Systems, methods, and computer program products that facilitate generating enhanced image representations

Abstract

The present application relates to systems, methods and computer program products that facilitate the generation of enhanced image representations. The present invention provides techniques for generating enhanced image representations from raw X-ray images using deep learning techniques. In another embodiment, a system is provided, the system comprising a memory and a processor, the memory storing a computer executable component, and the processor executing the computer executable component stored in the memory. The computer executable component may include a receiving component, an analysis component and an artificial intelligence component. The analysis component uses an AI-based model to analyze the raw X-ray image relative to a set of features of interest. The AI ​​component generates multiple enhanced image representations. Each enhanced image representation highlights a subset of the features of interest and suppresses the remaining features of interest outside the subset in the set.

Description

Technical Field

[0001] The present application generally relates to deep learning techniques, and more particularly to computer-implemented techniques for generating enhanced image representations using deep learning techniques. Background Art

[0002] X-ray imaging is a medical imaging modality widely used for first-line diagnosis and continuous assessment. In some cases, X-ray imaging is also used for detection of disease progression and routine diagnosis. X-ray imaging is a projection-based imaging modality, where each image pixel depicts the sum of information along a given projection depth. Substantially, an X-ray image is a two-dimensional projection of a three-dimensional map, where some of the available information is obscured due to the summation effect. For example, bone tissue provides a higher attenuation to X-rays during imaging, so that bone tissue will mask tissue changes in the same projection path, which correspond to tissues with lower attenuation to X-rays.

[0003] Sometimes, it may be beneficial to provide a medical image depicting a particular feature of interest that is not blocked by other features of interest. For example, early symptoms of coronavirus disease COVID-19 appear as ground-glass opacities (GGOs) in lung tissue. However, the lower lobes of the lung tissue where such GGOs may appear are typically obscured by liver tissue or heart tissue in X-ray images. Computed tomography (CT) imaging can be used to obtain an unobstructed view of the lower lobes of the lung tissue. However, CT imaging is generally more expensive and time-consuming than X-ray imaging. In addition, the radiation dose received by a patient from CT imaging is higher than the radiation dose received from X-ray imaging. Summary of the Invention

[0004] The following presents a summary of the invention to provide a basic understanding of one or more embodiments of the invention. The summary of the invention is not intended to identify key or important elements, nor is it intended to depict any scope of particular embodiments or any scope of the claims. Its sole purpose is to present concepts in a simplified form as a prelude to the more detailed description that is presented later. In one or more embodiments described herein, systems, computer-implemented methods, devices, and / or computer program products are provided that facilitate generating enhanced image representations using deep learning techniques.

[0005] According to one embodiment, a system is provided that includes a memory and a processor, the memory storing computer-executable components, and the processor executing the computer-executable components stored in the memory. The computer-executable components may include a receiving component, an analyzing component, and an artificial intelligence (AI) component. The receiving component receives a raw X-ray image. The analyzing component analyzes the raw X-ray image using an AI-based model with respect to a set of features of interest. The AI component generates a plurality of enhanced image representations. Each enhanced image representation highlights a subset of the features of interest and suppresses the remaining features of interest outside of that subset in the set.

[0006] In some embodiments, the computer-executable components may further include a training component that uses machine learning to train the AI-based model. In some embodiments, the training component utilizes computed tomography (CT) volumes as training data. In some embodiments, the AI-based model includes a first AI subsystem and a second AI subsystem, the first AI subsystem decomposing the raw X-ray image into a set of component images, and the second AI subsystem combining a subset of the component images to generate a reconstructed X-ray image of the raw X-ray image. In some embodiments, an error value is generated based on a comparison between the raw X-ray image and the reconstructed X-ray image. In some embodiments, a confidence score associated with the plurality of enhanced image representations is generated based on the error value.

[0007] In some embodiments, the computer-executable components may further include a presenting component that displays the raw X-ray image and an enhanced image representation from the plurality of enhanced image representations. In some embodiments, the computer-executable components may further include a selection component for selecting a particular feature of interest to selectively mask or generate.

[0008] In some embodiments, the elements described in connection with the disclosed computer-implemented method can be embodied in different forms, such as a computer system, a computer program product, or another form. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Many aspects, embodiments, objects, and advantages of the present invention will become apparent upon consideration of the following detailed description in conjunction with the accompanying drawings, in which like reference numerals represent like components throughout, and in which:

[0010] Figure 1 An exemplary raw X-ray image depicting a human chest region is shown in accordance with one or more embodiments described herein.

[0011] Figure 2 Shown is the use in accordance with one or more embodiments described herein Figure 1An exemplary enhanced image representation generated from an original X-ray image is shown.

[0012] Figure 3 Shown is a block diagram of an example system that facilitates generating an enhanced image representation according to one or more embodiments described herein.

[0013] Figure 4 Shown is a block diagram of another exemplary system that facilitates generating an enhanced image representation according to one or more embodiments described herein.

[0014] Figure 5 An exemplary process for generating a ground truth component image from computed tomography (CT) volume data according to one or more embodiments described herein is shown.

[0015] Figure 6 An exemplary network architecture of an artificial intelligence-based model that facilitates generation of enhanced image representations from raw X-ray images according to one or more embodiments described herein is shown.

[0016] Figure 7 Another exemplary network architecture for an artificial intelligence-based model that facilitates generation of enhanced image representations from raw X-ray images according to one or more embodiments described herein is shown.

[0017] Figure 8 Illustrative image data comparing an enhanced image representation generated from CT volume data with an enhanced image representation generated from an original X-ray image according to one or more embodiments described herein.

[0018] Figure 9 An enhanced image representation corresponding to a spatial location definition according to one or more embodiments described herein is shown.

[0019] Figure 10 Shows a flowchart of an example of a method for generating an enhanced image representation according to one or more embodiments described herein.

[0020] Figure 11 A block diagram of an exemplary, non-limiting operating environment is shown in which one or more embodiments described herein may be facilitated. Specific implementation method

[0021] The following detailed description is merely illustrative and is not intended to limit the implementation scheme and / or the application or use of the implementation scheme. In addition, it is not intended to be bound by any explicit or implicit information set forth in the aforementioned "Background Technology" or "Invention Summary" section or "Detailed Description of the Invention" section.

[0022] ​​The present subject matter discloses systems, computer-implemented methods, apparatuses, and / or computer program products for facilitating the generation of enhanced image representations from raw X-ray images using deep learning techniques. Refer to Figure 1 , which shows an exemplary raw X-ray image 100. Generally, radiography is a projection-based imaging modality, in which a radiation source emits an X-ray beam towards a radiation detector via an imaged medium (e.g., a patient). As the X-ray beam propagates through the medium, a portion of the X-ray beam energy is absorbed by the different materials that make up the medium at different rates depending on the density and composition of the given material.

