Non-spectral Computed Tomography (CT) scanner configured to generate spectral volumetric image data
阅读说明:本技术 被配置为生成谱体积图像数据的非谱计算机断层摄影(ct)扫描器 (Non-spectral Computed Tomography (CT) scanner configured to generate spectral volumetric image data ) 是由 白传勇 Y-M·朱 S·路 许世昱 党颢 H·赖 D·麦克奈特 汪蕙 于 2019-01-30 设计创作,主要内容包括:一种非谱计算机断层摄影扫描器(102)包括:辐射源(112),其被配置为发射X射线辐射;探测器阵列(114),其被配置为探测X射线辐射并且生成非谱数据;以及存储器(134),其被配置为存储谱图像模块(130),所述谱图像模块包括计算机可执行指令,所述计算机可执行指令包括被训练以产生谱体积图像数据的神经网络。所述神经网络是利用训练谱体积图像数据和训练非谱数据来训练的。所述非谱计算机断层摄影扫描器还包括:处理器(126),其被配置为利用经训练的神经网络来处理所述非谱数据以产生谱体积图像数据。(A non-spectral computed tomography scanner (102) comprising: a radiation source (112) configured to emit X-ray radiation; a detector array (114) configured to detect X-ray radiation and generate non-spectral data; and a memory (134) configured to store a spectral image module (130) comprising computer-executable instructions including a neural network trained to produce spectral volumetric image data. The neural network is trained using training spectral volumetric image data and training non-spectral data. The non-spectral computed tomography scanner further comprises: a processor (126) configured to process the non-spectral data with the trained neural network to produce spectral volumetric image data.)
1. A non-spectral computed tomography scanner (102), comprising:
a radiation source (112) configured to emit X-ray radiation;
a detector array (114) configured to detect X-ray radiation and generate non-spectral data;
a memory (134) configured to store a spectral image module (130) comprising computer-executable instructions including a neural network trained to produce spectral volumetric image data,
wherein the neural network is trained using training spectral volume image data and training non-spectral data; and
a processor (126) configured to process the non-spectral data with the trained neural network to produce spectral volumetric image data.
2. The scanner of claim 1, wherein the neural network is further trained using scanner geometry and physical information.
3. The scanner of any of claims 1 to 2, wherein the non-spectral data includes non-spectral projection data and the processor processes the non-spectral projection data with the trained neural network to produce the spectral volumetric image data.
4. The scanner of claim 3, wherein the training non-spectral data comprises training non-spectral projection data.
5. The scanner of claim 4, wherein the non-spectral data comprises non-spectral volumetric image data and the processor processes the non-spectral volumetric image data with the trained neural network to produce the spectral volumetric image data.
6. The scanner of claim 1, wherein the training non-spectral data comprises training non-spectral volumetric image data.
7. The scanner of claim 6, wherein the non-spectral volume image data is uncorrected spectral volume image data.
8. The scanner of claim 6, wherein the non-spectral volume image data is partially corrected spectral volume image data that is uncorrected for at least beam hardening and scattered radiation.
9. The scanner of any of claims 4 or 6 to 8, wherein the neural network is trained to minimize a difference between spectral volumetric image data generated by the neural network and the training non-spectral data.
10. The scanner of any of claims 6 to 7, wherein the non-spectral volume image data is corrected spectral volume image data.
11. A non-spectral computed tomography scanner comprising:
a radiation source configured to emit X-ray radiation;
a detector array configured to detect X-ray radiation and generate non-spectral data;
a memory configured to store a spectral image module comprising computer-executable instructions including a neural network trained to produce spectral volumetric image data; and
a processor configured to train the neural network with training spectral volume image data and training non-spectral data to generate spectral volume image data from the non-spectral data.
12. The scanner of claim 11, wherein the training non-spectral data comprises training non-spectral projection data.
13. The scanner of claim 12, wherein the neural network is trained to minimize a difference between spectral volumetric image data generated by the neural network and the training non-spectral projection data.
