Classifications

• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
  • A61B6/48—Diagnostic techniques
  • A61B6/482—Diagnostic techniques involving multiple energy imaging
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T12/00—Tomographic reconstruction from projections
  • G06T12/30—Image post-processing, e.g. metal artefact correction
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T5/00—Image enhancement or restoration
  • G06T5/60—Image enhancement or restoration using machine learning, e.g. neural networks
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T5/00—Image enhancement or restoration
  • G06T5/90—Dynamic range modification of images or parts thereof
  • G06T5/92—Dynamic range modification of images or parts thereof based on global image properties
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/10—Image acquisition modality
  • G06T2207/10072—Tomographic images
  • G06T2207/10081—Computed x-ray tomography [CT]
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/20—Special algorithmic details
  • G06T2207/20081—Training; Learning
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/20—Special algorithmic details
  • G06T2207/20084—Artificial neural networks [ANN]
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/30—Subject of image; Context of image processing
  • G06T2207/30004—Biomedical image processing

Definitions

• the following generally relates to imaging and more particularly to improving an image quality of virtual non-contrast images generated with data acquired with a spectral computed tomography (CT) scanner.
• CT computed tomography
• a CT scanner generally includes a single broadband x-ray tube mounted on a rotatable gantry opposite a detector array.
• the x-ray tube rotates around an examination region located between the x-ray tube and the detector array and emits polychromatic radiation that traverses the examination region.
• the detector array detects radiation that traverses the examination region and generates projection data.
• the projection data is reconstructed to generate volumetric image data, which can be used to generate one or more two-dimensional images.
• the images include pixels represented in terms of gray scale values corresponding to relative radiodensity. These values reflect the attenuation characteristics of the scanned subject and generally show structure such as anatomical structures within the subject.
• a contrast-enhanced CT scan includes scanning a subject that has been administered a radiocontrast agent. Procedures including a contrast-enhanced scan have also included a non-contrast-enhanced scan. For example, for coronary artery disease, contrast- enhanced images are useful to detect vessel narrowing and/or leakage, but are not well- suited to detect vessel calcification because of presence of iodine, and a separate non-enhanced scan is also performed for detecting vessel calcification. In another example, for renal masses, patients have undergone a non-contrast-enhanced scan and a contrast-enhanced
• nephrographic scan where attenuation values between the non-contrast-enhanced and the contrast-enhanced nephrographic images are compared to determine whether a mass is enhancing.
• performing two scans instead of one, increases total patient x-ray radiation dose and scan time, and reduces patient throughput.
• the detected radiation also includes spectral information as the absorption of the radiation by the subject and/or object is dependent on the energy of the photons traversing therethrough.
• spectral information can provide additional information such as information indicative of the elemental or material
• composition e.g., atomic number
• the projection data does not reflect the spectral characteristics as the signal output by the detector array is proportional to the energy fluence integrated over the energy spectrum.
• a CT scanner is also referred to herein as a non- spectral CT scanner.
• a CT scanner configured for spectral imaging leverages the spectral information.
• Such a CT scanner is referred to herein as a spectral CT scanner.
• a spectral CT scanner takes advantage of the differences in energy-dependent attenuation properties of materials in the subject (e.g., bone, soft tissue, fat, injected contrast agent) to provide material- specific images through material decomposition.
• An example spectral CT scanner is configured for dual-energy in which projection data is acquired at two different (e.g., higher and lower) energy levels.
• Example dual-energy configurations include: 1) one x-ray tube emitting x-ray radiation at one energy level and two layers of x-ray detectors respectively detecting lower energy x-rays and higher energy x-rays; 2) one x-ray tube with fast kV-switching and a single-layer detector, and 3) two x-ray tube / single-layer detector pairs angularly offset (e.g., 90 degrees) from each other.
• VNC virtual non-contrast
• TCT true non-contrast
• a spectral CT scanner can generate both contrast-enhanced images and VNC images from a single contrast-enhanced scan. In one instance, this mitigates having to perform both a non- contrast-enhanced scan and a contrast-enhanced scan for studies requiring both types of scans. As a result, total patient x-ray radiation dose and scan time are reduced, and patient throughput can be increased.
• VNC images tends to be degraded relative to TNC images.
• the signal can become smaller than ground truth or even missing relative to TNC images.
• the literature provides an example where one kidney stone is present and another is missing in the same VNC images, where the TNC image includes both.
• VNC images tend to exhibit higher noise and more artifact relative to TNC images. Higher noise may lead to reduced sensitivity and an increased false-positive rate in detection. Examples of artifact includes beam hardening and rim artifact. Unfortunately, such degradation in image quality can affect diagnosis.
• a spectral imaging system includes a radiation source configured to emit x-ray radiation and a detector array configured to detect x-ray radiation and generate spectral data.
• the spectral imaging system further includes a memory configured to store a virtual non-contrast image enhancing module that includes computer executable instructions including a neural network trained to produce image quality enhanced virtual non-contrast images.
• the neural network is trained with training spectral data and training non-contrast-enhanced images generated from a non-contrast-enhanced scan.
• the spectral imaging system further includes a processor configured to process the spectral data with the trained neural network to produce the image quality enhanced virtual non-contrast images.