[0023] The attenuated X-ray beam energy is detected by the radiation detector after passing through the medium. The detected energy generates a signal that represents the intensity of the X-ray beam energy incident on the radiation detector. These signals are processed to generate projection data that includes a line integral of the attenuation coefficients of the medium along the propagation path between the radiation source and the radiation detector. An X-ray image (e.g., X-ray image 100) is formed from such projection data using various reconstruction techniques (e.g., filtered backpropagation). Through such reconstruction techniques, a gray value is assigned to each pixel of the X-ray image, which is proportional to the attenuation coefficient value associated with that pixel in the projection data.

[0024] One aspect of using projection data that includes a line integral of attenuation coefficients to generate an X-ray image is that each pixel of a given X-ray image is a sum of information from multiple depths along the associated propagation path. Representing the depth information of each propagation path in a summative manner introduces occlusion effects into the X-ray image. For example, a material with a higher attenuation coefficient (e.g., bone tissue) will occlude or obscure other materials with a lower attenuation coefficient (e.g., lung tissue) in the same propagation path in the X-ray image. Such occlusion effects translate into subtle variations in the intensity values within regions of the X-ray image that are associated with multiple materials having different attenuation coefficients.

[0025] For example, X-ray image 100 includes multiple image regions that illustrate such occlusion effects. In this example, image region 102 is associated with a propagation path that includes both bone tissue and lung tissue. Image region 104 is associated with a propagation path that includes bone tissue, lung tissue, and liver tissue. Image region 106 is associated with a propagation path that includes bone tissue, lung tissue, and heart tissue. As Figure 1 shown, although there are multiple features of interest within each image region of X-ray image 100, it can be challenging to visually distinguish the individual features of interest due to the above-described occlusion effects. This is especially true in image regions where the variations in intensity values are particularly subtle (such as image region 106).

[0026] To mitigate such occlusion effects, the present disclosure utilizes deep learning techniques to generate enhanced image representations using the original X-ray image, as discussed in more detail below. Figure 2 illustrates the use according to one or more embodiments described herein Figure 1 exemplary enhanced image representations generated from the original X-ray image shown. As Figure 2 shown, each enhanced image representation highlights, emphasizes, or enhances features of interest that are different from the X-ray image 100. For example, the features of interest highlighted in the enhanced image representation 210 include multiple features of interest corresponding to various body tissues outside the lung tissue. The lung tissue is the feature of interest highlighted in the enhanced image representation 220. In the enhanced image representation 230, the bone tissue is the highlighted feature of interest. The epithelial tissue is the feature of interest highlighted in the enhanced image representation 240.

[0027] Figure 3 illustrates a block diagram of an exemplary system 300 for facilitating the generation of enhanced image representations according to one or more embodiments described herein. The system 300 includes an X-ray decomposition component 302 that generates one or more enhanced image representations 355 from the original X-ray image 350. The X-ray decomposition component 302 includes a memory 304 and one or more processors 306 that are used to store computer-executable components, and the processors are operatively coupled to the memory 304 to execute the computer-executable components stored in the memory 304. As Figure 3 shown, the computer-executable components include: a receiving component 308; an analyzing component 310; and an artificial intelligence (AI) component 312.

[0028] The receiving component 308 may receive the original X-ray image 350 for generating one or more enhanced image representations. In one embodiment, the receiving component 308 directly receives the original X-ray image 350 from an imaging device (e.g., an X-ray scanner) that generates, captures, and / or creates the original X-ray image 350. In one embodiment, the receiving component 308 receives the original X-ray image 350 from a remote computing device via a network interface. In one embodiment, the receiving component 308 receives the original X-ray image 350 from a database or data structure accessible to the X-ray decomposition component 302. In one embodiment, the database or data structure resides in the memory 304. In one embodiment, the X-ray decomposition component 302 may access the database or data structure via a network interface.

[0029] The analyzing component 310 may analyze the original X-ray image 350 relative to a set of features of interest using an AI-based model 320. In one embodiment, the AI-based model 320 is implemented using a network architecture 600, which is described below with respect toFigure 6 shown and described in more detail. When analyzing the original X-ray image 350, the analysis component 310 can apply an AI-based model 320 to identify a set of features of interest. For example, the set of features of interest can include bone tissue, epithelial tissue, lung tissue, and various body tissues outside the lung tissue, as described above with respect to Figure 2 discussed. The type of deep learning architecture for the AI-based model 320 can vary. In some embodiments, the AI-based model 320 can employ a convolutional neural network (CNN) architecture. Other suitable deep learning architectures for the AI-based model 320 can include, but are not limited to, recurrent neural networks, recursive neural networks, and classical neural networks.

[0030] The AI component 312 can generate multiple enhanced image representations using a set of component images obtained using the AI-based model 320. Each enhanced image representation highlights a subset of the features of interest and suppresses the remaining features of interest outside of that subset in the set. In one embodiment, the AI component 312 applies a filter that partitions the set of component images into multiple subsets of component images. In one embodiment, the multiple subsets of component images include a first subset of component images associated with the subset of features of interest highlighted by a given enhanced image representation. In one embodiment, the multiple subsets of component images include a second subset of component images associated with the remaining features of interest outside of the subset in the set, and the given enhanced image representation suppresses the second subset of component images. In one embodiment, the AI component 312 enhances the subset of features to generate an enhanced image representation by applying one or more image processing functions to one or more of the component images associated with the subset of features. Examples of suitable image processing functions include: edge enhancement, intensity rescaling, contrast enhancement, noise reduction, image smoothing, contrast stretching, gamma correction, etc. In one embodiment, the AI component 312 enhances the subset of features to generate an enhanced image representation by applying an identity transformation to one or more of the component images associated with the subset of features.

[0031] Figure 4 FIG. shows a block diagram of another exemplary system 400 that facilitates the generation of enhanced image representations according to one or more embodiments described herein. The system 400 includes an X-ray decomposition component 402 that generates one or more enhanced image representations 355 from the original X-ray image 350. The X-ray decomposition component 402 includes a memory 404 and one or more processors 406, the memory for storing computer-executable components, and the processors operatively coupled to the memory 404 to execute the computer-executable components stored in the memory 404. Similar to Figure 3Memory 304, and the computer-executable components stored in memory 404 include: a receiving component 308; an analysis component 310; and an AI component 312. As Figure 4 shown, the computer-executable components stored in memory 404 further include: a training component 408; a rendering component 410; and a selection component 412.

