Prediction of effective energy value

Abstract

The proposed concept aims to provide schemes, solutions, concepts, designs, methods, and systems related to training neural network models for predicting the effective energy values ​​of conventional CT images. In particular, embodiments aim to utilize single-energy CT images from spectral scanners, where the effective energy values ​​are known, in order to train a neural network model that can be used to predict the effective energy values ​​of conventional CT images, i.e., CT images that are not typically single-energy.

Description

[Technical Field]

[0001] The present invention relates to the field of predicting the effective energy values ​​of conventional CT images, and more particularly to the field of training a neural network model for predicting the effective energy values ​​of conventional CT images. [Background technology]

[0002] Dual-energy CT imaging is well-established. The simplest approach is to acquire two separate scans taken at two different kVp settings, i.e., at two different photon energies. Material resolution can be achieved in the projection region or image region in this case. Typically, there is a small but not negligible movement between the two scans due to the time delay between them. Alignment of low-dataset and high-dataset images is necessary and can be best performed in the image region. After image alignment, image-based material resolution of the CT images can be applied using the spatially matched images. Image-based material resolution of CT images is also well-established in the art.

[0003] In the theory of material resolution in image regions, the linear decay coefficient μ of each pixel in low- and high-CT images is considered. H,L (x,y,z)≡μ H,L This is approximated by the pixel values ​​in the image of the underlying material, i.e., a linear combination of the photoelectric effect (ph) and the Compton effect or scattering (sc).

number

number

number

[0004] However, one of the main challenges in image-based material degradation is finding the effective energy of the CT image in order to calculate a proper transfer function from the conventional input CT image to the material image. In clinical practice, the cumbersome system calibration procedure required to find the effective energy should be avoided because it slows down the workflow and is susceptible to drift.

[0005] Therefore, a solution is needed to predict the effective energy value of conventional CT images. [Means for solving the problem]

[0006] The present invention is defined by the claims.

[0007] According to an embodiment of one aspect of the present invention, a method for training a neural network model for predicting the effective energy value of conventional CT images is provided.

[0008] This method involves training a neural network model using a training algorithm configured to receive an array of training inputs and their respective known outputs, where the training inputs consist of spectral scanner single-energy CT images and the respective known outputs consist of the effective energy values ​​of the spectral scanner single-energy CT images.

[0009] Therefore, the proposed concept aims to provide schemes, solutions, concepts, designs, methods, and systems related to training neural network models for predicting the effective energy values ​​of conventional CT images.

[0010] In particular, the embodiment aims to utilize single-energy CT images from a spectral scanner, where the effective energy value is known, in order to train a neural network model that can be used to predict the effective energy value of conventional CT images, i.e., CT images that are typically not single-energy. In other words, the neural network model is trained using single-energy images from a spectral scanner, but the trained neural network model can be used to predict the effective energy value of conventional CT images obtained from either a spectral scanner or a conventional scanner. By training a neural network model to predict the effective energy value of a CT image using only the image, a method for estimating the effective energy of a CT image can be provided that does not require system calibration and additional system or image information. Actual kVp setting information of the scan is also not required in this case, and calibration procedures during clinical practice can be avoided.

[0011] In other words, it has been proposed that the neural network model can be trained using a training algorithm configured to receive single-energy CT images from a spectral scanner as training input and the effective energy values ​​of the single-energy CT images as known outputs. This could ultimately provide a neural network model capable of estimating the effective energy values ​​of conventional (i.e., non-single-energy) CT images solely from the images, thus enabling the calculation of a transfer function for the images and allowing material decomposition. This method also avoids the calibration procedure of CT systems, which slows down the workflow and is frequently affected by drift. Implementing this purely data-driven solution improves the ease of avoiding calibration and obtaining estimated effective energy values ​​of conventional CT images.

[0012] By training a neural network model to predict the effective energy value from a single input CT image, a method for estimating the effective energy value of a conventional CT image purely from the image alone may ultimately be provided.

[0013] The use of single-energy CT images from spectral scanners with known effective energy values ​​to train neural network models enables these models to become proficient in estimating effective energy values ​​even from non-single-energy CT images, i.e., conventional CT images. In other words, it has been realized that single-energy CT images obtainable from spectral scanners can be used to train neural network models to predict effective energy values ​​even from conventional CT images.

[0014] Ultimately, a method is provided to train a neural network model to predict the effective energy value of conventional CT images solely from the image itself, allowing the effective energy value of CT images to be estimated without additional or contextual information. The effective energy value of CT images can be utilized for future purposes, for example, to calculate transfer functions for material decomposition.

