Abstract: Systems and methods for training machine learning processes based on histo-projections, and for reconstructing medical images based on the trained machine learning processes, are disclosed. In some examples, a computing device receives image measurement data from an image scanning system, such as a positron emission tomography (PET) imaging system. The computing device applies a histogramming process to the image projection data and, based on applying the histogramming process, generates histo-projection data. Further, the computing device applies a trained machine learning process to the histo-projection data and, based applying the trained machine learning process to the histo-projection data, generates a reconstructed image. The computing device may provide the reconstructed image for display.

Abstract: Systems and methods for reconstructing medical images based on the trained deep learning processes, and for training deep learning processes, are disclosed. In some examples, image measurement data is received. A histo-image is generated based on the image measurement data. Further, an attenuation map, such as a ?-map, is received. An attenuation histo-image is generated based on the attenuation map. Further, a trained machine learning process, such as a trained neural network, is applied to features generated from the histo-image and the attenuation histo-image. Based on the application of the machine learning process to the histo-image and the attenuation histo-image, output image data characterizing an image volume is generated. In some examples, a machine learning process is trained based on histo-images and corresponding attenuation histo-images. The trained machine learning process may be employed to reconstruct images, such as positron emission tomography (PET) images.

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 of the scan data associated with the first imaging modality is performed based on the background radiation based attenuation map and a nuclear image is reconstructed from attenuation corrected scan data associated with the first imaging modality.

Abstract: Systems and methods for using paired high-resolution photon counting CT (PCCT) and PET images from the same patient to generate high-resolution PET images. A machine learning network is trained on paired images from patients. When the trained model is applied, a new patient's PET and PCCT images may be used to generate a high-resolution PET image for a medical diagnosis or further processing.

Abstract: Various systems and computer-implemented methods for background radiation based attenuation correction are disclosed. A first set of nuclear scan data including first scan data associated with a first imaging modality having a long-axial field of view and first background radiation data is received and a first background radiation attenuation map is generated by applying a trained machine-learning model to the first background radiation data. A first set of attenuation corrected scan data is generated by performing attenuation correction of the first scan data based only on the first background radiation attenuation map and a first image is reconstructed from the first set of attenuation corrected scan data. The disclosed background radiation based attenuation correction may be used for longer duration scans, repeat scans, and/or low-dose clinical applications, such as pediatric applications, theranostics, and/or other suitable applications.

Abstract: A method of identifying prompt gamma rays by triple-photon detection is disclosed. The method involves using two annihilation photons and a prompt gamma photon to determine the direction of the corresponding prompt gamma ray.

Abstract: Systems and methods include determination of a first time-of-flight offset for each of a plurality of crystals based on first annihilation radiation received by the plurality of crystals, determination of a second time-of-flight offset for each of the plurality of crystals based on radiation emitted by the plurality of crystals, determination, based on the second time-of-flight offsets, of a third time-of-flight offset for each of the plurality of crystals and associated with a response of the plurality of crystals to annihilation radiation, determination of whether the third time-of-flight offsets exceed a threshold, and, in response to a determination that the third time-of-flight offsets exceed the threshold, determine a fourth time-of-flight offset for each of the plurality of crystals based on second annihilation radiation received by the plurality of crystals.

Abstract: Systems and methods for reconstructing medical images based on the trained deep learning processes, and for training deep learning processes, are disclosed. In some examples, image measurement data is received. A histo-image is generated based on the image measurement data. Further, an attenuation map, such as a ?-map, is received. An attenuation histo-image is generated based on the attenuation map. Further, a trained machine learning process, such as a trained neural network, is applied to features generated from the histo-image and the attenuation histo-image. Based on the application of the machine learning process to the histo-image and the attenuation histo-image, output image data characterizing an image volume is generated. In some examples, a machine learning process is trained based on histo-images and corresponding attenuation histo-images. The trained machine learning process may be employed to reconstruct images, such as positron emission tomography (PET) images.

Abstract: Systems and methods for training end-to-end deep learning reconstruction processes, and for reconstructing medical images based on the trained deep learning processes, are disclosed. In some examples, input projection data is received. An untrained machine learning process is applied to the input projection data and, based on the application of the machine learning process to the projection data, an output image is generated. Further, a forward projection process is applied to the output image and, based on the application of the forward projection process to the output image, forward projected image data is generated. A loss value is then determined based on the forward projected image data and the input projection data. The loss value is then compared to a threshold value to determine whether the machine learning process is trained. The trained machine learning process may be employed to reconstruct images, such as positron emission tomography (PET) images.

Abstract: A framework for gantry alignment of a multimodality medical scanner. First image data of a non-radioactive structure is acquired by using intrinsic radiation emitted by scintillator crystals of detectors in a first gantry of the multimodality medical scanner. Second image data of the non-radioactive structure is acquired using a second gantry for another modality of the multimodality medical scanner. Image reconstruction may be performed based on the first and second image data of the non-radioactive structure to generate first and second reconstructed image volumes. A gantry alignment transformation that aligns the first and second reconstructed image volumes may then be determined.