[0032] The training component 408 can use machine learning to train an AI-based model 320. When training the AI-based model 320, the training component 408 retrieves multiple sets of training images from training data 450. Each set of training images includes an X-ray image and a corresponding ground truth (GT) composition image set, where the X-ray image is generated from the training data 450, and the corresponding GT composition image set is generated from the X-ray image generated from the training data 450. The training component 408 can generate such corresponding GT composition images to train the AI-based model 320, as described below with respect to Figure 5 shown and described.

[0033] In one embodiment, the training data 450 includes information from other imaging modalities (e.g., imaging modalities not used to generate, capture, and / or create the original X-ray image 350). Examples of other imaging modalities include: computed tomography (CT) volumes, magnetic resonance imaging (MRI) data, positron emission tomography (PET) data, etc. In one embodiment, the training data 450 includes information from different forms of the same imaging modality (e.g., different forms of the imaging modality used to generate, capture, and / or create the original X-ray image 350), such as multi-energy X-ray imaging data. In one embodiment, the training data 450 includes information from other imaging modalities, information from different forms of the same imaging modality, or a combination thereof.

[0034] Each corresponding GT composition image is defined by a different composition image definition that corresponds to one or more features of interest within the associated X-ray image. In one embodiment, each composition definition associated with a given set of training images is configured such that the recombination of the GT composition image set reconstructs the corresponding X-ray image. Defining each corresponding GT composition image with a different composition image definition provides the training component 408 with a mechanism that facilitates configuring the AI-based model 320 for application-specific decomposition.

[0035] For example, a particular application may involve presenting an unobstructed view of the lung tissue depicted in an X-ray image to a user of the system 400. As described above with respect to Figure 1As discussed, some regions of the lung tissue depicted in the X-ray image (e.g., region 106) may be occluded by other materials having different attenuation coefficient values. Such regions of the lung tissue may correspond to the lower lobe of the lung tissue, where the liver tissue or heart tissue in the X-ray image can typically obscure ground-glass opacities (GGOs). In this example, the training component 408 can utilize a first component definition and a second component definition, the first component definition corresponding to the lung tissue, and the second component definition corresponding to each remaining feature of interest outside the lung tissue (e.g., bone tissue) within the X-ray image. Presenting an unobstructed view of the lung tissue to the user of the system 400 using the first component definition and the second component definition of this example can facilitate early COVID-19 symptom recognition.

[0036] The training component 408 inputs the X-ray image from each retrieved training image set into the AI-based model 320. The AI-based model 320 decomposes each X-ray image into a set of component images and generates a reconstructed X-ray image using the set of component images. In one embodiment, the AI-based model 320 combines the set of component images to generate a reconstructed X-ray image. In one embodiment, the AI-based model 320 combines a subset of the component images to generate a reconstructed X-ray image.

[0037] For each training image set, the training component 408 utilizes a loss function that operates on the decomposed set of component images and the corresponding reconstructed X-ray image. The value obtained by the training component 408 from this loss function is backpropagated to guide the training of the AI-based model 320. In one embodiment, the loss function can be implemented using the loss function defined by Equation 1:

[0038]

[0039] According to Equation 1 above, X ′ i is the i-th decomposed component image, is the ground truth image corresponding to the i-th decomposed component image, μ is a scalar value that weights the reconstruction loss, X is the original X-ray image, X” is the reconstructed X-ray image generated using the decomposed component images, ρ i is a scalar value that weights the i-th regularization term, and (·) is a regularization function that operates on the i-th decomposed component image and the corresponding ground truth component image. In one embodiment, the regularization function (·) uses a structural similarity index measure (SSIM), a total variation loss measure, etc. to quantify the similarity between the i-th decomposed component image and the corresponding ground truth component image.

[0040] The loss function of Equation 1 includes a data fidelity term that promotes self - regularization of the decomposition process implemented by the AI - based model 320. This data fidelity term is represented in Equation 1 by the comparison between the original X - ray image X and the corresponding reconstructed X - ray image X”. This comparison also promotes training the AI - based model 320 to produce meaningful component images when decomposing the input X - ray image. Specifically, this comparison verifies that the set of decomposed component images is not arbitrary because they recombine to substantially form a replica of the original X - ray image.

[0041] In one embodiment, one or more loss functions other than the loss function defined by Equation 1 can be used to implement this loss function. In one embodiment, a combination of loss functions can be used to implement this loss function. In one embodiment, the combination of loss functions includes the loss function defined by Equation 1. In one embodiment, the combination of loss functions does not include the loss function defined by Equation 1.

[0042] In one embodiment, the training component 408 uses this data fidelity term to calculate the error associated with the original X - ray image and the reconstructed X - ray image. In one embodiment, the training component 408 generates a confidence score based on the error calculated using this data fidelity term. In one embodiment, this confidence score is a measure quantifying the quality of the decomposition process. In one embodiment, this confidence score indicates the accuracy associated with one or more enhanced image representations generated using the component images obtained through the decomposition process.

[0043] The loss function of Equation 1 also includes a regularization term that operates on each component image to reduce the variance of the AI - based model 320 while substantially not increasing the bias of the AI - based model 320. In one embodiment, this regularization term is used to mitigate style loss, adversarial loss, or a combination thereof.

[0044] In one embodiment, the training component 408 can train a post - processing AI model to perform image processing / analysis tasks on the enhanced image representation to improve the sensitivity of diagnostic decisions (e.g., early detection of lung diseases in X - ray images). Examples of image processing / analysis tasks include: organ segmentation, anomaly detection, anatomical feature characterization, medical image reconstruction, diagnosis, etc. In one embodiment, this post - processing AI model performs the image processing / analysis task based on the enhanced image representation, the associated original X - ray image, or a combination thereof.

[0045] In one embodiment, the training component 408 receives the training data 450 directly from one or more medical imaging devices (e.g., a computed tomography (CT) scanner). In one embodiment, the training component 408 receives the training data 450 from a remote computing device via a network interface. In one embodiment, the training component 408 receives the training data 450 from a database or data structure accessible to the X-ray decomposition component 402. In one embodiment, the database or data structure resides in the memory 404. In one embodiment, the X-ray decomposition component 402 can access the database or data structure via a network interface.

[0046] The rendering component 410 can generate a display for presentation to a user of the system 400. In one embodiment, the rendering component generates a display that includes the original X-ray image (e.g., the original X-ray image 350) and an enhanced image representation generated using the original X-ray image (e.g., the enhanced image representation 355).

