuation correction value to a second segment, and may generate the attenuation correction map 117 such that the second attenuation correction value is associated with a second portion of the PET measurement data 111 that corresponds to the second segment. The attenuation map generation engine 116 may assign attenuation correction values to various segments of the PET measurement data 111 until,for example, all segmented portions of the PET measurement data 111 have been assigned corresponding attenuation correction values.

[0026] In some instances, the attenuation map generation engine 116 may assign a predetermined attenuation correction value to unsegmented portions of the PET measurement data 111. For instance, the attenuation map generation engine 116 may assign an attenuation correction value of 0 to the unsegmented portions of the PET measurement data 111.

[0027] In some examples, the attenuation map generation engine 116 may assign multiple attenuation correction values to a single segment. For instance, the attenuation map generation engine 116 may assign a first attenuation correction value to one portion of the segment (e.g., the top half, bottom half, left side, right side, etc.), and a second attenuation correction value to another portion of the segment. The attenuation map generation engine 116 may generate the attenuation correction map 117 characterizing the associated attenuation correction values as described herein.

[0028] In some examples, the assigned attenuation correction values are based on predefined attenuation correction values for each type of segment. For instance, segment based attenuation correction data 113, stored in data repository 120, may identify one or more attenuation correction values for each type of segment (e.g, for each type of possible segment), where each type of segment corresponds to an anatomical region. For example, segment based attenuation correction data 113 may include a first attenuation correction value for a kidney segment, a second attenuation correction value for a heart segment, and a third attenuation correction value for a lung segment. Attenuation map generation engine 116 may obtain the segment based attenuation correction data 113 from the data repository 120, and may assign the attenuation correction values to the various segmented portions of the PET measurement data 111 identified by the segmentation data 115 based on the corresponding segments, thereby generating the attenuation correction map 117.

[0029] In some instances, the attenuation correction values of the segment based attenuation correction data 113 are based on historical attenuation correction values for each type of anatomical region (e.g., tissue or organ type). In some instance, the attenuation map generation engine 116 may generate average value of historical values based on the historicalattenuation correction values for a particular segment, and may assign the averaged attenuation correction value to the particular segment. In some examples, the segment based attenuation correction data 113 may include attenuation correction values (e.g., average attenuation correction values) based on a patient classification. For example, the segment based attenuation correction data 113 may include average attenuation correction values for each of various age ranges of patients (e.g., 5-10, 10-15, 15-21, 21-30, 30 and up years old), for each sex of a patient (e.g., average values for males, and average values for females), or for any other appropriate classification of patients.

[0030] In other examples, the attenuation map generation engine 116 may apply a trained neural network (e.g., a trained convolutional neural network (CNN)) to the segmentation data 115, where the trained neural network is configured to generate attenuation correction values based on a type of corresponding segment. Based on the application of the trained neural network to the segmentation data 115, the attenuation map generation engine 116 may generate the attenuation correction map 117.

[0031] For example, data repository 120 may store segment based attenuation correction neural network data 123 that characterizes a trained segment based attenuation correction neural network that is configured to generate attenuation correction values based on segmentation data 115. For instance, the segment based attenuation correction neural network data 123 may include parameters (e.g, hyperparameters, weights, coefficients, etc.) that characterize (e.g., define) the trained segment based attenuation correction neural network. The attenuation map generation engine 116 may read the segment based attenuation correction neural network data 123 from data repository 120, and may establish the trained segment based attenuation correction neural network based on the segment based attenuation correction neural network data 123. Further, the attenuation map generation engine 116 may receive the segmentation data 115, and may apply the trained segment based attenuation correction neural network to the segmentation data 115. Based on application of the trained segment based attenuation correction neural network to the segmentation data 115, the attenuation map generation engine 116 may generate the attenuation correction map 117 that includes attenuation correction values for the various segments of the PET measurement data 111.