14. The scanner according to any of claims 12 to 13, wherein the non-spectral data comprises non-spectral projection data, and the processor is further configured to process the non-spectral projection data with a trained neural network to produce spectral data from the non-spectral projection data.
15. The scanner of claim 11, wherein the training non-spectral data comprises training non-spectral volumetric image data.
16. The scanner of claim 15, wherein the neural network is trained to minimize a difference between spectral volumetric image data generated by the neural network and the training non-spectral volumetric image data.
17. The scanner according to any of claims 15 to 16, wherein the non-spectral data comprises non-spectral volumetric image data and the processor processes the non-spectral volumetric image data utilizing a trained neural network to produce spectral volumetric image data.
18. A computer readable storage medium encoded with computer readable instructions, which, when executed by a processor of a computing system, causes the processor to:
emitting X-ray radiation with a radiation source;
detecting the emitted X-ray radiation with a detector array and generating a signal indicative thereof;
reconstructing the signals and generating non-spectral volumetric image data;
training a neural network to produce spectral volumetric image data; and is
A trained neural network is employed to produce spectral volumetric image data from the generated non-spectral data.
19. The computer-readable storage medium of claim 18, wherein the generated non-spectral data is non-spectral projection data generated by a non-spectral computed tomography scanner.
20. The computer-readable storage medium of claim 19, wherein the generated non-spectral data is non-spectral volumetric image data generated by a non-spectral computed tomography scanner.
Technical Field
The following generally relates to imaging and, more particularly, to a non-spectral Computed Tomography (CT) scanner configured to generate spectral volumetric image data.
Background
CT scanners typically include an X-ray tube mounted on a rotatable gantry opposite a detector array. The X-ray tube rotates about an examination region located between the X-ray tube and the detector array, and emits polychromatic radiation that traverses the examination region. A detector array detects radiation traversing the examination region and generates projection data. The projection data is reconstructed to generate volumetric image data, which may be processed to generate one or two dimensions. Typically, corrections (e.g., scatter corrections, beam hardening corrections) are applied during reconstruction. The image includes pixels represented in terms of gray values corresponding to relative radiodensity. These values reflect the attenuation properties of the scanned object and typically show structures such as anatomical structures within the object.
The detected radiation also includes spectral information, since the absorption of radiation by the object and/or target depends on the energy of the photons passing through it. Such spectral information may provide additional information, for example, information indicative of the elements or material composition (e.g., atomic number) of the tissue and/or material of the object and/or target. Unfortunately, the projection data does not reflect spectral characteristics because the signal output by the detector array is proportional to the energy fluence integrated over the energy spectrum. Such CT scanners are also referred to herein as non-spectral CT scanners. A CT scanner configured for spectral imaging utilizes this spectral information to provide further information indicative of the composition of the element or material. Such a CT scanner is referred to herein as a spectral CT scanner.
Example dual-spectrum configurations include: 1) an X-ray tube emitting X-ray radiation at an energy level, and two layers of X-ray detectors detecting lower energy X-rays and higher energy X-rays, respectively; 2) one X-ray tube and single layer detector with fast kV switching, and 3) two X-ray tube/single layer detector pairs angularly offset from each other. Examples of such further information include low energy high contrast images, effective Z (atomic number) images, virtual monochromatic images, contrast agent quantification maps, virtual non-contrast images, electron density images, and/or other spectral information. Unfortunately, the additional detector layer(s), X-ray tube/detector array, and/or kVp switching circuitry increase the overall cost of the imaging system, and data acquisition, handling, and processing increase complexity.
Disclosure of Invention
Aspects described herein address the above-referenced matters and others.