[0047] The selection component 412 is for selecting a particular feature of interest for selective masking or generation. In one embodiment, the selection component 412 generates a user interface element that is configured to receive user input for selecting a particular feature of interest for selective masking or generation. In one embodiment, the selection corresponding to the particular feature of interest for selective masking or generation modifies the operation of the AI component 312. For example, the selection corresponding to the particular feature of interest for selective masking or generation can define a subset of the features of interest as discussed above with respect to the AI component 312. As another example, the selection corresponding to the particular feature of interest for selective masking or generation can define the remaining features of interest outside of this subset in the set as discussed above with respect to the AI component 312. In one embodiment, the selection component 412 forwards the user interface element to the rendering component 410 to be included on the display presented to the user of the system 400. The subset of the features of interest and suppresses the remaining features of interest outside of this subset in the set.

[0048] Figures 3 to 4 Each of the systems shown can be implemented via any type of computing device, such as the computer 1102 described in more detail below with respect to Figure 11 More detailed description. Figures 3 to 4 Each of the systems shown can include a single device or multiple devices that cooperate in a distributed environment. For example, the X-ray decomposition component 302 and / or 402 can be provided via multiple devices arranged in a distributed environment, which together provide the functions described herein. Additionally, other components not shown can also be included within this distributed environment.

[0049] Figure 5An exemplary process 500 for generating a GT composition image highlighting a given feature of interest from CT volume data in accordance with one or more embodiments described herein is shown. In Figure 5 this example, the given feature of interest is lung tissue. However, those skilled in the art will recognize that, in accordance with embodiments of the present disclosure, the composition image may highlight other features of interest or combinations of features of interest. For example, the composition image may highlight heart tissue or all of the patient's anatomical structures located anterior to the coronal plane with respect to the patient.

[0050] At block 510, the training component 408 retrieves CT volume data from the training data 450 corresponding to image data reconstructed from a particular CT scan operation. The retrieved CT volume data includes a plurality of voxels each assigned a radiodensity value. The radiodensity value quantifies the density of the medium (e.g., tissue) measured at a given location by generating the retrieved CT volume data via CT scanning. Each radiodensity value is represented as a dimensionless unit on a scale (i.e., Hounsfield unit (HU)), which linearly transforms the corresponding attenuation coefficient value into a scale where water and air are assigned HU values of 0 HU and -1000 HU, respectively. In one embodiment, the retrieved CT volume data includes corresponding HU values that fall within a range defined by -1024 HU and 1024 HU.

[0051] Each HU value represents the gray value assigned to a given voxel in the CT volume data. Air is typically associated with lower values on the Hounsfield scale (e.g., -1024 HU) and thus appears white in CT image data. Bone tissue is typically associated with higher values on the Hounsfield scale (e.g., 1024 HU) and thus appears black in CT image data. Other media (e.g., muscle tissue and adipose tissue) are associated with intermediate values on the Hounsfield scale (e.g., 50 HU and -100 HU, respectively) and thus appear as various shades of gray in CT image data.

[0052] At block 520, the training component 408 spatially transforms the CT volume data according to view requirements defined for the given feature of interest. Spatially transforming the CT volume data involves aligning the slices of the CT volume data with the expected orientation of the given feature of interest within the enhanced image representation. By way of example, image 525 represents a slice of the CT volume data depicting a feature of interest in lung tissue after spatial transformation of the CT volume data.

[0053] In one embodiment, the compositional image definition of the given feature of interest provides the view requirements for the training component 408 for spatial transformation. In one embodiment, the training component 408 performs spatial transformation using truss angle data or projection angle data associated with the CT volume data. In one embodiment, the training component 408 performs spatial transformation on the CT volume data by applying a transformation matrix that modifies the coordinate values associated with the CT volume data. In this embodiment, the application of the transformation matrix rotates the CT volume data according to the expected orientation of the given feature of interest within the GT compositional image.

[0054] At block 530, the training component 408 performs a masking process on the CT volume data based on the range of HU values defined for the given feature of interest. In one embodiment, the compositional image definition of the given feature of interest provides the range of HU values for the training component 408 for the masking process. The masking process performed by the training component 408 involves evaluating the HU value of each voxel of the CT volume data. Based on this evaluation, the training component 408 retains the HU values that fall within the defined range of HU values and removes or minimizes the HU values that fall outside the defined range of HU values.

[0055] In doing so, the training component 408 retains the voxels of the CT volume data corresponding to the given feature of interest while removing or minimizing the voxels corresponding to other features of interest. In one embodiment, the training component 408 minimizes each HU value that falls outside the defined range of HU values by replacing the HU value with a lower HU value on the Hounsfield scale (e.g., -1000 HU to the HU value assigned to air). Continuing the above example, image 535 represents a slice of the CT volume data depicted in image 525 after the masking process, where the voxels corresponding to the lung tissue feature of interest are retained while the voxels corresponding to other features of interest are minimized.

[0056] At block 540, the training component 408 converts the HU values of the CT volume data to corresponding linear attenuation coefficient values. In one embodiment, the training component 408 uses the CT scan parameters associated with the CT volume data to convert the HU values to corresponding linear attenuation coefficient values. At block 550, the training component 408 applies a projection matrix to the CT volume data. The projection matrix applied by the training component 408 maps each voxel from the three-dimensional space of the CT volume data into a two-dimensional image plane to generate a two-dimensional intensity image. In one embodiment, the projection matrix is an orthogonal projection matrix.

[0057] At block 560, the training quantization component 408 performs a quantization process on the two-dimensional intensity image to generate a GT component image depicting a given region of interest. The quantization process performed by the training component 408 involves converting a continuous range of gray values represented in the two-dimensional intensity image into a discrete set of gray values. Continuing with the above example, image 565 represents the GT component image after the attenuation coefficient conversion at block 540, the application of the projection matrix at block 550, and the masking process at block 560, and this GT component image corresponds to a slice of the CT volume data depicted in image 535. In one embodiment, this discrete set of gray values corresponds to the dynamic range of the GT component image. In one embodiment, the training component 408 stores the GT component image in the training data 450 after performing the quantization process.

[0058] Figure 6 An exemplary network architecture 600 of an AI-based model according to one or more embodiments described herein is shown, and this AI-based model facilitates generating an enhanced image representation from an original X-ray image. As Figure 6 shown, the exemplary network architecture 600 includes a first AI subsystem 610 and a second AI subsystem 620. The first AI subsystem 610 decomposes an input (original) X-ray image X into a set of component images (i.e., component images X’ 1 to component image X′ N ). In one embodiment, each component image in this set of component images is associated with a different component image definition utilized by the training component 408 when training the first AI subsystem 610.

[0059] Each component image in this set of component images corresponds to a different region of interest feature included in the input X-ray image X. In other words, each component image highlights, emphasizes, or enhances a specific region of interest feature included in the input X-ray image X. Thus, each component image in the set of component images output by the first AI subsystem 610 corresponds to Figures 3 to 4 the image data for which the AI component 312 is used to generate one or more enhanced image representations.