[0032] Further, image volume reconstruction engine 118 may apply a trained machine learning process to the PET measurement data 111 and the attenuation correction map 117 to reconstruct an attenuation corrected image. For example, image volume reconstruction engine 118 may obtain, from data repository 116, trained neural network data 135 characterizing parameters (e.g., hyperparameters, weights, coefficients, etc.) of a trained neural network 119 (e.g., a trained CNN). Image volume reconstruction engine 118 may establish the trained neural network 119 based on the trained neural network data 135. For instance, image volume reconstruction engine 118 may configure the neural network 119 based on the parameters of the trained neural network data 135. Once established, image volume reconstruction engine 118 may apply the trained neural network 119 to the PET measurement data 111 and the attenuation correction map 117 to generate final image volume 191 characterizing the reconstructed, and attenuation corrected, image. For instance, the established neural network 119 may apply an attenuation correction process (e.g., an attenuation correction algorithm) to the PET measurement data 111 to generate the final image volume 191. The trained neural network 119 can include, for example, any suitable reconstruction algorithm that reconstructs images based on emission data and attenuation maps, for example. For instance, in some examples, the trained neural network 119 can be a filtered back projection (FBP) algorithm (also known as an FBP model), a Maximum Likelihood Expectation Maximization (MLEM) algorithm, an ordered subset expectation maximization (OSEM) algorithm, or any other suitable reconstruction algorithm. The image reconstruction system 104 may store the final image volume 191 in data repository 120.

[0033] FIG. 2 illustrates a computing device 200 that can be employed by the image reconstruction system 104. For instance, computing device 200 can implement one or more of the functions of the segmentation engine 114, the attenuation map generation engine 116, and the image volume reconstruction engine 118 of the image reconstruction system 104 described herein.

[0034] Computing device 200 can include one or more processors 201, working memory 202, one or more input-output devices 203, instruction memory 207, a transceiver 204, one or more communication ports 209, and a display 206, all operatively coupled to one or more databuses 208. Data buses 208 allow for communication among the various devices. Data buses 208 can include wired, or wireless, communication channels.

[0035] Processors 201 can include one or more distinct processors, each having one or more cores. Each of the distinct processors can have the same or different structure. Processors201 can include one or more central processing units (CPUs), one or more graphics processing units (GPUs), application specific integrated circuits (ASICs), digital signal processors (DSPs), and the like.

[0036] Processors 201 can be configured to perform a certain function or operation by executing code, stored on instruction memory 207, embodying the function or operation. For example, processors 201 can be configured to perform one or more of the functions, methods, or operations disclosed herein.

[0037] Instruction memory 207 can store instructions that can be accessed (e.g., read) and executed by processors 201. For example, instruction memory 207 can be a non-transitory, computer-readable storage medium such as a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), flash memory, a removable disk, CD-ROM, any non-volatile memory, or any other suitable memory. For example, instruction memory 207 can store instructions that, when executed by one or more processors 201, cause one or more processors 201 to perform one or more of the functions of image reconstruction system 104, such as one or more of the functions of the segmentation engine 114, the attenuation map generation engine 116, and the image volume reconstruction engine 118.

[0038] Processors 201 can store data to, and read data from, working memory 202. For example, processors 201 can store a working set of instructions to working memory 202, such as instructions loaded from instruction memory 207. Processors 201 can also use working memory202 to store dynamic data created during the operation of computing device 200. Working memory 202 can be a random access memory (RAM) such as a static random access memory (SRAM) or dynamic random access memory (DRAM), or any other suitable memory.

[0039] Input-output devices 203 can include any suitable device that allows for data input or output. For example, input-output devices 203 can include one or more of a keyboard, atouchpad, a mouse, a stylus, a touchscreen, a physical button, a speaker, a microphone, or any other suitable input or output device. For example, a user (e.g., medical professional) may provide input to the computing device 200 using an input-output device 203. Based on the input, the computing device 200 may display an attenuation corrected image, such as final image volume 191, on display 206.

[0040] Communication port(s) 209 can include, for example, a serial port such as a universal asynchronous receiver / transmitter (UART) connection, a Universal Serial Bus (USB) connection, or any other suitable communication port or connection. In some examples, communication port(s) 209 allows for the programming of executable instructions in instruction memory 207. In some examples, communication port(s) 209 allow for the transfer (e.g., uploading or downloading) of data, such as PET measurement data 111.