[0015] In some embodiments, the array of training inputs includes spectral scanner monoenergetic CT images from 40 keV to 140 keV with a step size of 0.1 keV or more. This is within the typical range of the effective energy of conventional CT images and allows the model to estimate the effective energy of CT images with an accuracy of 0.1 keV.

[0016] In some embodiments, the spectral scanner monoenergetic CT images are, in particular, the entire single-axis CT images.

[0017] In some embodiments, the array of training inputs further includes adjusted spectral scanner monoenergetic CT images, and the adjustments include slice inversion, rotation, shift, zoom, cropping, and use of different noise levels. This increases the amount of training data on which the neural network is trained and also allows the training to be invariant to changing image characteristics due to reconstruction parameters or anatomical variations.

[0018] In some embodiments, the neural network model includes a convolutional neural network (CNN) or a machine learning regression model or a classification model. CNNs have proven particularly successful in the analysis of images and are known to be able to classify images with much lower error rates than other types of neural networks. Regression models are known to be particularly useful for analyzing the relationship between independent features, such as pixels of a CT image, and a dependent variable, such as effective energy. Classification models are particularly useful for assigning data to discrete classes based on a particular set of features.

[0019] In some embodiments, the array of training inputs further includes corresponding aggregated feature vectors for each spectral scanner monoenergetic CT image. Using the corresponding aggregated feature vectors together with the monoenergetic CT images has been found to improve the performance of the neural network model.

[0020] In some embodiments, the aggregated feature vectors are preprocessed using principal component analysis (PCA). It has been found that focusing on the most important feature vector components can further improve the training performance of the neural network model.

[0021] In some embodiments, the aggregated feature vector includes a histogram of Hounsfield unit (HU) values ​​within each spectral scanner single-energy CT image. It has been found that specific use of the HU value histogram within the CT images during training can significantly improve the performance of the neural network model.

[0022] Furthermore, a method for estimating the effective energy value of a conventional CT image is provided, the method comprising the steps of acquiring a first CT image and processing the first CT image using a neural network model trained according to any of the methods disclosed herein to generate a first estimated effective energy value of the first CT image.

[0023] After training according to the method described above, the neural network model may be able to actually process CT images and generate an estimated effective energy value for the processed CT images.

[0024] In some embodiments, a conventional method for estimating the effective energy value of a CT image further includes a step of using the first estimated effective energy value in a further image processing step of the first CT image, such as bone beam hardening correction. Once the estimated effective energy value of the CT image is generated, it can be used to improve the image processing step. For example, the estimated effective energy value of the CT image can be used in bone beam hardening correction of the CT image.

[0025] In some embodiments, a method for estimating the effective energy value of a conventional CT image further comprises the steps of dividing a first CT image into a plurality of CT subimages and processing the plurality of CT subimages using a neural network model to generate an estimated effective energy value for each CT subimage. This may take into account the fact that the effective energy value may vary within a conventional CT image. For example, a CT image containing strong beam hardening artifacts is more likely to produce an image in which the effective energy value varies within the image. The estimated effective energy value for each subimage can then be used to compute a different local transfer function for each subimage, which can then be utilized to apply material resolution to each subimage using each local transfer function, respectively.

[0026] In some embodiments, a conventional method for estimating the effective energy value of a CT image further includes the steps of: acquiring a second CT image; processing the second CT image using a neural network model to generate a second estimated effective energy value for the second CT image; and calculating a transfer function based on the first estimated effective energy value of the first CT image and the second estimated effective energy value of the second CT image. This may be particularly useful in the case of dual-energy CT imaging, where two CT images, one low-energy and one high-energy, are generated in each scan. Once the effective energy values ​​are estimated for both CT images, the transfer function can be calculated based on the estimated effective energy values.

[0027] In some embodiments, the above method further includes the step of performing material decomposition based on a first CT image and a second CT image using a calculated transfer function. Spectral imaging enabled by material decomposition is an important application of CT images.

[0028] In some embodiments, a conventional method for estimating the effective energy value of a CT image, which includes the step of acquiring a second CT image, further includes the steps of acquiring at least a third CT image, processing at least the third CT image using a neural network model to generate a third estimated effective energy value for at least the third CT image, calculating a transfer function based on the third estimated effective energy value for at least the third CT image, and performing material decomposition based on the first, second, and third CT images using the calculated transfer function. This may be particularly useful in the case of high-energy material decomposition, for example, in a photon-counting CT system, where more than two images are generated for each scan.

[0029] Also provided are computer programs having coding means for implementing any of the methods disclosed herein when executed on a processing system.