[0060] The second AI subsystem 620 reconstructs the input X-ray image X by combining at least one subset of this set of component images to generate a reconstructed X-ray image X″. As discussed above, the comparison between the input (or original) X-ray image and the corresponding reconstructed X-ray image facilitates the self-regularization of the AI-based model, and this corresponding reconstructed X-ray image is generated using the component images associated with the input X-ray image. For example, this comparison between the input X-ray image and the corresponding reconstructed X-ray image provides an input to the data fidelity term of the loss function as discussed above with respect to Figure 4 the training component 408.

[0061] Figure 7 Another exemplary network architecture 700 of an AI-based model according to one or more embodiments described herein is shown, which promotes generating an enhanced image representation from an original X-ray image. As Figure 7 shown, the exemplary network architecture 700 includes a first AI subsystem 710 and a second AI subsystem 720. Similar to Figure 6 the first AI subsystem 610, Figure 7 the first AI subsystem 710 of decomposes an input (or original) X-ray image X into a set of constituent images. Similarly, the second AI subsystem 720 reconstructs the input X-ray image X by combining at least one subset of the set of constituent images to generate a reconstructed X-ray image X″.

[0062] Figure 6 The comparison between Figure 7 shows that the first AI subsystem 710 decomposes the input X-ray image into fewer constituent images than the first AI subsystem 610. This difference shows an aspect of the definition of constituent images within the embodiments of the present disclosure. In one embodiment, there is a correspondence between the multiple constituent image definitions used to decompose the original X-ray image during training of the training component 408 when training the AI subsystem and the multiple constituent images output by the convolutional neural network.

[0063] During the training of Figure 6 the first AI subsystem 610, the training component 408 utilizes at least four constituent image definitions. Thus, the first AI subsystem 610 decomposes the input X-ray image into at least four constituent images (i.e., constituent images X’ 1 to X′ N ). During the training of Figure 7 the first AI subsystem 710, the training component 408 utilizes two constituent image definitions. These two constituent image definitions include: a first constituent image definition associated with lung tissue; and a second constituent image definition associated with each remaining feature of interest outside the lung tissue (e.g., bone tissue) included in the original X-ray image. Thus, the first AI subsystem 710 decomposes the input X-ray image into two constituent images (i.e., constituent images X’ 1 and X 2 ′).

[0064] Figure 8 Exemplary image data according to one or more embodiments described herein is shown, which compares an enhanced image representation 810 generated from CT volume data with an enhanced image representation 840 generated from an original X-ray image. In Figure 8 , the lung tissue is a feature of interest highlighted in each of the enhanced image representations 810 and 840. Using as described above with respect to Figure 5The exemplary process 500 discussed is used to generate an enhanced image representation 810. An enhanced image representation 830 is generated using a composite image obtained by providing an original X-ray image 820 as an input to the AI-based model 320. Figure 8 Enhanced image representation 830 is depicted as highlighting each remaining feature of interest outside of the lung tissue. Enhanced image representation 840 is a difference image formed by subtracting enhanced image representation 830 from original X-ray image 820. As Figure 8 As can be seen, enhanced image representations 810 and 840 exhibit comparable spatial resolution. In some embodiments, an enhanced image representation derived from an original X-ray image (e.g., enhanced image representation 840) may exhibit a reduction in dynamic range relative to an enhanced X-ray image derived from CT volume data.

[0065] Figure 9 illustrates another aspect of composition definitions within an embodiment of the present disclosure. Specifically, each composition definition used by the training component 408 to decompose the raw X-ray image when training the AI-based model can be application-specific. Each composition image or enhanced image representation generated by the AI ​​component 312 based on such composition images discussed above corresponds to a tissue type definition. For example, Figure 2 Enhanced image representations are shown highlighting lung tissue, bone tissue, and epithelial tissue (ie, enhanced image representations 220, 230, and 240, respectively).

[0066] If Figure 9 As shown, the training component 408 can utilize other types of component definitions to decompose the original X-ray image when training the AI-based model. For example, the training component 408 utilizes spatial position definitions to decompose the original X-ray image into component images associated with enhanced image representations 910 and 920 when training the AI-based model. In this example, the AI ​​component 312 utilizes such component images to generate an enhanced image representation 910 and generates an enhanced image representation 920, and the enhanced image representation 910 highlights the features of interest associated with the front area of ​​the original X-ray image, and the enhanced image representation 920 highlights the features of interest associated with the back area of ​​the original X-ray image. Other spatial position definitions can distinguish between the left and right areas of the original X-ray image, the upper and lower areas of the original X-ray image, etc.

[0067] For another example, the training component 408 can utilize component definitions that distinguish natural and extraneous features of interest within the raw X-ray image to decompose the raw X-ray image when training the artificial intelligence-based model. In this example, the first component definition can correspond to an implant or a peripherally inserted central catheter (PICC) line, and the second component definition can correspond to epithelial tissue.

[0068] ​As another example, the training component 408 can utilize compositional definitions that distinguish the latent properties of features of interest within the original X-ray image. Such compositional definitions can facilitate material decomposition processing. In this example, a first compositional definition can correspond to a first matrix material of a multi-material object, and a second compositional definition can correspond to a second matrix material of the multi-material object.

[0069] Figure 10 is a flowchart illustrating an example of a method for generating an enhanced image representation according to one or more embodiments described herein. At block 1002, method 1000 includes receiving, by a receiving component (e.g., Figures 3 to 4 receiving component 308), an original X-ray image. At block 1004, method 1000 includes analyzing, by an analysis component (e.g., Figures 3 to 4 analysis component 310), the original X-ray image using an AI-based model to identify a set of features of interest. At block 1006, method 1000 includes generating, by an AI component (e.g., Figures 3 to 4 AI component 312), a plurality of enhanced image representations. Each enhanced image representation highlights a subset of the features of interest and suppresses the remaining features of interest outside of that subset within the set.

[0070] In one embodiment, method 1000 further includes training, by the training component, the AI-based model using machine learning. In one embodiment, method 1000 further includes the training component utilizing information as training data from other imaging modalities or different forms of the same imaging modality, including: computed tomography (CT) volumes, magnetic resonance imaging (MRI) data, positron emission tomography (PET) data, multi-energy X-ray imaging data, or combinations thereof. In one embodiment, method 1000 further includes the training component using the set of original X-ray images and the corresponding set of enhanced image representations to generate at least one of a diagnosis or a segmentation to improve the sensitivity of diagnostic decision-making. In one embodiment, method 1000 further includes the training component calculating an error associated with the original X-ray image and the reconstructed X-ray image. In one embodiment, method 1000 further includes the training component generating a confidence score associated with the plurality of enhanced image representations.