[0041] Display 206 can display user interface 205. User interface 205 can enable user interaction with computing device 200. For example, user interface 205 can be a user interface for an application that allows for the viewing of final image volumes 191. In some examples, a user can interact with user interface 205 by engaging input-output devices 203. In some examples, display 206 can be a touchscreen, where user interface 205 is displayed on the touchscreen.

[0042] Transceiver 204 allows for communication with a network, such as a Wi-Fi network, an Ethernet network, a cellular network, or any other suitable communication network. For example, if operating in a cellular network, transceiver 204 is configured to allow communications with the cellular network. Processor(s) 201 is operable to receive data from, or send data to, a network via transceiver 204.

[0043] FIG. 3 illustrates an exemplary trained neural network 170 (e.g., a trained CNN) that can be implemented by, for example, the segmentation engine 114 of the image reconstruction system 104. The trained neural network 170 uses spatial relationships within PET distributions to define segmentations. In this example, PET measurement data I l l is inputted into the trained neural network 170. Based on inputting the PET measurement data, the trained neural network generates segmentation data 115 characterizing one or more anatomical regions, i.e., segments, of the PET measurement data 111. For instance, the generated segmentation data115 may include first segment data 115A that characterizes a first anatomical region (e.g., a lung). The generated segmentation data 115 may also include second segment data 115B, third segment data 115C, and fourth segment data 115D. The second segment data 115B may characterize a second anatomical region (e.g., a heart), while the third segment data 115C may characterize a third anatomical region (e.g., a kidney). Additionally, the fourth segment data 115D may characterize a fourth anatomical region (e.g., a liver). Each of the first segment data 115A, second segment data 115B, third segment data 115C, and fourth segment data 115D may include values defining corresponding regions (e.g., image locations) of the PET measurement data.

[0044] FIG. 4A illustrates a nuclear image 400 that includes a front-view (i.e., coronal view) image 402 and side-view (i.e., sagittal view) image 404. For instance, the front-view image 402 may be based on a top view scan of a patient (e.g, by image scanning system 102), and the side-view image 404 may be based on a side view scan of a patient (e.g., by image scanning system 102).

[0045] FIG. 4B illustrates a segmented image 420 that includes a front-view segmentation image 422 and side-view segmentation image 424 based on a segmentation of the front- view image 402 and side-view image 404 of FIG. 4A, respectively. Here, the segmentations were defined in 3D across the entire volume simultaneously. As such, the figures show 2D slices through 3D data. The segmentation of the nuclear image 400 may be based on any of the exemplary segmentation processes described herein. For instance, the front- view segmentation image 422 may be generated based on an application of a segmentation process to the front-view image 402, which may generate segmentation data identifying various anatomical regions such as, for example, a first segment 426 (e.g, a right lung), a second segment 428 (e.g., a liver), a third segment 430 (e.g., hip bone), a fourth segment 432 (e.g., a heart), a fifth segment 434 (e.g., a left lung), and a sixth segment 436 (e.g., a brain). Similarly, the side-view segmentation image 424 may be generated based on an application of the segmentation process to the side-view image 404, which may generate segmentation data identifying various anatomical regions. The segmentation process may include, for example, application of a clustering algorithm (e.g., a K-means clustering algorithm).

[0046] FIG. 4C illustrates an attenuation corrected image 440 that includes a front-view attenuation correction image 442 and a side-view attenuation correction image 444. The frontview attenuation correction image 442 may be generated based on assigned attenuation correction values to the various segments identified in the front-view segmentation image 422 of FIG. 4B. As described herein, the attenuation correction values may be assigned based on historical average attenuation correction values for each type of segment (e.g., each type of anatomical region). In some instances, the attenuation correction values are assigned based on an application of a trained neural network, such as one established based on segment based attenuation correction neural network data 123, to the segmentation data identifying the various anatomical regions of the segmented image 420 of FIG. 4B.