[0030] Furthermore, a system is provided for training a neural network model to predict the effective energy values ​​of conventional CT images. This system has a processor configuration configured to train a neural network model using a training algorithm configured to receive an array of training inputs and their respective known outputs, the training inputs comprising spectral scanner single-energy CT images, and the respective known outputs comprising the effective energy values ​​of spectral scanner single-energy CT images.

[0031] Furthermore, a system for estimating the effective energy value of a conventional CT image is provided, which has a processor configuration configured to use a neural network model trained according to any of the methods of claims 1 to 7 to process (330) the first CT image and generate a first estimated effective energy value of the first CT image.

[0032] Therefore, a concept may be proposed for training a neural network model to predict the effective energy value of conventional CT images.

[0033] These and other aspects of the present invention will become apparent from and be explained with reference to the embodiments described below.

[0034] For a better understanding of the present invention and to more clearly illustrate how the present invention is carried out, the accompanying drawings are referenced merely as examples. [Brief explanation of the drawing]

[0035] [Figure 1] This is a schematic diagram of an imaging system such as a CT scanner. [Figure 2A] This is a simplified flowchart of a method for training a neural network model to predict the effective energy value of conventional CT images. [Figure 2B] This is a more detailed explanation of the method shown in Figure 2A. [Figure 3] This is a flowchart illustrating a conventional method for estimating the effective energy value of CT images. [Figure 4] This is a flowchart illustrating a method for performing material decomposition based on multiple CT images. [Figure 5] This is a simplified block diagram of a system for training a neural network model to predict the effective energy values ​​of conventional CT images. [Figure 6] Examples of computers in which one or more parts of the embodiment are employed are shown. [Modes for carrying out the invention]

[0036] The present invention will be described with reference to the drawings.

[0037] The detailed descriptions and specific examples illustrate exemplary embodiments of the apparatus, systems, and methods, but should be understood to be for illustrative purposes only and not intended to limit the scope of the invention. These and other features, aspects, and advantages of the apparatus, systems, and methods of the invention will be better understood from the following description, the appended claims, and the appended drawings. The drawings are for illustrative purposes only and are not drawn to a specific scale. Also, the same reference numerals are used throughout the drawings to indicate the same or similar parts.

[0038] Modifications of the disclosed embodiments can be understood and implemented by those skilled in the art in carrying out the claimed invention, based on a review of the drawings, disclosures, and appended claims. In the claims, the words “comprising” do not exclude other components or steps, and the indefinite articles “a” or “an” do not exclude plurality.

[0039] Please understand that the drawings are merely schematic and are not drawn to a specific scale. Also, please understand that the same reference number is used throughout the drawings to indicate the same or similar parts.

[0040] Implementations in accordance with this disclosure relate to various techniques, methods, schemes, and / or solutions for training neural network models for predicting the effective energy values ​​of conventional CT images. According to the proposed concept, several solutions can be implemented individually or together. That is, these possible solutions may be described separately below, but two or more of these possible solutions may be implemented in one combination or another.

[0041] Embodiments of the present invention aim to utilize single-energy CT images from a spectral scanner with known effective energy values ​​to train a neural network model that can be used to predict the effective energy values ​​of conventional (i.e., non-single-energy) CT images. By training a neural network model to predict the effective energy values ​​of CT images using only the images, a method for estimating the effective energy of a CT image without requiring system calibration and additional system or image information may be provided. Actual kVp setting information of the scan is also not required. In this case, calibration procedures during clinical practice can be avoided.

[0042] Therefore, the proposed concept aims to provide schemes, solutions, concepts, designs, methods, and systems related to training neural network models for predicting the effective energy values ​​of conventional CT images.

[0043] Figure 1 schematically shows an imaging system 100, such as a CT scanner configured for spectral (multi-energy) imaging, used to acquire training data. The imaging system 100 generally includes a fixed gantry 102 and a rotatable gantry 104 that is rotatably supported by the fixed gantry and rotates around the examination area 106 about the z-axis. A sofa-like object support 120 supports an object or subject within the examination area.

[0044] A radiation source 108, such as an X-ray tube, is rotatably supported by a rotatable gantry 104 and rotates with the rotatable gantry, emitting radiation across the inspection area 106. In one example, the radiation source includes a single broad-spectrum X-ray tube. In another example, the radiation source includes a single X-ray tube configured to switch between at least two different emission voltages (e.g., 80 kVp and 140 kVp) during a scan. In yet another example, the radiation source includes two or more X-ray tubes configured to emit radiation with different average spectra. In yet another example, the radiation source includes a combination thereof.