[0071] In one embodiment, method 1000 further includes the presenting component displaying the original X-ray image and the enhanced X-ray image. In one embodiment, method 1000 further includes the presenting component generating a display that compares the differences between the original X-ray image and the enhanced image representation. In one embodiment, method 1000 further includes the selection component providing a specific feature of interest to selectively mask or generate.

[0072] In one embodiment, method 1000 further includes decomposing the original X-ray image into a set of component images using a first AI subsystem and combining a subset of the component images using a second AI subsystem to generate a reconstructed X-ray image. In one embodiment, an error value is generated based on a comparison between the original X-ray image and the reconstructed X-ray image.

[0073] Figure 11 Non-limiting backgrounds may be provided for various aspects of the subject matter disclosed herein, which are intended to provide a general description of a suitable environment in which various aspects of the subject matter disclosed herein may be implemented. Figure 11 A block diagram of an exemplary, non-limiting operating environment is shown in which one or more embodiments described herein may be facilitated. For the sake of brevity, repeated descriptions of similar elements employed in other embodiments described herein are omitted.

[0074] Reference Figure 11 , a suitable operating environment 1100 for implementing various aspects of the present disclosure may also include a computer 1102. The computer 1102 may also include a processing unit 1104, a system memory 1106, and a system bus 1108. The system bus 1108 couples system components including, but not limited to, the system memory 1106 to the processing unit 1104. The processing unit 1104 may be any of a variety of available processors. Dual microprocessors and other multi-processor architectures may also be used as the processing unit 1104. The system bus 1108 may be any of a variety of types of bus structures, including a memory bus or memory controller, a peripheral bus or external bus, and / or a local bus using various available bus architectures, including but not limited to Industry Standard Architecture (ISA), Micro Channel Architecture (MCA), Extended ISA (EISA), Intelligent Drive Electronics (IDE), VESA Local Bus (VLB), Peripheral Component Interconnect (PCI), Card Bus, Universal Serial Bus (USB), Advanced Graphics Port (AGP), FireWire (IEEE11124), and Small Computer System Interface (SCSI).

[0075] The system memory 1106 may also include volatile memory 1110 and non-volatile memory 1112. The Basic Input / Output System (BIOS) (basic routines that transfer information between elements contained within the computer 1102, such as during startup) is stored in the non-volatile memory 1112. The computer 1102 may also include removable / non-removable, volatile / non-volatile computer storage media. Figure 11Illustrated is, for example, a disk storage device 1114. The disk storage device 1114 may also include, but is not limited to, devices such as disk drives, floppy disk drives, tape drives, Jaz drives, Zip drives, LS-100 drives, flash memory cards, or memory sticks. The disk storage device 1114 may also include a separate storage medium or a storage medium in combination with other storage media. To facilitate connection of the disk storage device 1114 to the system bus 1108, a removable or non-removable interface, such as interface 1116, is typically used. Figure 11 Also depicted is software that acts as a mediator between the user and the basic computer resources described in the appropriate operating environment 1100. Such software may also include, for example, an operating system 1118. The operating system 1118, which may be stored on the disk storage device 1114, is used to control and allocate the resources of the computer 1102.

[0076] System applications 1120 utilize the operating system 1118 for resource management through, for example, program modules 1122 and program data 1124 stored in the system memory 1106 or on the disk storage device 1114. It should be recognized that the present disclosure may be implemented with various operating systems or combinations of operating systems. The user inputs commands or information into the computer 1102 through an input device 1136. The input device 1136 includes, but is not limited to, pointing devices such as a mouse, trackball, stylus, touchpad, keyboard, microphone, joystick, gamepad, satellite dish, scanner, TV tuner card, digital camera, digital video camera, webcam, etc. These and other input devices are connected to the processing unit 1104 via the interface port 1130 through the system bus 1108. The interface port 1130 includes, for example, serial ports, parallel ports, game ports, and universal serial bus (USB). Some of the output devices 1134 use the same type of ports as the input device 1136. Thus, for example, a USB port can be used to provide input to the computer 1102 and output information from the computer 1102 to the output device 1134. An output adapter 1128 is provided to illustrate the presence of some output devices 1134, such as monitors, speakers, and printers, as well as other output devices 1134 that require a special adapter. By way of example and not limitation, the output adapter 1128 includes video and sound cards that provide a connection means between the output device 1134 and the system bus 1108. It should be noted that other devices and / or systems of devices provide both input capabilities and output capabilities, such as a remote computer 1140.

[0077] The computer 1102 can operate in a networked environment using a logical connection to one or more remote computers, such as remote computer 114. The remote computer 1140 can be a computer, server, router, network PC, workstation, microprocessor-based device, peer device, or other common network node, etc., and generally can also include many or all of the elements described with respect to computer 1102. For purposes of brevity, only the memory storage device 1142 is shown for the remote computer 1140. The remote computer 1140 is logically connected to the computer 1102 via a network interface 1138 and then physically connected via a communication link 1132. The network interface 1138 encompasses wired and / or wireless communication networks, such as local area networks (LANs), wide area networks (WANs), cellular networks, etc. LAN technologies include Fiber Distributed Data Interface (FDDI), Copper Distributed Data Interface (CDDI), Ethernet, Token Ring, etc. WAN technologies include, but are not limited to, point-to-point links, circuit-switched networks such as Integrated Services Digital Network (ISDN) and its variants thereon, packet-switched networks, and Digital Subscriber Line (DSL). The communication link 1132 refers to the hardware / software for connecting the network interface 1138 to the system bus 1108. Although the communication link 1132 is shown within the computer 1102 for clarity, the communication link can also be external to the computer 1102. For purposes of illustration only, the hardware / software for connecting to the network interface 1138 can also include internal and external technologies, such as modems, including conventional telephone-grade modems, cable modems, and DSL modems, ISDN adapters, and Ethernet cards.

[0078] One or more embodiments described herein may be a system, a method, an apparatus, and / or a computer program product at any possible technical detail level of integration. The computer program product may include a computer-readable storage medium (or multiple computer-readable storage media) having computer-readable program instructions thereon for causing a processor to perform aspects of one or more embodiments. The computer-readable storage medium may be a tangible device that can retain and store instructions for use by an instruction execution device. The computer-readable storage medium may be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of the computer-readable storage medium may also include the following: a portable computer floppy disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as a punch card or raised structures in grooves having instructions recorded thereon, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium should not be construed to be a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., an optical pulse through an optical fiber cable) or an electrical signal transmitted through a wire. In this regard, in various embodiments, the computer-readable storage medium as used herein may include non-transitory and tangible computer-readable storage media.