[0047] For example, a first attenuation correction value may be assigned to the first segment 426, while a second attenuation correction value may be assigned to the second segment 428. Similarly, a third attenuation correction value may be assigned to the third segment 430, while a fourth attenuation correction value may be assigned to the fourth segment 432. Further, a fifth attenuation correction value may be assigned to the fifth segment 434, while a sixth attenuation correction value may be assigned to the sixth segment 436. Similarly, the side-view attenuation correction image 444 may be generated based on assigned attenuation correction values to the various segments identified in the side-view segmentation image 424 of FIG. 4B.

[0048] FIG. 5 is a flowchart of an exemplary method 500 to reconstruct an image. The method can be performed by any suitable circuitry that may include hardware and / or software, such as by the computing device 200 or image reconstruction system 104 described herein.

[0049] Beginning at block 502, emission data from an image scanning system is received. For instance, the image reconstruction system 104 may receive PET measurement data 111 from the image scanning system 102. At block 504, segmentation data characterizing a plurality of segments is generated. The segmentation data is generated based on an application of a segmentation process to the emission data. For instance, and as described herein, image reconstruction system 104 may apply a trained K-means segmentation process to the PET measurement data 111 and, in response, generate segmentation data 115 characterizing various anatomical regions of the PET measurement data 111.

[0050] Proceeding to block 506, a determination is made as to whether there are any segments. If there is at least one segment, the method proceeds to block 508, where attenuation correction data is generated the segment. The attenuation correction data includes at least one attenuation correction value based on the type of corresponding segment. For example, and as described herein, the image reconstruction system 104 may assign one or more predetermined attenuation correction values to the segment based on the segment’s type (e.g., lung segment, heart segment, etc.).

[0051] In other examples, the image reconstruction system 104 may apply a trained neural network to the segmentation data, where the trained neural network is configured to generate attenuation correction values based on the type of corresponding segment. For example, the image reconstruction system 104 may establish a trained neural network based on segment based attenuation correction neural network data 123, and may apply the trained neural network to the segmentation data to generate the one or more attenuation correction values for the segment.

[0052] From block 508, the method proceeds back to block 506 to determine if there are any more segments identified by the segmentation data. If there is at least one segment for which attenuation correction values have not yet been determined, the method proceeds back to block 508 to generate the attenuation correction values for that segment. Otherwise, if there are no additional segments to determine attenuation correction values for, the method proceeds to block 510.

[0053] Further, at block 510, an image is reconstructed based on the emission data and the attenuation correction data. For example, the image reconstruction system 140 may apply any suitable reconstruction algorithm to the emission data and the attenuation correction data to generate a corrected image, such as final image volume 191. At block 512, the reconstructed image is stored in a data repository, such as data repository 120.

[0054] FIG. 6 is a flowchart of an exemplary method 600 to train a neural network (e.g., a CNN) to generate attenuation correction values based on segmentation data. The method can be performed by any suitable circuitry that may include hardware and / or software, such as by the computing device 200 or image reconstruction system 104 described herein.

[0055] Beginning at block 602, the computing device 200 inputs labelled segmentation data characterizing segmented emission data into an untrained neural network. The segmentation data may be labeled with, for instance, corresponding attenuation correction values for each of the segments (e.g, during a supervised learning process). For example, the segmentation data may characterize, for each of a plurality of scanned PET images, various anatomical regionssegments). Furthermore, the attenuation correction values are associated with the various types of segments, and may be based on historical averages of attenuation correction values. The computing device 200 may input a predetermined number of epochs of the labelled segmentation data into the untrained neural network.

[0056] At block 604, the computing device 200 inputs unlabeled segmentation data characterizing segmented emission data to the neural network (e.g., during a validation process). Based on the inputted unlabeled segmentation data, the neural network generates output data. At block 606, the computing device 200 may compare the output data to ground truth data and, based on the comparison, generate a metric value. The metric value may be, for instance, a metric value computed from a loss function, such as computed precision values, computed recall values, a computed area-under-curve (AUC) value, any receiver operating characteristic (ROC) curve or precision-recall (PR) curve value, or any other suitable metric value.