[0045] The radiation sensing detector array 110 corresponds to an arc across the inspection area 106 at an angle opposite to the radiation source 108. The radiation sensing detector array detects radiation passing through the inspection area and generates an electrical signal (projection data) indicating it. If the radiation source includes a single broad-spectrum X-ray tube, the radiation sensing detector array includes energy-resolved detectors (e.g., a direct-conversion photon count detector, at least two sets of scintillators (multilayer) with different spectral sensitivities). In kVp switching and multi-tube configurations, the detector array may include a single-layer detector, a direct-conversion photon count detector, and / or a multilayer detector. The direct-conversion photon count detector may include a conversion material such as CdTe, CdZnTe, Si, Ge, GaAs, or other direct-conversion materials. An example of a multilayer detector includes a two-stage detector, such as the two-stage detector described in U.S. Patent No. 7,968,853 B2.

[0046] The reconstruction device 118 receives spectral projection data from the detector array 110 and reconstructs spectral volumetric image data such as sCCTA image data, high-energy images, low-energy images, photoelectric images, Compton scattering images, iodine images, calcium images, virtual non-contrast images, bone images, soft tissue images, and / or other basic material images. The reconstruction device can also reconstruct non-spectral volumetric image data by combining spectral projection data and / or spectral volumetric image data, for example. Generally, spectral projection data and / or spectral volumetric image data include data from at least two different energies and / or energy ranges.

[0047] Console 112 functions as an operator console. The console is operablely connected to an output device 116, such as a monitor, and an input device 114, such as a keyboard or mouse. Software residing in console 112 allows the operator to interact with and / or operate the imaging system 100 via a graphical user interface (GUI) or otherwise.

[0048] Referring here to Figures 2A and 2B, a flowchart of the method 200 for training a neural network model 250 for predicting the effective energy values ​​of conventional CT images, according to the proposed embodiment, is shown.

[0049] This method comprises step 210 of training a neural network model 250 using a training algorithm 240 configured to receive a training input 220 and an array of known outputs 230, where the training input comprises a spectral scanner single-energy CT image and each known output comprises the effective energy value of the spectral scanner single-energy CT image. The effective energy of a CT image depends on the attenuation characteristics of the object being scanned and therefore varies from CT image to CT image, even for the same patient; for example, an axial scan of a patient including the leg will have a different effective energy than an axial scan of the leg. The trained neural network model can process and estimate the effective energy values ​​of CT images even from a different CT system than the one from which the training data was acquired.

[0050] In some embodiments, the neural network model 250 includes a convolutional neural network (CNN) or a machine learning regression or classification model. CNNs have proven particularly successful in image analysis and are known to be able to classify images with a much lower error rate than other types of neural networks. Regression models have been found particularly useful for analyzing the relationship between independent features, e.g., pixels in a CT image, and dependent variables, e.g., effective energy. Classification models are particularly useful for assigning data, e.g., effective energy values, to discrete classes based on a particular set of features. Examples of CNN models that may be used include DenseNet, ResNet-like, or EfficientNet. A machine learning regression model that may be used is a random forest.

[0051] The structure of an artificial neural network (or simply a neural network) is inspired by the human brain. A neural network consists of layers, each layer containing multiple neurons. Each neuron contains a mathematical operation. In particular, each neuron may contain different weighted combinations of a single type of transformation (e.g., the same type of transformation, such as a sigmoid, but with different weights). In the process of processing input data, the mathematical operation of each neuron is performed on the input data to produce a numerical output, and the output of each layer in the neural network is sequentially fed to the next layer, which provides the final output.

[0052] Several types of layers exist, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs). This exemplary embodiment of the present invention employs a CNN-based learning algorithm because CNNs have proven particularly successful in video analysis and can classify frames in a video with a much lower error rate than other types of neural networks.

[0053] A CNN typically contains multiple layers, including convolutional layers, pooling layers, and fully connected layers. Convolutional layers consist of a set of learnable filters that extract features from the input. Pooling layers are a type of nonlinear downsampling that reduces data size by combining the outputs of multiple neurons in one layer with a single neuron in the next layer. Fully connected layers connect each neuron in one layer to all neurons in the next layer.

[0054] The methods for training machine learning algorithms are well known. Typically, such methods involve obtaining a training dataset containing training input data entries and corresponding training output data entries. The initialized machine learning algorithm is applied to each input data entry to generate predicted output data entries. The error between the predicted output data entries and the corresponding training output data entries is used to refine the machine learning algorithm. This process can be repeated until the errors converge and the predicted output data entries are sufficiently similar to the training output data entries (e.g., ±1%). This is generally known as supervised learning.

[0055] For example, the weighting of the mathematical operations of each neuron can be modified until the error converges. Known methods for modifying neural networks include gradient descent and backpropagation algorithms.