• a spectral imaging system in another aspect, includes a radiation source configured to emit x-ray radiation and detector array configured to detect x-ray radiation and generate spectral data.
• the spectral imaging system further includes a memory configured to store a virtual non-contrast image enhancing module that includes computer executable instructions including a neural network.
• the spectral imaging system further includes a processor configured to train the neural network to produce image quality enhanced virtual non-contrast images with training spectral data and training non-contrast-enhanced images generated from a non-contrast-enhanced scan.
• a computer readable storage medium is encoded with computer readable instructions, which, when executed by a computer processor of a computing system, causes the computer processor to: emit x-ray radiation with a radiation source, detect emitted x-ray radiation with a detector array and generate a spectral data, train a neural network trained to produce image quality enhanced virtual non-contrast images, and process the spectral data with the trained neural network to produce the image quality enhanced virtual non-contrast images.
• the invention may take form in various components and arrangements of components, and in various steps and arrangements of steps.
• the drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention.
• FIGURE 1 schematically illustrates an example spectral computed tomography scanner with a VNC image enhancing module that includes a neural network trained to improve an image quality of VNC images generated by the scanner.
• FIGURE 2 schematically illustrates an example of training the neural network with a set of training VNC images and a set of training TNC images.
• FIGURE 3 schematically illustrates an example of employing the trained neural network of FIGURE 2 with VNC images generated by the spectral computed tomography scanner to produce image quality improved VNC images.
• FIGURE 4 schematically illustrates an example of training the neural network with a set of training dual-energy images and the set of training TNC images.
• FIGURE 5 schematically illustrates an example of employing the trained neural network of FIGURE 4 with contrast-enhanced dual-energy images generated by the spectral computed tomography scanner to produce image quality improved VNC images.
• FIGURE 6 schematically illustrates an example of training the neural network with a set of training dual-energy projection data and the set of TNC images.
• FIGURE 7 schematically illustrates an example of employing the trained neural network of FIGURE 6 with contrast-enhanced dual-energy projection data generated by the spectral computed tomography scanner to produce image quality improved VNC images.
• FIGURE 8 illustrates an example method for training a neural network with training VNC and TNC images and employing the trained neural network to produce image quality improved VNC images.
• FIGURE 9 illustrates an example method for training a neural network with training dual-energy and the TNC images and employing the trained neural network to produce image quality improved VNC images.
• FIGURE 10 illustrates an example method for training a neural network with training dual-energy projection data and the TNC images and employing the trained neural network to produce image quality improved VNC images.
• VNC virtual non-contrast
• Contrast-enhanced dual-energy spectral data such as low and high kVp data from a spectral CT scanner, are reconstructed to generate contrast- enhanced low and high kVp images, which are decomposed to generate spectral images, including the VNC image.
• the two kVp images will include beam hardening artifacts, e.g., due to the contrast (i.e. photoelectric absorption).
• the two kVp images will likely include residual beam hardening artifacts.
• the residual beam hardening artifacts can be further reduced during the decomposition, but the resulting VNC image will still include residual beam hardening artifacts, which can reduce image quality.
• a neural network is trained with a set of training spectral images as input and a set of training TNC images as reference data.
• the neural network learns to map the set of training spectral images to the set of training TNC images.
• the set of training spectral images includes spectral images with different noise levels and the set of training TNC images as reference data includes low noise TNC images.
• the mapping is non-linear such that the residual beam hardening artifacts in the input VNC images are further reduced and/or do not show in the final VNC images, and noise in the input VNC images is reduced in the final VNC images.
• this approach can produce contrast-enhanced images and VNC images with an image quality similar or equivalent to TNC images.
• FIGURE 1 schematically illustrates an example computed tomography (CT) scanner 102 configured for spectral CT imaging, e.g., multi-energy imaging such as dual energy imaging.
• CT scanner 102 includes a stationary gantry 104 and a rotating gantry 106, which is rotatably supported by the stationary gantry 104 and rotates around an examination region 108 (and a portion of an object or subject therein) about a longitudinal or z-axis 110.
• CT computed tomography
• a radiation source 112 such as an x-ray tube, is supported by and rotates with the rotating gantry 106 around the examination region 108.
• the radiation source 112 emits x-ray radiation that is collimated to form a generally fan, wedge, or cone shaped x-ray radiation beam that traverses the examination region 108.
• the radiation source 112 is a single x-ray tube configured to emit broadband (polychromatic) radiation for a single selected peak emission voltage (kVp) of interest.
• the radiation source 112 is configured to switch between at least two different emission voltages (e.g., 70 keV, 100 keV, 120 keV, etc.) during a scan.
• the radiation source 112 includes two or more x-ray tubes angular offset on the rotating gantry 104 with each configured to emit radiation with a different mean energy spectrum.
• US 8,442,184 B2 describes a system with kVp switching and multiple x- ray tubes, and is incorporated herein by reference in its entirety.
• a radiation sensitive detector array 114 subtends an angular arc opposite the radiation source 112 across the examination region 108.
• the detector array 114 includes one or more rows of detectors that are arranged with respect to each other along the z-axis direction and detects radiation traversing the examination region 108.
• the detector array 114 For a contrast- enhanced scan, the detector array 114 generates contrast-enhanced spectral projection data (line integrals) such as contrast-enhanced high energy projection data and contrast-enhanced low energy projection data.