In one aspect, a non-spectral computed tomography scanner includes: a radiation source configured to emit X-ray radiation; a detector array configured to detect X-ray radiation and generate non-spectral data; and a memory configured to store a spectral image module comprising computer-executable instructions including a neural network trained to produce spectral volumetric image data. The neural network is trained using training spectral volumetric image data and training non-spectral data. The non-spectral computed tomography scanner further comprises: a processor configured to process the non-spectral data with a trained neural network to produce spectral volumetric image data.
In another aspect, a non-spectral computed tomography scanner includes: a radiation source configured to emit X-ray radiation; a detector array configured to detect X-ray radiation and generate non-spectral data; and a memory configured to store a spectral image module comprising computer-executable instructions including a neural network trained to produce spectral volumetric image data. The non-spectral computed tomography scanner further comprises: a processor configured to train the neural network with training spectral volume image data and training non-spectral data to generate spectral volume image data from the non-spectral data.
In another aspect, a computer-readable storage medium is encoded with computer-readable instructions that, when executed by a computer processor of a computing system, cause the computer processor to: emitting X-ray radiation with a radiation source; detecting the emitted X-ray radiation with a detector array and generating a signal indicative thereof; reconstructing the signals and generating non-spectral volumetric image data; training a neural network to produce spectral volumetric image data; and employing the trained neural network to produce spectral volumetric image data from the generated non-spectral data.
Still other aspects of the present application will be appreciated by those of ordinary skill in the art upon reading and understand the appended description.
Drawings
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 the purpose of illustrating preferred embodiments and are not to be construed as limiting the invention.
Fig. 1 schematically shows an example CT scanner comprising a spectral image module with a neural network for predicting spectral volumetric image data from non-spectral data.
Fig. 2 schematically shows an example of training a neural network using spectral and non-spectral (corrected or uncorrected) volumetric image data.
Fig. 3 schematically illustrates an example of training a neural network using spectral volumetric image data and non-spectral projection data.
FIG. 4 schematically illustrates a system for predicting spectral volumetric image data from a non-spectral volumetric image using the neural network trained in FIG. 2.
FIG. 5 schematically illustrates a system for predicting spectral volumetric image data from a non-spectral projection data image using the neural network trained in FIG. 3.
Fig. 6 schematically illustrates an example neural network.
Fig. 7 illustrates an example method according to fig. 1, 2 and 4.
Fig. 8 illustrates an example method according to fig. 1, 3 and 5.
Detailed Description
The following describes a method for generating spectral data from non-spectral projection data and/or non-spectral volumetric image data and system information generated by a non-spectral Computed Tomography (CT) scanner. Typically, non-spectral projection data includes unique spectral information hidden in the data. For example, even though the attenuation along a particular ray path may be the same for both high-Z low-density targets and low-Z high-density targets, the physical effects (i.e., compton scattering and photoelectric absorption) may be different, where the low-Z high-density targets cause more compton scattering, 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 minimized or do not show up in non-spectral volume images reconstructed using conventional CT reconstruction techniques due to effective scatter correction and beam-hardening compensation, among other things.
In one approach described herein, a neural network is trained using at least projection data (again, which includes unique spectral information), reference spectral data (i.e., data from the same scan and also from the same CT scanner), and system information. Through training, the neural network learns how to generate spectral data from non-spectral projection data and system information. For example, neural networks learn how to generate spectral data corresponding to compton scattering for high-Z low-density and low-Z high-density targets and photoelectric absorption for high-Z low-density and low-Z high-density targets. Further training may be performed to fine tune parameters of the neural network so that the neural network can process the training non-spectral data and produce spectral data within a given error of the training spectral data. Once trained, the trained neural network may process non-spectral projection data and system information from a scan of the patient and utilize the neural network to generate spectral data for the patient. As further described herein, the training data may include non-spectral volumetric image data.