[0056] The training input data entries 220 for the CNN that may be used in method 200 correspond to single-energy CT images of a spectral scanner. The training output data entries 230 correspond to the effective energy values ​​of each spectral scanner single-energy CT image. Furthermore, several preprocessing methods may be employed to improve the training samples. In other words, the first CNN can be trained using a training algorithm 240 configured to receive a training input and an array of each known output, where the training input includes single-energy CT images of a spectral scanner, and each known output includes the effective energy value of the single-energy CT image of the spectral scanner. In this way, the CNN is trained to output the effective energy value of the single-energy CT image of a spectral scanner, given a single-energy CT image of a spectral scanner.

[0057] Single-energy CT images from a spectral scanner can be easily created using a spectral CT system by generating single-energy images corresponding to various single-energy levels. Note that actual human patient data is required, not just phantom data.

[0058] In some embodiments, the array of training inputs 220 includes spectral scanner single-energy CT images ranging from 40 keV to 140 keV with a step size of 0.1 keV or larger. This is a typical range of effective energy for conventional CT images, allowing the model to estimate the effective energy of the CT images with an accuracy of 0.1 keV.

[0059] In some embodiments, the spectral scanner single-energy CT image is, in particular, the entire single-axis CT image.

[0060] In some embodiments, the training input array 220 further includes single-energy CT images from a tuned spectral scanner, where the tune-up includes slice inversion, rotation, shift, zoom, cropping, and the use of different noise levels. For example, a low-pass filter may be applied to the single-energy CT images. This increases the amount of training data on which the neural network model 250 is trained, and also allows the training to be invariant to image characteristics that change due to reconstruction parameters or anatomical changes. In other words, providing multiple versions of CT images at a given target energy makes the training somewhat independent. This technique is also called data augmentation.

[0061] In some embodiments, the array of training inputs 220 further includes corresponding aggregated feature vectors of each spectral scanner single-energy CT image. Using the corresponding aggregated feature vectors with the single-energy CT images has been shown to improve the performance of the neural network model 250. For example, the corresponding aggregated feature vectors of each spectral scanner's single-energy CT image may be used in a multilayer perceptron (MLP), random decision forest, or adaptive boosting during training. In some embodiments, the aggregated feature vectors may be sufficient on their own, essentially replacing the original CT images.

[0062] In some embodiments, the aggregated feature vectors are preprocessed using principal component analysis (PCA). It has been found that focusing on the most important feature vector components can further improve the training performance of the neural network model 250.

[0063] In some embodiments, the aggregated feature vector includes a histogram of Hounsfield unit (HU) values ​​within each spectral scanner single-energy CT image. The specific use of the histogram of HU values ​​within the CT images during training has been shown to significantly improve the performance of the neural network model 250.

[0064] Referring here to Figure 3, a diagram of a method 300 for estimating the effective energy value of a conventional CT image is provided.

[0065] In step 310, a first CT image is obtained. The first CT image does not need to be single-energy and may be any conventional CT image. The conventional CT image may include scattering correction and beam hardening correction.

[0066] In step 320, the first CT image is segmented into multiple CT subimages. This allows for the fact that the effective energy can vary within the first CT image. For example, a CT image containing strong beam hardening artifacts is more likely to produce an image where the effective energy varies within the image. However, step 320 is optional, and this method may proceed with the entire single first CT image, i.e., without performing step 320. In step 330, the multiple CT subimages (or the entire single first CT image) are processed using the neural network trained in method 200 to generate an estimated effective energy value for each CT subimage (or generate a single estimated effective energy value for the entire first CT image). In some cases, each subimage may have the same estimated effective energy value, in which case the single effective energy value may be considered the value for the entire image.

[0067] In step 340, the estimated effective energy value of the first CT image (or sub-image) may be used in further image processing steps of the first CT image. For example, the estimated effective energy value of the CT image may be used for bone beam hardening correction of the CT image.

[0068] Referring here to Figure 4, a diagram of method 400 for performing material decomposition based on multiple (i.e., at least two) CT images is provided.

[0069] In step 310, the first CT image is obtained. In step 410, the second CT image is obtained. In step 420, at least the third CT image is obtained.

[0070] In step 430, the first, second, and at least third CT images are processed using the neural network trained in method 200 so as to generate the first, second, and third effective energy values ​​for the first, second, and at least third CT images, respectively.

[0071] In step 440, the transfer function is calculated based on the first, second, and at least third estimated effective energy values of the first, second, and at least third CT images, respectively. Estimating the effective energy values of two CT images and calculating the transfer function based on both estimated effective energy values can be particularly useful in the case of dual-energy CT imaging, where two conventional CT images are generated in each scan, a low-energy scan and a high-energy scan. Examples of dual-energy CT systems are high-speed kVp switching systems, two-layer detector systems, and dual-source systems. The addition of at least a third CT image, and the subsequent estimation of its effective energy value and use of that value in the calculation of the transfer function can be particularly useful in the case of multi-energy material decomposition, where more than two images are generated for each scan, for example in a photon-counting CT system.