[0057] Further, and at block 608, the computing device 200 may determine whether the neural network is trained based on the metric value. For instance, the computing device 200 may compare the metric value to a threshold value. If the metric value exceeds (or, in some examples, is below) the threshold value, the neural network is trained, and the method proceeds to block 610. If, however, the metric value does not exceed (or, in some examples, is not below) the threshold value, then the method proceeds back to block 602 to continue training the neural network.

[0058] At block 610, the computing device 200 stores parameters (e.g., hyperparameters, weights, coefficients, etc.) characterizing the trained neural network in a data repository. For example, the computing device 200 may store the parameters as segment based attenuation correction neural network data 123 in data repository 120.

[0059] The following is a list of non-limiting illustrative embodiments disclosed herein:

[0060] Illustrative Embodiment 1: An apparatus comprising: a memory device storing instructions; and at least one processor communicatively coupled the memory device, wherein the at least one processor is configured to execute the instructions to: receive emission data from an image scanning system; apply a segmentation process to the emission data and, based on the segmentation process, generate segmentation data characterizing a plurality of segments; generate attenuation correction data comprising at least one attenuation correction value for each of the plurality of segments; and reconstruct an image based on the emission data and the attenuation correction data.

[0061] Illustrative Embodiment 2: The apparatus of illustrative embodiment 1, wherein the at least one processor is configured to execute the instructions to determine the at least one attenuation correction value for each of the plurality of segments based on a type of corresponding segment.

[0062] Illustrative Embodiment 3: The apparatus of any of illustrative embodiments 1-2, wherein the plurality of segments comprises a first segment identifying a first anatomical region, and a second segment identifying a second anatomical region.

[0063] Illustrative Embodiment 4: The apparatus of illustrative embodiment 3, wherein the at least one attenuation correction value for each of the plurality of segments comprises a firstattenuation correction value for the first segment, and a second attenuation correction value for the second segment, wherein the first attenuation correction value differs from the second attenuation correction value.

[0064] Illustrative Embodiment 5: The apparatus of any of illustrative embodiments 1-4, wherein the at least one attenuation correction values are based on historical attenuation correction values.

[0065] Illustrative Embodiment 6: The apparatus of any of illustrative embodiments 1-5, wherein the at least one attenuation correction values is an average of the historical attenuation correction values.

[0066] Illustrative Embodiment 7: The apparatus of any of illustrative embodiments 1-6, wherein the at least one processor is configured to execute the instructions to: for each of the plurality of segments: determine a type of segment; and based on the type of segment, receive, from a data repository, the at least one attenuation correction value for the type of segment.

[0067] Illustrative Embodiment 8: The apparatus of any of illustrative embodiments 1-7, wherein the at least one processor is configured to execute the instructions to apply a trained neural network to the segmentation data and, based on the application of the trained neural network to the segmentation data, generate the attenuation correction data.

[0068] Illustrative Embodiment 9: The apparatus of any of illustrative embodiments 1-8, wherein the at least one processor is configured to execute the instructions to apply a trained neural network to the emission data and the attenuation correction data and, based on the application of the trained neural network to the emission data and the attenuation correction data, reconstruct the image.

[0069] Illustrative Embodiment 10: The apparatus of any of illustrative embodiments 1-9, wherein the emission data is positron emission tomography measurement data.

[0070] Illustrative Embodiment 11 : The apparatus of any of illustrative embodiments 1-10, wherein the segmentation process is a K-means clustering process.

[0071] Illustrative Embodiment 12: The apparatus of any of illustrative embodiments 1-11, wherein the at least one processor is configured to execute the instructions to store the image in a data repository.

[0072] Illustrative Embodiment 13: The apparatus of any of illustrative embodiments 1-12, wherein the at least one processor is configured to execute the instructions to provide the image for display.

[0073] Illustrative Embodiment 14: A method by at least one processor, the method comprising:receiving emission data from an image scanning system; applying a segmentation process to the emission data and, based on the segmentation process, generating segmentation data characterizing a plurality of segments; generating attenuation correction data comprising at least one attenuation correction value for each of the plurality of segments; and reconstructing an image based on the emission data and the attenuation correction data.

[0074] Illustrative Embodiment 15: The method of illustrative embodiment 14, comprising determining the at least one attenuation correction value for each of the plurality of segments based on a type of corresponding segment.

