Canon Medical Systems Corporation

Toshiba Medical Systems Corporation

University of California

Evren Asma

Wenyuan Qi

Li Yang

Chung Chan

Yi Qiang

Jeffrey Kolthammer

Jian Zhou

Peng Peng

Wenli Wang

Xiaoli Li

Yujie Lu

Jinyi Qi

Tiantian Li

Zhaoheng Xie

Zhou Yu

Kent C Burr

Manabu Teshigawara

Xiaofeng Niu

Ting Xia

Ryo Okuda

Qiang Yi

A method, apparatus, and non-transitory computer-readable storage medium for image denoising whereby a deep image prior (DIP) neural network is trained to produce a denoised image by inputting the second medical image to the DIP neural network and combining a converging noise and an output of the DIP network during the training such that the converging noise combined with the output of the DIP network approximates the first medical image at the end of the training, wherein the output of the DIP network represents the denoised image.

Apparatus and methods for unsupervised image denoising using double over-parameterization

A method, system, and computer readable medium to perform nuclear medicine scatter correction estimation, sinogram estimation and image reconstruction from emission and attenuation correction data using deep convolutional neural networks. In one embodiment, a Deep Convolutional Neural network (DCNN) is used, although multiple neural networks can be used (e.g., for angle-specific processing). In one embodiment, a scatter sinogram is directly estimated using a DCNN from emission and attenuation correction data. In another embodiment a DCNN is used to estimate a scatter-corrected image and then the scatter sinogram is computed by a forward projection.

Scatter estimation for PET from image-based convolutional neural network

A PET scanner includes gamma-ray detector rings that form a bore through which an imaging subject is translated, a length of the bore defining an axial length of the PET scanner, the gamma-ray detector rings being movable along the axial length, the gamma-ray detector rings including gamma-ray detector modules therein, and processing circuitry configured to receive PET data associated with a plurality of transaxial slices of the imaging subject, the PET data including a first set of spatial information and timing information corresponding to a first data acquisition period for the gamma-ray detector modules in a first axial position and a second set of spatial information and timing information corresponding to a second data acquisition period for the gamma-ray detector modules in a second axial position, and reconstruct a PET image based on the first set of spatial and timing information and the second set of spatial and timing information.

Pet reconstruction for adjustable pet system

A deep learning (DL) convolution neural network (CNN) reduces noise in positron emission tomography (PET) images, and is trained using a range of noise levels for the low-quality images having high noise in the training dataset to produce uniform high-quality images having low noise, independently of the noise level of the input image. The DL-CNN network can be implemented by slicing a three-dimensional (3D) PET image into 2D slices along transaxial, coronal, and sagittal planes, using three separate 2D CNN networks for each respective plane, and averaging the outputs from these three separate 2D CNN networks. Feature-oriented training can be implemented by segmenting each training image into lesion and background regions, and, in the loss function, applying greater weights to voxels in the lesion region. Other medical images (e.g. MRI and CT) can be used to enhance resolution of the PET images and provide partial volume corrections.

Apparatus and method for medical image reconstruction using deep learning to improve image quality in positron emission tomography (PET)

Upon receiving list-mode data by detecting radiation emitted from a radiation source positioned with a field of view of a medical imaging scanner, each photon included in the list-mode data can be classified according to one or more interaction properties, such as energy or number of crystals interacted with. Grouped pairs of photons can be generated based on the classifying, and a corresponding interaction-property-specific correction kernel (e.g., a corresponding interaction-property-specific point spread function correction kernel) can be selected for each group. Corresponding interaction-property-specific correction kernels can then be utilized to generate higher quality images.

Event property-dependent point spread function modeling and image reconstruction for PET

Existing, low quality images can be restored using reconstruction or a combination of post-reconstruction techniques to generate a real patient phantom. The real patient phantom (RPP) can then be simulated in Monte Carlo simulations of a higher performance system and a lower performance system. Alternatively, the RPP can be simulated in the higher performance system, and a real scan can be performed by an existing, lower performance system. The higher performance system can be differentiated from the lower performance system in a variety of ways, including a higher resolution time of flight measurement capability, a greater sensitivity, smaller detector crystals, or less scattering. A neural network can be trained using the images produce by the higher performance system as the target, and the images produced by the lower performance system as the input. After training, the trained neural network can be used to output input images taken in a lower performance system with higher performance system characteristics.