[0072] In step 450, material decomposition based on the first CT image, the second CT image, and at least the third CT image is performed using the transfer function calculated in step 440. If only two CT images are obtained, the material decomposition is performed based on only the first and second CT images using the transfer function calculated using the first and second estimated effective energy values of the first and second CT images, respectively. Spectral imaging enabled by material decomposition is an important application of CT images.

[0073] More specifically, for each energy channel e, the conventional CT image has to be reconstructed, and the effective energy E for each energy channel, i.e., for each CT image e has to be estimated, for example, using any of the methods disclosed herein. If the number of materials in the CT image (e.g., 2) is less than the number of energy channels (e.g., 3), the forward model can be given as follows.

Equation

[0074] Equation (4) is the pseudo-inverse A TA or weighted pseudo-inverse A T C -1 A can be inverted using C, where C is the covariance matrix between input CT images. A weighted pseudo-inverse function is one embodiment of the above transfer function.

[0075] Taking the example of acquiring and processing two CT images, the effective energy value can be estimated for each image pair at a given z position. The matrix given in Equation 3 can be formulated, inverted, and applied to an image pair to calculate the material image ph and sc for a given z position.

[0076] Referring now to Figure 5, a simplified block diagram of a system 500 for training a neural network model to predict the effective energy values ​​of conventional CT images is shown according to one embodiment. The system configured for training a neural network model to predict the effective energy values ​​of conventional CT images has a processor configuration 510.

[0077] The processor configuration 510 is configured to train a neural network model using a training algorithm configured to receive a training input and an array of known outputs, wherein the training input comprises a spectral scanner single-energy CT image, and each known output comprises the effective energy value of the spectral scanner single-energy CT image.

[0078] Furthermore, referring to Figure 5, a simplified block diagram of system 520 for estimating the effective energy value of conventional CT images using a neural network model is also shown. This system uses a neural network model trained to predict the effective energy value of conventional CT images according to the present invention. System 520 has a processor configuration 530.

[0079] Referring now to Figure 6, an example of a computer 600 in which one or more parts of the embodiment may be employed is shown. The various operations described above may utilize the functions of computer 600. In this regard, it should be understood that the system function blocks may run on a single computer or may be distributed across several computers and locations (e.g., connected via the Internet).

[0080] Computer 600 includes, but is not limited to, PCs, workstations, laptops, PDAs, palm devices, servers, and storage devices. Generally, from a hardware architecture standpoint, computer 600 may include one or more processors 610, memory 620, and one or more I / O devices 630 that are communicatively coupled via a local interface (not shown). The local interface may be, for example, one or more buses or other wired or wireless connections as known in the art, but is not limited to. The local interface may have additional elements such as controllers, buffers (caches), drivers, repeaters, and receivers to enable communication. Furthermore, the local interface may include address, control, and / or data connections to enable appropriate communication between the aforementioned components.

[0081] The processor 610 is a hardware device for executing software that can be stored in the memory 620. The processor 610 can be virtually any custom-made or commercially available processor, a central processing unit (CPU), a digital signal processor (DSP), or an auxiliary processor among several processors associated with the computer 600, and the processor 610 may be a semiconductor-based microprocessor (in the form of a microchip) or a microprocessor.

[0082] Memory 620 may include one or a combination of volatile memory elements (such as dynamic random access memory (DRAM), static random access memory (SRAM), random access memory (RAM), and non-volatile memory elements (such as ROM, erasable programmable read-only memory (EPROM), electronically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), tape, compact disc read-only memory (CD-ROM), disk, floppy disk, cartridge, cassette, etc.). Furthermore, memory 620 may incorporate electrical, magnetic, optical, and / or other types of storage media. It should be noted that memory 620 may have a distributed architecture in which various components are located apart from each other but can be accessed by processor 610.

[0083] The software in memory 620 may include one or more separate programs, each program including an ordered list of executable instructions for implementing a logical function. The software in memory 620 includes, in exemplary embodiments, a suitable operating system (O / S) 650, a compiler 660, source code 670, and one or more applications 680. As shown, application 680 has multiple functional components for performing the features and operations of the exemplary embodiment. Application 680 of computer 600 may represent various applications, computing units, logic, functional units, processes, operations, virtual entities and / or modules, in exemplary embodiments, but application 680 is not limited to these.