[0075] Illustrative Embodiment 16: The method of any of illustrative embodiments 14- 15, wherein the plurality of segments comprises a first segment identifying a first anatomical region, and a second segment identifying a second anatomical region.

[0076] Illustrative Embodiment 17: The method of illustrative embodiment 16, wherein the at least one attenuation correction value for each of the plurality of segments comprises a first attenuation correction value for the first segment, and a second attenuation correction value for the second segment, wherein the first attenuation correction value differs from the second attenuation correction value.

[0077] Illustrative Embodiment 18: The method of any of illustrative embodiments 14-17, wherein the at least one attenuation correction values are based on historical attenuation correction values.

[0078] Illustrative Embodiment 19: The method of any of illustrative embodiments 14-18, wherein the at least one attenuation correction values is an average of the historical attenuation correction values.

[0079] Illustrative Embodiment 20: The method of any of illustrative embodiments 14-19, comprising: for each of the plurality of segments: determining a type of segment; and based on the type of segment, receiving, from a data repository, the at least one attenuation correction value for the type of segment.

[0080] Illustrative Embodiment 21 : The method of any of illustrative embodiments 14-20, comprising applying a trained neural network to the segmentation data and, based on the application of the trained neural network to the segmentation data, generating the attenuation correction data.

[0081] Illustrative Embodiment 22: The method of any of illustrative embodiments 14-21, comprising applying a trained neural network to the emission data and the attenuationcorrection data and, based on the application of the trained neural network to the emission data and the attenuation correction data, reconstructing the image.

[0082] Illustrative Embodiment 23 : The method of any of illustrative embodiments 14-22, wherein the emission data is positron emission tomography measurement data.

[0083] Illustrative Embodiment 24: The method of any of illustrative embodiments 14-23, wherein the segmentation process is a K-means clustering process.

[0084] Illustrative Embodiment 25: The method of any of illustrative embodiments 14-24, comprising storing the image in a data repository.

[0085] Illustrative Embodiment 26: The method of any of illustrative embodiments 14-25, comprising providing the image for display.

[0086] Illustrative Embodiment 27 : A non-transitory computer readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform operations comprising: receiving emission data from an image scanning system; applying a segmentation process to the emission data and, based on the segmentation process, generating segmentation data characterizing a plurality of segments; generating attenuation correction data comprising at least one attenuation correction value for each of the plurality of segments; andreconstructing an image based on the emission data and the attenuation correction data.

[0087] Illustrative Embodiment 28: The non-transitory computer readable medium of illustrative embodiment 27, wherein the instructions, when executed by the at least one processor, further cause the at least one processor to perform operations comprising determining the at least one attenuation correction value for each of the plurality of segments based on a type of corresponding segment.

[0088] Illustrative Embodiment 29: The non-transitory computer readable medium of any of illustrative embodiments 27-28, wherein the plurality of segments comprises a first segment identifying a first anatomical region, and a second segment identifying a second anatomical region.

[0089] Illustrative Embodiment 30: The non-transitory computer readable medium of illustrative embodiment 29, wherein the at least one attenuation correction value for each of the plurality of segments comprises a first attenuation correction value for the first segment, and a second attenuation correction value for the second segment, wherein the first attenuation correction value differs from the second attenuation correction value.

[0090] Illustrative Embodiment 31 : The non-transitory computer readable medium of any of illustrative embodiments 27-30, wherein the at least one attenuation correction values are based on historical attenuation correction values.

[0091] Illustrative Embodiment 32: The non-transitory computer readable medium of any of illustrative embodiments 27-31, wherein the at least one attenuation correction values is an average of the historical attenuation correction values.

[0092] Illustrative Embodiment 33 : The non-transitory computer readable medium of any of illustrative embodiments 27-32, wherein the instructions, when executed by the at least one processor, further cause the at least one processor to perform operations comprising: for each of the plurality of segments: determining a type of segment; and based on the type of segment, receiving, from a data repository, the at least one attenuation correction value for the type of segment.