Neural network for improved performance of medical imaging systems

A guided pairing method includes generating a singles list by detecting a plurality of singles at a plurality of detector elements in a detector array, the plurality of singles falling within a plurality of detection windows; for each detection window of the plurality of detection windows in the singles list having exactly two singles of the plurality of singles, determining the line of responses (LORs) for each of the two singles of the plurality of singles; for each detection window of the plurality of detection windows in the singles list having more than two singles of the plurality of singles, determining all coincidences possible based on the more than two singles; generating a weight for said each coincidence of the coincidences based on the determined LORs for said each of the two singles of the plurality of singles; and pairing the more than two singles based on the generated weight for said each coincidence of the coincidences.

Method and apparatus for guided pairing of multi-coincidences for time of flight positron emission tomography

A system and method for training a neural network to denoise images. One noise realization is paired to an ensemble of training-ready noise realizations, and fed into a neural network for training. Training datasets can also be retrospectively generated based on existing patient studies to increase the number of training datasets.

Noise to noise ensemble learning for pet and CT image and data denoising

A method of normalizing detector elements in an imaging system is described herein. The method includes a line source that is easier to handle for a user, and decouples the normalization of the detector elements into a transaxial domain and an axial domain in order to isolate errors due to positioning of the line source. Additional simulations are performed to augment the real scanner normalization. A simulation of a simulated line source closely matching the real line source can be performed to isolate errors due to physical properties of the crystals and position of the crystals in the system, wherein the simulated detector crystals are otherwise modeled uniformly. A simulation of a simulated cylinder source can be performed to determine errors due to other effects stemming from gaps between the detector crystals.

Method and system for PET detector efficiency normalization

A deep learning (DL) convolution neural network (CNN) reduces noise in positron emission tomography (PET) images, and is trained using a range of noise levels for the low-quality images having high noise in the training dataset to produceuniform high-quality images having low noise, independently of the noise level of the input image. The DL-CNN network can be implemented by slicing a three-dimensional (3D) PET image into 2D slices along transaxial, coronal, and sagittal planes, using three separate 2D CNN networks for each respective plane, and averaging the outputs from these three separate 2D CNN networks. Feature-oriented training can be implemented by segmenting each training image into lesion and background regions, and, in the loss function, applying greater weights to voxels in the lesion region. Other medical images (e.g. MRI and CT) can be used to enhance resolution of the PET images and provide partial volume corrections.

The present disclosure is related to removing scatter from a SPECT scan by utilizing a radiative transfer equation (RTE) method. An attenuation map and emission map are acquired for generating scatter sources maps and scatter on detectors using the RTE method. The estimated scatter on detectors can be removed to produce an image of a SPECT scan with less scatter. Both first-order and multiple-order scatter can be estimated and removed. Additionally, scatter caused by multiple tracers can be determined and removed.

Total SPECT scatter estimation and correction using radiative transfer equation

The present disclosure relates to an apparatus for estimating scatter in positron emission tomography, comprising processing circuitry configured to acquire an emission map and an attenuation map, each representing an initial image reconstruction of a positron emission tomography scan, calculate, using a radiative transfer equation (RTE) method, a scatter source map of a subject of the positron emission tomography scan based on the emission map and the attenuation map, estimate, using the RTE method and based on the emission map, the attenuation map, and the scatter source map, scatter, and perform an iterative image reconstruction of the positron emission tomography scan based on the estimated scatter and raw data from the positron emission tomography scan of the subject.

Method and apparatus for scatter estimation in positron emission tomography

A method of imaging includes obtaining a plurality of dynamic sinograms, each dynamic sinogram representing detection events of gamma rays at a plurality of detector elements, summing the plurality of dynamic sinograms to generate an activity map based on a radioactivity level of the gamma rays; reconstructing, using the plurality of dynamic sinograms, a plurality of dynamic images, each of the plurality of dynamic images corresponding to one of the each of the plurality of dynamic sinograms, and generating, using the plurality of dynamic sinograms and the activity map, at least one parametric image.

Growing community of inventors

Buffalo Grove, IL, United States of America

Evren Asma