• the detector array 114 includes an energy-resolving detector such as a multi-layer scintillator/photo -sensor detector (e.g., US 7,968,853 B2, which is incorporated herein by reference in its entirety).
• the detector array 114 includes a photon counting (direct conversion) detector (e.g., WO 2009/072056 A2, which is incorporated herein by reference in its entirety).
• the radiation source 112 includes the broadband, kVp switching and/or multiple X-ray tube radiation sources.
• the radiation source 112 includes kVp switching and/or multiple X-ray tube radiation sources.
• a reconstructor 116 processes the contrast-enhanced spectral projection data and generates contrast-enhanced spectral volumetric image data such as contrast-enhanced high energy volumetric image data and contrast-enhanced low energy volumetric image data.
• the reconstructor 116 can also generate contrast-enhanced non-spectral volumetric image data, e.g., by first combining the high energy projection data and the low energy projection data and then reconstructing the combined projection data and/or by combining the high energy volumetric image data and low energy volumetric image data.
• Spectral and non spectral images can be derived therefrom.
• the reconstructor 116 can be implemented with a processor such as a central processing unit (CPU), a microprocessor, etc.
• a basis material decomposer 118 decomposes the contrast-enhanced projection data (projection domain decomposition) and/or the contrast- enhanced spectral volumetric image data and/or images (image domain decomposition) into basis components.
• the reconstructor 116 reconstructs the projection data basis components to generate basis component volumetric image data and/or images such as VNC volumetric image data and/or images.
• the basis material decomposer 118 decomposes contrast-enhanced spectral volumetric image data and/or images into basis component volumetric image data and/or images such as VNC volumetric image data and/or images.
• separate basis material decomposers are used for projection and image domain decomposition.
• the basis material decomposer 118 is omitted or not employed.
• the basis material decomposer 118 employs two material decomposition algorithms, each assuming a water-iodine basis material pair or a calcium- iodine basis material pair, to generate VNC images without iodine.
• Another approach is to perform an N-dimensional cluster analysis to decompose the images into components such as iodine or other materials including soft tissue, calcium, etc.
• Other examples are described in Song, et al,“Virtual Non-Contrast CT Using Dual-Energy Spectral CT: Feasibility of Coronary Artery Calcium Scoring,” Korean J Radiol 2016;17(3):321 -329, and US
• the decomposer 118 can be implemented with a processor or the like.
• a subject support 120 such as a couch, supports a subject or object (e.g., a phantom) 122 in the examination region 108.
• the subject support 120 is movable in coordination with performing an imaging procedure so as to guide the subject or object with respect to the examination region 108 for loading, scanning, and/or unloading the subject or object.
• An injector 124 is configured to inject or administer a material such as one or more (e.g., iodine) contrast agents to the subject or object 122 to be scanned for a perfusion scan.
• a contrast agent can additionally or alternatively be manually administered by a clinician or the like. Where the contrast agent is manually administered to the subject or object 122, the injector 124 can be omitted.
• An operator console 126 includes a human readable output device 128 such as a display monitor, a filmer, etc. and an input device 130 such as a keyboard, mouse, etc.
• the console 126 further includes a processor 132 (e.g., a CPU, a microprocessor, etc.) and computer readable storage medium 134 (which excludes transitory medium) such as physical memory.
• the computer readable storage medium 134 includes a VNC image enhancing module 136.
• the illustrated VNC image enhancing module 136 includes computer executable instructions for processing VNC images to produce improved image quality VNC image (image quality enhanced virtual non-contrast images) similar or equivalent to TNC images generated from a non-contrast-enhanced scan.
• the instructions include a neural network trained with training spectral data and TNC images.
• the spectral CT scanner 102 can produce contrast-enhanced images and VNC images with image quality diagnostically similar or equivalent to TNC images generated from a non-contrast-enhanced scan using data acquired with a single contrast-enhanced scan.
• FIGURES 2-7 schematically illustrate different embodiments for training the neural network.
• the neural network is trained to take into account differences in the delivery of contrast material in the input data.
• the training data sets are constructed such that the distribution of the data feature (i.e., contrast delivery) well represents the distribution of the data feature in real clinical scenarios.
• the training data sets can come from both real clinical scans and simulation.
• An advantage of a simulation is that contrast delivery can be controlled in every data set (including contrast-enhanced spectral CT data and TNC images) and thereby a large number of data sets can be generated that strictly follow a presumed distribution of the data feature.
• the input and training data can be acquired such that contrast delivery is consistent across the data.
• FIGURE 2 schematically illustrates example training of a neural network 202 with training VNC images generated using material decomposition and training TNC images from a non-contrast-enhanced CT scan.
• Parameters of the neural network are updated and converge through a number of iterations using an objective function 204.
• the objective function 204 includes a mathematical function that minimizes an error between the VNC volumetric image data or images output by the neural network 202 and TNC images, e.g., based on a mean squared and/or other difference between the VNC volumetric image data or images output by the neural network 202 and TNC images.
• the parameters are updated until the error falls below of predetermined threshold.
• Other stopping criteria include a predetermined number of iterations, a predetermined time duration, etc.
• FIGURE 3 schematically illustrates an embodiment in which the VNC image enhancing module 136 includes the trained neural network 202 of FIGURE 2.