[0093] Illustrative Embodiment 34: The non-transitory computer readable medium of any of illustrative embodiments 27-33, wherein the instructions, when executed by the at least one processor, further cause the at least one processor to perform operations comprising applying a trained neural network to the segmentation data and, based on the application of the trained neural network to the segmentation data, generating the attenuation correction data.

[0094] Illustrative Embodiment 35: The non-transitory computer readable medium of any of illustrative embodiments 27-34, wherein the instructions, when executed by the at least one processor, further cause the at least one processor to perform operations comprising applying a trained neural network to the emission data and the attenuation correction data and, based onthe application of the trained neural network to the emission data and the attenuation correction data, reconstructing the image.

[0095] Illustrative Embodiment 36: The non-transitory computer readable medium of any of illustrative embodiments 27-35, wherein the emission data is positron emission tomography measurement data.

[0096] Illustrative Embodiment 37: The non-transitory computer readable medium of any of illustrative embodiments 27-36, wherein the segmentation process is a K-means clustering process.

[0097] Illustrative Embodiment 38: The non-transitory computer readable medium of any of illustrative embodiments 27-37, wherein the instructions, when executed by the at least one processor, further cause the at least one processor to perform operations comprising storing the image in a data repository.

[0098] Illustrative Embodiment 39: The non-transitory computer readable medium of any of illustrative embodiments 27-38, wherein the instructions, when executed by the at least one processor, further cause the at least one processor to perform operations comprising providing the image for display.

[0099] Illustrative Embodiment 40: An apparatus comprising: a means for receiving emission data from an image scanning system;a means for applying a segmentation process to the emission data and, based on the segmentation process, generating segmentation data characterizing a plurality of segments; a means for generating attenuation correction data comprising at least one attenuation correction value for each of the plurality of segments; and a means for

[0100] Illustrative Embodiment 41 : The apparatus of illustrative embodiment 14, comprising determining the at least one attenuation correction value for each of the plurality of segments based on a type of corresponding segment.

[0101] Illustrative Embodiment 42: The apparatus of any of illustrative embodiments 14-15, wherein the plurality of segments comprises a first segment identifying a first anatomical region, and a second segment identifying a second anatomical region.

[0102] Illustrative Embodiment 43 : The apparatus of illustrative embodiment 16, wherein the at least one attenuation correction value for each of the plurality of segments comprises a first attenuation correction value for the first segment, and a second attenuation correction value for the second segment, wherein the first attenuation correction value differs from the second attenuation correction value.

[0103] Illustrative Embodiment 44: The apparatus of any of illustrative embodiments 14-17, wherein the at least one attenuation correction values are based on historical attenuation correction values.

[0104] Illustrative Embodiment 45 : The apparatus of any of illustrative embodiments14-18, wherein the at least one attenuation correction values is an average of the historical attenuation correction values.

[0105] Illustrative Embodiment 46: The apparatus of any of illustrative embodiments 14-19, comprising: for each of the plurality of segments: a means for determining a type of segment; and a means for, based on the type of segment, receiving, from a data repository, the at least one attenuation correction value for the type of segment.

[0106] Illustrative Embodiment 47 : The apparatus of any of illustrative embodiments 14-20, comprising a means for applying a trained neural network to the segmentation data and, based on the application of the trained neural network to the segmentation data, generating the attenuation correction data.

[0107] Illustrative Embodiment 48 : The apparatus of any of illustrative embodiments 14-21, comprising a means for applying a trained neural network to the emission data and the attenuation correction data and, based on the application of the trained neural network to the emission data and the attenuation correction data, reconstructing the image.

[0108] Illustrative Embodiment 49: The apparatus of any of illustrative embodiments14-22, wherein the emission data is positron emission tomography measurement data.

[0109] Illustrative Embodiment 50: The apparatus of any of illustrative embodiments14-23, wherein the segmentation process is a K-means clustering process.

[0110] Illustrative Embodiment 51 : The apparatus of any of illustrative embodiments 14-24, comprising a means for storing the image in a data repository.

[0111] Illustrative Embodiment 52: The apparatus of any of illustrative embodiments 14-25, comprising a means for providing the image for display.