[0084] The operating system 650 controls the execution of other computer programs and provides scheduling, input / output control, file and data management, memory management, communication control, and related services. It is intended by the inventors that application 680 for carrying out exemplary embodiments may be applicable to all commercially available operating systems.

[0085] Application 680 may be a source program, an executable program (object code), a script, or any other entity including a set of instructions to be executed. If it is a source program, it is usually translated to work properly in relation to O / S 650 via a compiler (such as compiler 660), assembler, interpreter, etc., which may or may not be contained in memory 620. Furthermore, Application 680 can be written as an object-oriented programming language, which has classes of data and methods, or a procedural programming language, which has routines, subroutines, functions, and such languages ​​as C, C++, C#, Pascal, BASIC, API calls, HTML, XHTML, XML, ASP scripts, JavaScript, FORTRAN, COBOL, Perl, Java, ADA, .NET, etc.

[0086] The I / O device 630 may include, but is not limited to, input devices such as a mouse, keyboard, scanner, microphone, or camera. Furthermore, the I / O device 630 may include, but is not limited to, output devices such as a printer or display. Finally, the I / O device 630 may further include, but is not limited to, devices that communicate both inputs and outputs, such as a network interface card or modulator / demodulator (for accessing remote devices, other files, devices, systems, or networks), radio frequency (RF) or other transceivers, telephone interfaces, bridges, routers, etc. The I / O device 630 also includes components for communicating over various networks such as the Internet or an intranet.

[0087] If computer 600 is a PC, workstation, intelligent device, etc., the software in memory 620 may also include a Basic Input / Output System (BIOS) (omitted for simplification). The BIOS is a set of essential software routines that initialize and test the hardware at startup, boot the OS 650, and support data transfer between hardware devices. The BIOS is stored in some type of read-only memory, such as ROM, PROM, EPROM, EEPROM, etc., so that the BIOS can be executed when computer 800 is started.

[0088] When computer 600 is operating, processor 610 is configured to execute software stored in memory 620, communicate data with memory 620, and generally control the operation of computer 600 according to the software. Application 680 and O / S 650 are read, either whole or in part, by processor 610, buffered within processor 610, and then executed.

[0089] If Application 680 is implemented in software, it should be noted that Application 680 may be virtually stored on any computer-readable medium for use by or in connection with any computer-related system or method. In the context of this specification, the computer-readable medium may be an electronic, magnetic, optical or other physical device or means capable of housing or storing a computer program for use by or in connection with a computer-related system or method.

[0090] Application 680 may be implemented on any computer-readable medium intended for use by or associated with an instruction execution system, apparatus, or device, such as a computer-based system, a system including a processor, or other system capable of fetching instructions from and executing instructions from an instruction execution system, apparatus, or device. In the context of this document, “computer-readable medium” can be any means by which a program can be stored, communicated, propagated, or transferred for use by or associated with an instruction execution system, apparatus, or device. A computer-readable medium may, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, devices, or propagation media.

[0091] The methods shown in Figures 1 to 3 and the system shown in Figure 4 may be implemented in hardware, software, or a combination of both (for example, as firmware running on a hardware device). As long as one embodiment is partially or entirely implemented in software, the functional steps shown in the process flowchart may be executed by one or more appropriately programmed physical computing devices such as a central processing unit (CPU) or graphics processing unit (GPU). Each process, and its individual component steps shown in the flowchart, may be executed by the same or different computing devices. According to one embodiment, a computer-readable storage medium stores a computer program containing computer program code configured to cause one or more physical computing devices to perform the encoding or decoding method described above when the program is executed on one or more physical computing devices.

[0092] The storage medium may include volatile and non-volatile computer memory such as RAM, PROM, EPROM, and EEPROM, optical discs (such as CDs, DVDs, and BDs), and magnetic storage media (such as hard disks and tapes). The various storage media may be mounted within a computing device or may be transportable so that one or more programs stored on the storage medium are loaded into a processor.

[0093] As long as one embodiment is implemented in hardware, either partially or entirely, the blocks shown in the block diagram of Figure 4 may be separate physical components, logical subdivisions of a single physical component, or all may be implemented as an integrated single physical component. The functionality of one block shown in the drawing may be divided into multiple components in the implementation, or the functionality of multiple blocks shown in the drawing may be combined into a single component in the implementation. Hardware components suitable for use in embodiments of the present invention include, but are not limited to, conventional microprocessors, application-specific integrated circuits (ASICs), and field-programmable gate arrays (FPGAs). One or more blocks may be implemented as a combination of dedicated hardware for performing some functions and one or more programmed microprocessors and associated circuits for performing other functions.