• the trained neural network 202 receives the VNC images (which contain high noise and/or artifacts relative to TNC images) generated by the basis material decomposer 118 (FIGURE 1) and outputs image quality improved VNC images, which includes VNC images with less noise and/or artifact than the input VNC images and which has sufficient image quality for diagnosis.
• FIGURE 4 schematically illustrates a variation of FIGURE 2 in which the input to the neural network 202 is training dual-energy (high and low energy) images.
• FIGURE 5 schematically illustrates an embodiment in which the VNC image enhancing module 136 includes the neural network 202 trained in FIGURE 4.
• the trained neural network 202 receives contrast-enhanced dual-energy images generated by the reconstructor 116 (FIGURE 1) and outputs image quality improved VNC images.
• the basis material decomposer 118 is either omitted from the system 102 of FIGURE 1 or not utilized.
• the neural network 202 handles the material decomposition as it is capable of performing nonlinear mapping of data, and material decomposition often involves solving polynomial equations (e.g., for high energy and low energy) after approximations.
• the noise in the basis material images can be lower than the noise from conventional material
• decomposition e.g., because conventional material decomposition only uses measured data on one detector pixel or reconstructed value of one image voxel at a time, but neural network uses a small patch, which includes more than one pixel or voxel.
• projection data includes unique spectral information hidden in the data.
• the physical effects i.e., Compton scatter and photoelectric absorption
• the physical effects can be different with low-Z, high density objects leading to more Compton scatter, and these differences are hidden in the acquired raw projection data.
• Another example is that there are more beam hardening artifacts for high-Z materials relative to low-Z materials. These differences are reduced and/or removed and do not show up in the reconstructed volumetric image data, e.g., to effective scatter correction, beam hardening compensation, etc.
• corrections that remove spectral information are omitted from the reconstruction.
• the neural network 202 of FIGURES 2-3 is trained with VNC images generated from volumetric image data that was reconstructed without using a correction, such as scatter correction and/or beam-hardening compensation, that removes the unique spectral information
• the neural network 202 of FIGURES 4-5 is trained with high and low energy images reconstructed without using a correction, such as scatter correction and/or beam-hardening compensation, that removes the unique spectral information. This allows the training to take into account the unique spectral information.
• FIGURE 6 schematically illustrates a variation of FIGURE 2 in which the input to the neural network 202 is training dual-energy (high and low energy) projection data.
• FIGURE 7 schematically illustrates an embodiment in which the VNC image enhancing module 136 includes the neural network 202 trained in FIGURE 6.
• the trained neural network 202 receives contrast-enhanced dual-energy projection data generated by the detector array 114 (FIGURE 1) and outputs image quality improved VNC images.
• the basis material decomposer 118 is either omitted from the system 102 of FIGURE 1 or not utilized.
• a variation includes two of the three embodiments described in FIGURES 2- 7.
• one variation includes the neural network of FIGURES 2 and 4.
• Another variation includes the neural network of FIGURES 2 and 6.
• Another variation includes the neural network of FIGURES 4 and 6.
• Another variation includes all three of the three embodiments described in FIGURES 2-7.
• Another variation includes a single neural network trained with different types of training such as the training data in FIGURES 2 and 4, 2 and 6, 2 and 6, or 2, 4 and 6.
• FIGURE 8 illustrates an example method in accordance with an embodiment herein.
• a set of training VNC images is received, as described herein and/or otherwise.
• a set of training TNC images is received, as described herein and/or otherwise.
• the set of training VNC images and the set of training TNC images are employed to train a neural network, as described herein and/or otherwise.
• VNC images are received, as described herein and/or otherwise.
• the received VNC images are processed with the trained neural network to produce image quality improved VNC images.
• FIGURE 9 illustrates another example method in accordance with an embodiment herein.
• a set of training dual-energy images is received, as described herein and/or otherwise.
• a set of training TNC images is received, as described herein and/or otherwise.
• the set of training dual-energy images and the set of training TNC images are employed to train a neural network, as described herein and/or otherwise.
• contrast-enhanced dual-energy images are received, as described herein and/or otherwise.
• the contrast-enhanced dual-energy images are processed with the trained neural network to produce image quality improved VNC images.
• FIGURE 10 illustrates another example method in accordance with an embodiment herein.
• a set of training dual-energy projection data is received, as described herein and/or otherwise.
• a set of training TNC images is received, as described herein and/or otherwise.
• the set of training dual-energy projection data and the set of training TNC images are employed to train a neural network, as described herein and/or otherwise.
• contrast-enhanced dual-energy projection data is received, as described herein and/or otherwise.
• the contrast-enhanced dual-energy projection data is processed with the trained neural network to produce image quality improved VNC images.
• the above may be implemented by way of computer readable instructions, encoded or embedded on computer readable storage medium, which, when executed by a computer processor(s), cause the processor(s) to carry out the described acts. Additionally, or alternatively, at least one of the computer readable instructions is carried by a signal, carrier wave or other transitory medium, which is not computer readable storage medium.
• the word“comprising” does not exclude other elements or steps, and the indefinite article“a” or“an” does not exclude a plurality.
• a single processor or other unit may fulfill the functions of several items recited in the claims.
• the mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measured cannot be used to advantage.
• a computer program may be stored/distributed on a suitable medium, such as an optical storage medium or a solid-state medium supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems. Any reference signs in the claims should not be construed as limiting the scope.