In one case, the above allows the non-spectral CT scanner to generate not only non-spectral projection data and/or non-spectral volumetric image data, but also spectral projection data and/or spectral volumetric image data. Accordingly, a non-spectral CT scanner may generate spectral projection data and/or spectral volumetric image data without including additional hardware and/or circuitry of the spectral CT scanner, such as a multi-slice detector array, a plurality of X-ray tube/detector array pairs, and/or kVp switching circuitry. Furthermore, it mitigates the data acquisition, handling and processing complexity of the spectral imaging system.
Fig. 1 schematically illustrates an example non-spectral Computed Tomography (CT)
A
A radiation
A
A
The
The
With the configurations described herein,
Figure 2 schematically illustrates an example training of a neural network of the
In this example, the neural network is trained using data generated by a CT scanner 200 configured with a dual-layer detector array 202, which includes a first layer 204 and a second layer 206. A non-limiting example of a multi-layer detector array is described in US 7,968,853B 2 entitled "Double decker detector for spectral CT", filed on 10.2006, which is hereby incorporated by reference in its entirety. A beam 208 generated by an X-ray tube 210 traverses an examination region 212 and an object or subject 214 disposed therein and is detected by the dual layer detector array 202.
The first layer 204 produces first projection data that is processed by a non-spectral reconstructor 216 that employs a non-spectral reconstruction algorithm to produce non-spectral volumetric image data. In this example, the non-spectral reconstructor 216 applies beam hardening corrections, scatter corrections, and/or one or more other corrections, and the resulting non-spectral volumetric image data is referred to as corrected non-spectral volumetric image data.
The second layer 206 generates second projection data that is provided to a spectral reconstructor 218 along with the first projection data. The energy spectra of the first projection data and the second projection data are different, wherein the energy spectrum of the second projection data is higher than the energy spectrum of the first projection data. The spectral reconstructor 218 employs one or more spectral algorithms to generate one or more sets of spectral volumetric image data.
The
In one case, the
By using the known function f, I can be obtained byWithout artifacts=αμp+βμsAnd IArtifacts=(αμ,βμsS) are combined to solve for alpha and beta. The
The trainable parameters of the
In one case, the objective function 222 is implemented as a cost function l (w), for example, as shown in equation 1:
equation 1:
where NSI is the jth set of training non-spectral volumetric image data, Sinfo is the jth set of system specific information, SI is the jth set of training spectral volumetric image data, and T represents a transform that transforms NSI and Sinfo into SI using trainable parameters w,representing a regularization term, λ, that drives the equation towards a small parameter, w2Denotes a hyper-parameter, and k denotes an index. The function T parameterized by w is solved iteratively by minimizing a cost function.
In this example, the non-spectral reconstructor 216 applies at least scatter and/or beam hardening corrections, as discussed above. Such correction reduces or removes spectral information. For example and as described above, the photoelectric effect and compton scattering difference between low-Z high-density structures and high-Z low-density structures is reflected in the non-spectral projection data. Generally, the projection data for low-Z high-density structures will include more scatter. Scatter correction may remove or reduce this difference. In addition, there are more beam hardening artifacts for high Z materials than for low Z materials. Beam hardening corrections may remove or reduce such differences. In one variation, corrections to remove spectral information (e.g., scatter correction and/or beam hardening) are omitted. The resulting data is referred to herein as partially corrected (some corrections omitted) or uncorrected (all corrections omitted) non-spectral volumetric image data.
Figure 3 schematically illustrates another example training of a neural network of the
The
Fig. 2 and 3 depict the following examples: therein, the neural network is trained using data generated by a CT scanner (CT scanner 200) configured with a dual-layer detector array 202. In one variation, the neural network is trained using data generated by a spectral CT scanner configured with more than two layers. In yet another variation, a spectral CT scanner is configured with a photon counting (direct conversion) detector, and a neural network is trained with data generated therefrom.