[0094] Modifications of the disclosed embodiments can be understood and implemented by those skilled in the art in carrying out the claimed invention, based on a review of the drawings, disclosures, and appended claims. In the claims, the words “comprising” do not exclude other components or steps, and the indefinite articles “a” or “an” do not exclude plurality. A single processor or other unit may fulfill the functions of several items enumerated in the claims. The mere fact that certain means are described in different dependent claims does not imply that combinations of these means cannot be used advantageously. Where a computer program is mentioned above, the computer program may be stored / distributed on a suitable medium, such as an optical or 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 telecommunications systems. Where the term “adapted” is used in a claim or specification, the term “adapted” is meant to be equivalent to the term “configured.” No reference numeral in a claim should be construed as limiting in scope.

[0095] The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or part of an instruction having one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions shown in a block may occur out of the order shown in the figure. For example, two blocks shown consecutively may actually be executed substantially simultaneously, or blocks may sometimes be executed in reverse order depending on the functions they contain. It should also be noted that each block in a block diagram and / or flowchart, as well as combinations of blocks in a block diagram and / or flowchart, can be implemented by a purpose-specific hardware-based system that performs a specified function or operation, or a combination of purpose-specific hardware and computer instructions. The flowcharts and block diagrams illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or part of an instruction having one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions shown in a block may occur outside the order shown in the diagram. For example, two consecutively shown blocks may actually be executed substantially simultaneously, or blocks may sometimes be executed in reverse order depending on the functions they contain. It should also be noted that each block in a block diagram and / or flowchart, as well as any combination of blocks in a block diagram and / or flowchart, can be implemented by a purpose-specific hardware-based system that performs a specified function or operation, or a combination of purpose-specific hardware and computer instructions.

Claims

1. In conventional methods for training neural network models to predict the effective energy values ​​of CT images, A step of training a neural network model using a training algorithm configured to receive a training input and an array of known outputs, wherein the training input comprises a single-energy CT image from a spectral scanner, and each known output comprises the effective energy value of the single-energy CT image from the spectral scanner. A method having.

2. The method according to claim 1, wherein the array of training inputs includes spectral scanner single-energy CT images at 40 keV to 140 keV with a step size of 0.1 keV or larger.

3. The method according to claim 1 or 2, wherein the array of training inputs further comprises adjusted spectral scanner single-energy CT images, the adjustments include slice inversion, rotation, shift, zoom, cropping, and the use of different noise levels.

4. The method according to any one of claims 1 to 3, wherein the neural network model includes a convolutional neural network, a machine learning regression model, or a classification model.

5. The method according to any one of claims 2 to 4, wherein the array of training inputs further includes corresponding aggregated feature vectors of each spectral scanner single-energy CT image.

6. The method according to claim 5, wherein the aggregated feature vectors are preprocessed using principal component analysis (PCA).

7. The method according to claim 5 or 6, wherein the aggregated feature vector includes a histogram of HU values ​​in each spectral scanner single-energy CT image.

8. A conventional method for estimating the effective energy value of CT images, The first step is to acquire a CT image, The steps of processing the first CT image using the neural network model trained by any of the methods of claims 1 to 7 to generate a first estimated effective energy value of the first CT image, A method that possesses this.

9. A further image processing step of the first CT image, such as bone beam hardening correction, using a first estimated effective energy value, The method according to claim 8, further comprising:

10. The steps include segmenting the first CT image into a plurality of CT subimages, The steps include processing the plurality of CT subimages using the neural network model to generate an estimated effective energy value for each CT subimage, The method according to claim 8 or 9, further comprising:

11. The steps include acquiring a second CT image, The steps include processing the second CT image using the neural network model to generate a second estimated effective energy value for the second CT image, The steps include: calculating a transfer function based on the first estimated effective energy value of the first CT image and the second estimated effective energy value of the second CT image; The method according to any one of claims 8 to 10, further comprising:

12. Using the calculated transfer function, perform material decomposition based on the first CT image and the second CT image. The method according to claim 11, further comprising:

13. A computer program that, when executed on a processing system, includes coding means for implementing the method described in any one of claims 1 to 12.

14. In a system for training a neural network model to predict the effective energy value of conventional CT images, the system is: A neural network model is trained using a training algorithm configured to receive a training input and an array of known outputs, wherein the training input includes a single-energy CT image from a spectral scanner, and each known output includes the effective energy value of the said single-energy CT image from the spectral scanner. Processor configuration configured as follows, A system that has

15. In conventional systems for estimating the effective energy value of CT images, The first CT image was obtained, The first CT image is processed by To generate a first estimated effective energy value of the first CT image, the neural network model trained by any of the methods described in any of claims 1 to 7 is used. Processor configuration configured as follows, A system that has

Images