[0079] The computer-readable program instructions described herein can be downloaded to a corresponding computing / processing device from a computer-readable storage medium or downloaded to an external computer or external storage device via a network (e.g., the Internet, a local area network, a wide area network, and / or a wireless network). The network can include copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium within the corresponding computing / processing device. The computer-readable program instructions for performing the operations of one or more implementations can be assembler instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state-setting data, configuration data for integrated circuits, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages such as Smalltalk, C++, etc., and procedural programming languages such as the "C" programming language or similar programming languages. The computer-readable program instructions can be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network (including a local area network (LAN) or a wide area network (WAN)), or can establish a connection with an external computer (e.g., through the Internet using an Internet service provider). In some implementations, an electronic circuit, including, for example, a programmable logic circuit, a field-programmable gate array (FPGA), or a programmable logic array (PLA), can execute the computer-readable program instructions by utilizing the state information of the computer-readable program instructions to personalize the electronic circuit in order to perform aspects of one or more implementations.

[0080] Aspects of one or more embodiments are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions. These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions executed via the processor of the computer or other programmable data processing apparatus create a means for implementing the functions / acts specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that can direct a computer, a programmable data processing apparatus, and other devices to function in a particular manner, such that the computer-readable storage medium having instructions stored therein comprises an article of manufacture including instructions for implementing aspects of the functions / acts specified in one or more blocks of the flowchart and / or block diagram. The computer-readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus, or other device to produce a computer-implemented process, such that the instructions executed on the computer, other programmable apparatus, or other device implement the functions / acts specified in one or more blocks of the flowchart and block diagram.

[0081] The flowchart 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 described herein. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which includes one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions noted in the blocks may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It should also be noted that each block of the block diagrams and flowchart illustrations, and combinations of blocks in the block diagrams and flowchart illustrations, can be implemented by a system based on dedicated hardware that performs a particular function or act, or combinations of dedicated hardware and computer instructions.

[0082] While the subject matter has been described above in the general context of computer-executable instructions of a computer program product that runs on one or more computers, those skilled in the art will recognize that the present disclosure may also be implemented, or may be implemented in conjunction with, other program modules. Generally, program modules include routines, programs, components, data structures, etc. that perform particular tasks or implement particular abstract data types. In addition, those skilled in the art will understand that the computer-implemented methods of the present invention may be practiced with other computer system configurations, which include single-processor or multi-processor computer systems, small computing devices, mainframe computers, and computers, hand-held computing devices (e.g., PDAs, telephones), microprocessor-based or programmable consumer or industrial electronic devices, and the like. The illustrated aspects may also be practiced in a distributed computing environment where tasks are performed by remote processing devices linked through a communication network. However, some (if not all) aspects of the present disclosure may be practiced on a stand-alone computer. In a distributed computing environment, program modules may be located in local and remote memory storage devices. For example, in one or more embodiments, computer-executable components may be executed from a memory that may include one or more distributed memory units or may be constituted by one or more distributed memory units. As used herein, the terms "memory" and "memory unit" may be used interchangeably. In addition, the code of the computer-executable components in one or more embodiments described herein may be executed in a distributed manner, e.g., multiple processors may combine or cooperate to execute code from one or more distributed memory units. As used herein, the term "memory" may cover a single memory or memory unit at one location, or multiple memories or memory units at one or more locations.

[0083] As used in this application, terms such as "component", "system", "platform", "interface", etc. can refer to and can include computer-related entities or entities related to an operating machine with one or more specific functions. Entities disclosed herein can be hardware, a combination of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, an executing thread, a program, and a computer. By way of illustration, an application program running on a server and the server can both be components. One or more components can reside within a process or an executing thread, and a component can be located on one computer and / or distributed between two or more computers. In another example, corresponding components can execute according to various computer-readable media storing various data structures. Components can communicate via local and / or remote processes, such as according to a signal having one or more data packets (e.g., data from one component that interacts with another component in a local system, a distributed system, and / or a network (such as the Internet via a signal with other systems)). As another example, a component can be a device having a specific function provided by a mechanical part operated by an electrical or electronic circuit, where the electrical or electronic circuit is operated by a software or firmware application executed by a processor. In this case, the processor can be inside or outside the device and can execute at least a part of the software or firmware application. As yet another example, a component can be a device that provides a specific function through an electronic component rather than a mechanical part, where the electronic component can include a processor or other devices for executing software or firmware that at least partially imparts functionality to the electronic component. In one aspect, a component can, for example, emulate an electronic component via a virtual machine within a cloud computing system.

[0084] As used herein, the term "facilitate" is in the context of a system, device, or component that "facilitates" one or more actions or operations, with respect to the nature of a complex computing environment in which multiple components and / or multiple devices can participate in some computing operations. Non-limiting examples of actions that may or may not involve multiple components and / or multiple devices include sending or receiving data, establishing a connection between devices, determining intermediate results for obtaining a result (e.g., including using ML and / or AI techniques to determine intermediate results), etc. In this regard, a computing device or component can facilitate an operation by participating in any part of the operation to be completed. Thus, when an operation of a component is described herein, it should be understood that in cases where the operation is described as being facilitated by a component, the operation can optionally be completed through the cooperation of one or more other computing devices or components, such as but not limited to: sensors, antennas, audio and / or visual output devices, other devices, etc.

[0085] In addition, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or". That is, unless otherwise specified or clear from the context, "X employs A or B" is intended to mean any natural inclusive permutation. That is, if X employs A; X employs B; or X employs both A and B, then "X employs A or B" is satisfied in any of the foregoing instances. Further, the articles "a" and "an" as used in this specification and the appended drawings are generally to be construed to mean "one or more" unless otherwise specified or clear from the context to be in the singular form. As used herein, the terms "example" and / or "exemplary" are used to mean serving as an example, instance, or illustration. To avoid doubt, the subject matter disclosed herein is not limited by such examples. Further, any aspect or design described herein as "example" and / or "exemplary" is not necessarily to be construed as more preferred or advantageous than other aspects or designs, nor does it imply exclusion of equivalent exemplary structures and techniques known to those of ordinary skill in the art.