In another variation, the neural network is trained using data generated by a spectral CT scanner configured with multiple X-ray tubes. In this case, the non-spectral projection data is for one of the X-ray tube/detector array pairs. In another variation, the neural network is trained using data generated by a spectral CT scanner configured with kVp switching circuitry. In this case, the non-spectral projection data is for one of the kvps. An example of a system configured with multiple X-ray tubes and/or kVp switching circuits is described in US 8,442,184B 2, filed on 1/6/2009 and entitled "spectra ct," which is incorporated herein by reference in its entirety.
FIG. 4 schematically illustrates a
Fig. 6 schematically illustrates an example of a suitable
In this example, the
Typically, a convolutional layer includes convolutional neurons that apply a convolutional operation to an input. The batch normalization layer normalizes the output of the convolutional layer before processing the output by the rectifying linear unit. The rectifying-linear unit applies a non-linear activation function to its input. The
Examples of convolutional networks are further described in the following: gouk et al, "Fast Sliding Window Classification with volumetric New Networks" (IVNVZ' 14Proceedings of the29th International Conference on Image and Vision Computing New Zealand, page 114-; "full volumetric networks for the management of" (Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015); and Ronneberger et al "U-Net: conversion network for biological Image Segmentation" (Medical Image Computing and Computer-Assisted interaction (MICCAI), Springer, LNCS, volumes 9351: 234-241, 2015).
Fig. 7 illustrates an example method according to embodiments herein.
It should be recognized that the order of the actions in this method is not limiting. Accordingly, other sequences are contemplated herein. In addition, one or more acts may be omitted, and/or one or more additional acts may be included.
At 702, spectral volume image data is generated with a spectral imaging system, as described herein and/or otherwise.
At 704, non-spectral volumetric image data is generated utilizing the same spectral imaging system, as described herein and/or otherwise.
At 706, the spectral volume image data, the non-spectral volume image data, and the system information are employed to train a neural network to predict the spectral volume image data from the non-spectral volume image data.
At 708, a trained neural network is employed to predict spectral volumetric image data from non-spectral volumetric image data from a non-spectral scanner and system information for the non-spectral scanner.
Fig. 8 illustrates an example method according to embodiments herein.
It should be recognized that the order of the actions in this method is not limiting. Accordingly, other sequences are contemplated herein. In addition, one or more acts may be omitted, and/or one or more additional acts may be included.
At 802, spectral volume image data is generated with a spectral imaging system, as described herein and/or otherwise.
At 804, non-spectral projection data is generated using the same spectral imaging system, as described herein and/or otherwise.
At 806, the neural network is trained using the spectral volume image data, the non-spectral projection data, and the system information to predict the spectral volume image data from the non-spectral projection data.
At 808, the trained neural network is employed to predict spectral volumetric image data from non-spectral projection data from the non-spectral scanner and system information of the non-spectral scanner.
The spectral image includes a single type or a plurality of types of spectral images. For example, in one case, the spectral image includes an effective low keV image (e.g., for maximum tissue contrast). In another case, the spectral image includes an effective Z image. In yet another case, the spectral image includes both an effective low keV image and an effective Z image. In yet another case, the spectral image includes one or more other types of spectral images.
Where the spectral image includes multiple types of spectral images, a different neural network may be trained for each of the multiple types. In another case, a combined neural network is trained for at least two different types of spectral images. Multiple and/or combined trained neural networks are employed to predict one or more spectral volumetric image data sets for one or more types of spectral images from non-spectral data. In the above example, this includes predicting an effective low keV image and an effective Z image.
The above may be implemented by computer readable instructions encoded or embedded on a computer readable storage medium, which when executed by a computer processor(s) cause the processor(s) to perform 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 that is not a computer readable storage medium.
While the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive; the invention is not limited to the disclosed embodiments. Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims.
In the claims, the word "comprising" does not exclude other elements or steps, and the word "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 measures 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/wireless telecommunication systems. Any reference signs in the claims shall not be construed as limiting the scope.
- 上一篇:一种医用注射器针头装配设备
- 下一篇:使用心音的慢性血压监测