[0112] The apparatuses and processes are not limited to the specific embodiments described herein. In addition, components of each apparatus and each process can be practiced independent and separate from other components and processes described herein.

[0113] The previous description of embodiments is provided to enable any person skilled in the art to practice the disclosure. The various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein can be applied to other embodiments without the use of inventive faculty. The present disclosure is not intended to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

What is claimed is:

1. An apparatus comprising: a memory device storing instructions; and at least one processor communicatively coupled the memory device, wherein the at least one processor is configured to execute the instructions to: receive emission data from an image scanning system; apply a segmentation process to the emission data and, based on the segmentation process, generate segmentation data characterizing a plurality of segments; generate attenuation correction data comprising at least one attenuation correction value for each of the plurality of segments; and reconstruct an image based on the emission data and the attenuation correction data.

2. The apparatus of claim 1, wherein the at least one processor is configured to execute the instructions to determine the at least one attenuation correction value for each of the plurality of segments based on a type of corresponding segment.

3. The apparatus of claim 1, wherein the plurality of segments comprises a first segment identifying a first anatomical region, and a second segment identifying a second anatomical region.

4. The apparatus of claim 3, wherein the at least one attenuation correction value for each of the plurality of segments comprises a first attenuation correction value for the first segment, anda second attenuation correction value for the second segment, wherein the first attenuation correction value differs from the second attenuation correction value.

5. The apparatus of claim 1, wherein the at least one attenuation correction values are based on historical attenuation correction values.

6. The apparatus of claim 1, wherein the at least one attenuation correction values is an average of the historical attenuation correction values.

7. The apparatus of claim 1, wherein the at least one processor is configured to execute the instructions to: for each of the plurality of segments: determine a type of segment; and based on the type of segment, receive, from a data repository, the at least one attenuation correction value for the type of segment.

8. The apparatus of claim 1, wherein the at least one processor is configured to execute the instructions to apply a trained neural network to the segmentation data and, based on the application of the trained neural network to the segmentation data, generate the attenuation correction data.

9. The apparatus of claim 1, wherein the at least one processor is configured to execute the instructions to apply a trained neural network to the emission data and the attenuation correctiondata and, based on the application of the trained neural network to the emission data and the attenuation correction data, reconstruct the image.

10. The apparatus of claim 1, wherein the emission data is positron emission tomography measurement data.

11. The apparatus of claim 1, wherein the segmentation process is a K-means clustering process.

12. The apparatus of claim 1, wherein the at least one processor is configured to execute the instructions to store the image in a data repository.

13. The apparatus of claim 1, wherein the at least one processor is configured to execute the instructions to provide the image for display.

14. A method by at least one processor, the method comprising: receiving emission data from an image scanning system; applying a segmentation process to the emission data and, based on the segmentation process, generating segmentation data characterizing a plurality of segments; generating attenuation correction data comprising at least one attenuation correction value for each of the plurality of segments; and reconstructing an image based on the emission data and the attenuation correction data.

15. The method of claim 14, comprising determining the at least one attenuation correction value for each of the plurality of segments based on a type of corresponding segment.

16. The method of claim 14, wherein the plurality of segments comprises a first segment identifying a first anatomical region, and a second segment identifying a second anatomical region.

17. The method of claim 16, wherein the at least one attenuation correction value for each of the plurality of segments comprises a first attenuation correction value for the first segment, and a second attenuation correction value for the second segment, wherein the first attenuation correction value differs from the second attenuation correction value.

18. The method of claim 14, wherein the at least one attenuation correction values are based on historical attenuation correction values.

19. A non-transitory computer readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform operations comprising: receiving emission data from an image scanning system; applying a segmentation process to the emission data and, based on the segmentation process, generating segmentation data characterizing a plurality of segments; generating attenuation correction data comprising at least one attenuation correction value for each of the plurality of segments; andreconstructing an image based on the emission data and the attenuation correction data.

20. The non-transitory computer readable medium of claim 19, wherein the instructions, when executed by the at least one processor, further cause the at least one processor to perform operations comprising determining the at least one attenuation correction value for each of the plurality of segments based on a type of corresponding segment.

Images