Abstract: Various systems and computer-implemented methods for Radio Frequency (RF) coil attenuation correction are disclosed. PET time-of-flight (TOF) data generated by a PET imaging modality collocated with an MR imaging modality is received. RF coil attenuation data is extracted from the PET TOF data and an initial RF coil attenuation map is generated. A trained model configured to improve a signal to noise ratio of the initial RF coil attenuation map is applied to generate a final RF coil attenuation map. Attenuation correction of the PET TOF data is performed based on the final RF coil attenuation map. An image is reconstructed from attenuation corrected PET TOF data.

Abstract: A framework for gantry alignment of a multimodality medical scanner. First image data of a non-radioactive structure is acquired by using intrinsic radiation emitted by scintillator crystals of detectors in a first gantry of the multimodality medical scanner. Second image data of the non-radioactive structure is acquired using a second gantry for another modality of the multimodality medical scanner. Image reconstruction may be performed based on the first and second image data of the non-radioactive structure to generate first and second reconstructed image volumes. A gantry alignment transformation that aligns the first and second reconstructed image volumes may then be determined.

Abstract: A framework for gantry alignment of a multimodality medical scanner. First image data of a non-radioactive structure is acquired by using intrinsic radiation emitted by scintillator crystals of detectors in a first gantry of the multimodality medical scanner. Second image data of the non-radioactive structure is acquired using a second gantry for another modality of the multimodality medical scanner. Image reconstruction may be performed based on the first and second image data of the non-radioactive structure to generate first and second reconstructed image volumes. A gantry alignment transformation that aligns the first and second reconstructed image volumes may then be determined.

Abstract: Positron emission tomography (PET) with partially known attenuation accounts for the missing attenuation. Since a computed tomography (CT) scan may provide attenuation for less than all the locations used in PET reconstruction, artificial intelligence corrects for the missing attenuation. For example, the unknown attenuation or attenuation correction factors are estimated by the artificial intelligence. The known and estimated attenuations or correction factors are used in the PET reconstruction, providing more uniform PET sensitivity and better accounting for scatter. As another example, the artificial intelligence alters intensity of the activity in some locations to account for reconstruction with missing attenuation information, correcting for sensitivity variation and/or lack of scatter information for some locations.

Abstract: Systems and methods of dynamic PET imaging are disclosed. A system includes a positron emission tomography (PET) imaging modality configured to execute a first scan to acquire a first PET dataset and a processor. The first PET dataset includes dynamic PET data. The processor is configured to back-project the first PET dataset to generate a plurality of histo-image frames, input each of the plurality of histo-image frames to a trained neural network, and receive a dynamic PET output from the trained neural network. Each of the histo-image frames corresponds to a first axial position of the PET imaging modality.

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 of the scan data associated with the first imaging modality is performed based on the background radiation based attenuation map and a nuclear image is reconstructed from attenuation corrected scan data associated with the first imaging modality.

Abstract: Various systems and computer-implemented methods for background radiation based attenuation correction are disclosed. A first set of nuclear scan data including first scan data associated with a first imaging modality having a long-axial field of view and first background radiation data is received and a first background radiation attenuation map is generated by applying a trained machine-learning model to the first background radiation data. A first set of attenuation corrected scan data is generated by performing attenuation correction of the first scan data based only on the first background radiation attenuation map and a first image is reconstructed from the first set of attenuation corrected scan data. The disclosed background radiation based attenuation correction may be used for longer duration scans, repeat scans, and/or low-dose clinical applications, such as pediatric applications, theranostics, and/or other suitable applications.

Abstract: A computer-implemented method for generating a motion corrected image is provided. The method includes receiving listmode data collected by an imaging system; producing two or more histo-image frames or two or more histo-projection frames based on the listmode data; providing the two or more histo-image frames or two or more histo-projection frames to an Artificial Intelligence (AI) system; receiving two or more AI reconstructed images from the AI system based on the two or more histo-image frames or the two or more histo-projection frames; and generating a motion estimation in reconstructed images by using a motion free AI reconstructed image frame as a reference frame.

Abstract: A framework for gantry alignment of a multimodality medical scanner. First image data of a non-radioactive structure is acquired by using intrinsic radiation emitted by scintillator crystals of detectors in a first gantry of the multimodality medical scanner. Second image data of the non-radioactive structure is acquired using a second gantry for another modality of the multimodality medical scanner. Image reconstruction may be performed based on the first and second image data of the non-radioactive structure to generate first and second reconstructed image volumes. A gantry alignment transformation that aligns the first and second reconstructed image volumes may then be determined.

Abstract: An image reconstruction system generates a motion estimation using images that have been reconstructed using AI processing. The system receives listmode data collected by an imaging system and produces two or more histo-images based on the listmode data. The system provides the two or more histo-images to an AI system and receives two or more AI reconstructed images back from the AI system based on the two or more histo-images. The system generates a motion estimation based on the two or more AI reconstructed images.