Landscapes

• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• General Physics & Mathematics (AREA)
• Theoretical Computer Science (AREA)
• Health & Medical Sciences (AREA)
• Life Sciences & Earth Sciences (AREA)
• Medical Informatics (AREA)
• Optics & Photonics (AREA)
• Heart & Thoracic Surgery (AREA)
• Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
• Biophysics (AREA)
• Pathology (AREA)
• Radiology & Medical Imaging (AREA)
• Biomedical Technology (AREA)
• High Energy & Nuclear Physics (AREA)
• Molecular Biology (AREA)
• Surgery (AREA)
• Animal Behavior & Ethology (AREA)
• General Health & Medical Sciences (AREA)
• Public Health (AREA)
• Veterinary Medicine (AREA)
• Apparatus For Radiation Diagnosis (AREA)
• Image Analysis (AREA)

Priority Applications (5)

Applications Claiming Priority (2)

Publications (1)

Family

ID=65243573

Family Applications (1)

Country Status (6)

Cited By (7)

Families Citing this family (3)

Citations (5)

Family Cites Families (9)

• 2019
  • 2019-01-30 US US16/965,225 patent/US11813101B2/en active Active
  • 2019-01-30 EP EP19702422.7A patent/EP3746979B1/en active Active
  • 2019-01-30 ES ES19702422T patent/ES2968760T3/es active Active
  • 2019-01-30 CN CN201980011363.6A patent/CN111684486B/zh active Active
  • 2019-01-30 JP JP2020541482A patent/JP7346429B2/ja active Active
  • 2019-01-30 WO PCT/EP2019/052157 patent/WO2019149711A1/en not_active Ceased

Patent Citations (5)

Non-Patent Citations (5)

Also Published As

Similar Documents

Legal Events