[0086] As used in this specification, the term "processor" can generally refer to any computing processing unit or device, including but not limited to a single-core processor; a single processor with software multithreading execution capabilities; a multi-core processor; a multi-core processor with software multithreading execution capabilities; a multi-core processor with hardware multithreading technology; a parallel platform; and a parallel platform with distributed shared memory. Additionally, a processor can refer to an integrated circuit, an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a field-programmable gate array (FPGA), a programmable logic controller (PLC), a complex programmable logic device (CPLD), discrete gate or transistor logic components, discrete hardware components, or any combination thereof designed to perform the functions described herein. Further, a processor can utilize nanoscale architectures (such as but not limited to molecule- and quantum-dot-based transistors, switches, and gates) in order to optimize space usage or enhance the performance of a user device. A processor can also be implemented as a combination of computing processing units. In this disclosure, terms such as "storage", "storage device", "data storage", "data storage device", "database", and substantially any other information storage component related to the operation and functionality of a component are used to refer to a "memory component", an entity embodied in a "memory", or a component that includes a memory. It should be recognized that the memory and / or memory components described herein can be volatile memory or non-volatile memory, or can include both volatile and non-volatile memory. By way of illustration and not limitation, non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM), flash memory, or non-volatile random access memory (RAM) (e.g., ferroelectric RAM (FeRAM)). For example, volatile memory can include RAM, which can serve as an external cache memory. By way of illustration and not limitation, RAM can be provided in various forms, such as synchronous RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), direct Rambus RAM (DRRAM), direct Rambus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM). Additionally, the memory components of the systems or computer-implemented methods disclosed herein are intended to include but not be limited to including these and any other suitable types of memory.

[0087] The foregoing description includes only examples of systems and computer-implemented methods. Of course, it is not possible to describe every conceivable combination of components or computer-implemented methods for the purposes of describing the present disclosure, but one of ordinary skill in the art will recognize that many further combinations and permutations of the present disclosure are possible. In addition, to the extent that the terms "including", "having", "possessing", etc. are used in the detailed description, the claims, the appendices, and the drawings, such terms are intended to be inclusive in a manner similar to the term "comprising", as "comprising" is interpreted when used as a transitional word in the claims.

[0088] The description of the various embodiments has been presented for purposes of illustration, but is not intended to be exhaustive or limited to the disclosed embodiments. Many modifications and variations will be apparent to one of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terms used herein were chosen to best explain the principles of the embodiments, the practical application, or technical improvement over technologies found in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A system for facilitating generation of an enhanced image representation, the system comprising: a memory storing computer executable components; and a processor that executes the computer executable components stored in the memory, wherein the computer executable components include: A receiving component, wherein the receiving component receives an original X-ray image; an analysis component that uses an artificial intelligence-based model to analyze the raw X-ray image with respect to a feature set of interest and generates a set of component images of the raw X-ray image corresponding to the feature set of interest; and An artificial intelligence component that generates a plurality of enhanced image representations by dividing the set of constituent images into a first subset associated with features of interest that are to be highlighted and a second subset associated with features of interest that are to be suppressed for the corresponding enhanced images, and applying an image processing function to the first subset and the second subset of constituent images to respectively highlight and suppress the features of interest in the corresponding enhanced images.

2. The system of claim 1, further comprising a training component that employs machine learning to train the artificial intelligence-based model using a set of training images retrieved from training data.

3. The system of claim 2, wherein the training component utilizes information as training data derived from other imaging modalities or different forms of the same imaging modality, including: Computed tomography volumes, magnetic resonance imaging data, positron emission tomography data, multi-energy X-ray imaging data, or a combination thereof.

4. The system of claim 1, further comprising a presentation component that displays the original X-ray image and an enhanced image representation from the plurality of enhanced image representations.

5. The system of claim 2, wherein the training component is configured to train a post-processing artificial intelligence model to perform a diagnostic task or an organ segmentation task based on the original X-ray image and the enhanced image representation.

6. The system of claim 5, wherein the diagnostic task or the organ segmentation task comprises detection of lung disease.

7. The system of claim 1 , wherein the artificial intelligence-based model comprises a first artificial intelligence subsystem that decomposes the original X-ray image into a set of constituent images and a second artificial intelligence subsystem that combines subsets of the constituent images to generate a reconstructed X-ray image of the original X-ray image, wherein an error value is generated based on a comparison between the original X-ray image and the reconstructed X-ray image.

8. The system of claim 7, wherein confidence scores associated with the plurality of enhanced image representations are generated based on the error values.

9. The system of claim 2, wherein the training component is configured to obtain values ​​from a loss function, which are back-propagated to guide the training of the artificial intelligence-based model.

10. A method for facilitating generation of an enhanced image representation, the method comprising: The receiving component receives the original X-ray image; Analyzing the raw X-ray image using an artificial intelligence based model by an analysis component to identify a set of features of interest; Generating, by the analysis component, a set of component images of the original X-ray image corresponding to the set of features of interest; dividing, by the artificial intelligence component, the set of constituent images into a first subset associated with features of interest to be highlighted and a second subset associated with features of interest to be suppressed for the corresponding enhanced image; applying, by the artificial intelligence component, an image processing function to the first subset and the second subset of constituent images to respectively highlight and suppress features of interest in the corresponding enhanced images; as well as A plurality of enhanced image representations are generated by the artificial intelligence component from the first subset and the second subset of constituent images.

11. The method according to claim 10, further comprising: Machine learning is used by the training component to train the artificial intelligence based model using a set of training images retrieved from the training data.

12. The method according to claim 11, further comprising: Information is utilized by the training component as training data originating from other imaging modalities or different forms of the same imaging modality, including: computed tomography volumes, magnetic resonance imaging data, positron emission tomography data, multi-energy X-ray imaging data, or combinations thereof.

13. The method according to claim 11, further comprising: The post-processing artificial intelligence model is trained by the training component to perform a diagnostic task or an organ segmentation task based on the original X-ray image and the enhanced image representation.

14. The method according to claim 10, further comprising: A display is generated by a presentation component that compares a difference between the original X-ray image and an enhanced image representation from the plurality of enhanced image representations.

15. The method according to claim 10, further comprising: decomposing the raw X-ray image into the set of component images using a first artificial intelligence subsystem; and combining a subset of the component images using a second artificial intelligence subsystem to generate a reconstructed X-ray image, Wherein an error value is generated based on a comparison between the original X-ray image and the reconstructed X-ray image.

16. The method according to claim 15, further comprising: Confidence scores associated with the plurality of enhanced image representations are generated using the error values.

17. The method according to claim 11, further comprising: Values ​​are obtained from the loss function by the training component and are back-propagated to guide the training of the artificial intelligence-based model.

18. A computer program product that facilitates generating an enhanced image representation, the computer program product comprising a non-transitory computer readable medium having program instructions embodied therewith, the program instructions executable by a processing component to cause the processing component to: receiving a raw X-ray image; analyzing the raw X-ray image using an artificial intelligence-based model to identify a set of features of interest; generating a set of component images of the original X-ray image corresponding to the set of features of interest; dividing the set of constituent images into a first subset associated with features of interest to be highlighted and a second subset associated with features of interest to be suppressed for the corresponding enhanced image; applying an image processing function to the first subset and the second subset of constituent images to respectively highlight and suppress features of interest in the corresponding enhanced images; as well as A plurality of enhanced image representations are generated from the first subset and the second subset of constituent images.

Images