Optimization of CT image formation in simulated X-rays

Abstract

Projection data obtained by scanning the subject is retrieved. The retrieved projection data is processed to reconstruct a three-dimensional image. A two-dimensional image is generated based on the reconstructed three-dimensional image. An uncertainty in the identification of the disease or uncertainty is identified in the two-dimensional image. The projection data or three-dimensional image is reprocessed into an updated three-dimensional image, with at least one parameter of the reprocessing being based on the uncertainty in the identification of the disease or uncertainty. An updated two-dimensional image is then generated based on the updated three-dimensional image.

Description

[Technical Field]

[0001] The present invention relates generally to systems and methods for simulating conventional X-ray images from computed tomography (CT) data, and more particularly to optimizing the simulated X-ray images based on disease or condition classification using adjustable parameters in a model of the imaging chain. [Background technology]

[0002] CT imaging offers advantages over conventional planar X-ray (CXR) imaging. Thus, CT imaging has replaced X-ray imaging in a variety of clinical settings and is increasingly being adopted in addition to, and in place of, such X-ray imaging.

[0003] This is particularly true for low-dose and ultra-low-dose CT imaging (ULDCT), which is currently aimed at replacing CXR in additional settings such as routine chest imaging in the outpatient setting.

[0004] One of the main advantages of CT imaging over CXR is that CT provides additional information, particularly three-dimensional spatial information. CXR also has a relatively low sensitivity and high false-negative rate in many clinical scenarios. Because of the additional information associated with CT imaging, it is also better suited to various image processing and artificial intelligence (AI)-based diagnostic techniques.

[0005] On the other hand, CXR has higher spatial resolution than conventional CT images, especially ULDCT images, and is less susceptible to noise.

[0006] One reason CT imaging has not been more widely adopted in routine clinical situations is that CT image times are substantially higher than CXR, in part because radiologists are more familiar with CXR images and therefore more comfortable interpreting and diagnosing such images than conventional planar x-ray images. Summary of the Invention [Problem to be solved by the invention]

[0007] However, as radiologists continue to rely on CXR images, they are forgetting the advantages of CT imaging, both in terms of additional information and analytical capabilities.

[0008] Therefore, to incorporate some of the advantages associated with CT images into synthetic CXR images, it is known to create digitally reconstructed radiograph (DRR) images. Such DRR images can be constructed by digitally tracing x-rays through a 3D CT volume. In clinical settings, such reconstructions can be used by radiologists in a variety of settings where CXR images are traditionally used.

[0009] Classification models for diseases, conditions, disorders, etc. (hereafter collectively referred to as "disease classification models") are known in the context of CXR images. However, such models cannot be applied to the CT images themselves, and therefore, when image processing is applied to the CT data before generating the DDR, information related to disease classification cannot be leveraged to optimize the reconstruction.

[0010] There is a need for CT imaging systems and methods, particularly ULDCT imaging systems and methods, that can present data to radiologists in a format that is more easily interpreted and more likely to be adopted. Additionally, there is a need for systems that can utilize disease classification to optimize image reconstruction. [Means for solving the problem]

[0011] Systems and methods are provided for converting three-dimensional CT data into two-dimensional images, including, for example, retrieving projection data acquired by scanning a subject from multiple angles about a central axis.

[0012] In some embodiments, a computer-implemented method for processing medical CT data includes retrieving projection data acquired by scanning a subject.

[0013] The method then includes processing the projection data to reconstruct a three-dimensional image. Once processed, the method generates a two-dimensional image based on the three-dimensional image and identifies the disease or uncertainty in identifying the disease in the two-dimensional image.

[0014] The method then reprocesses the projection data or the three-dimensional image into an updated three-dimensional image, where at least one parameter of the reprocessing is based on the uncertainty in the identification of the disease or uncertainty. The method then generates an updated two-dimensional image based on the updated three-dimensional image.

[0015] In some embodiments, generating the two-dimensional image is performed by a neural network.

[0016] In some embodiments, the reprocessing of the projection data or the three-dimensional images is performed by a neural network.

[0017] In some embodiments, the three-dimensional image is denoised using a first value of at least one parameter. In some such embodiments, the updated three-dimensional image is denoised using a second value of the at least one parameter. The second value is based on an uncertainty in the identification of the disease or uncertainty.

[0018] In some embodiments, applying artificial intelligence (AI)-based super-resolution is applied to the three-dimensional image using a first value for at least one parameter such that the three-dimensional image results in a high-resolution image. In some such embodiments, applying AI-based super-resolution is applied to the updated three-dimensional image using a second value for the at least one parameter such that the updated three-dimensional image results in another high-resolution image. The second value is based on an uncertainty in the identification of the disease or uncertainty.

[0019] In some embodiments, the 3D image and / or the updated 3D image are denoised by a trained convolutional neural network (CNN).

[0020] In some embodiments, the two-dimensional image is generated by the DRR network using a first value of the at least one parameter. In some such embodiments, an updated two-dimensional image is generated by using the DRR network using a second value of the at least one parameter. The second value is based on an uncertainty in the identification of the disease or uncertainty.

[0021] In some embodiments, at least one physical element is identified in the three-dimensional image and / or the updated three-dimensional image. The at least one physical element is removed or masked out prior to generating the output image. In some such embodiments, the at least one physical element is a plurality of ribs or a heart. In some such embodiments, the at least one physical element is removed or masked out based on a disease or uncertainty in identifying a disease in the two-dimensional image.

[0022] In some embodiments, the disease is identified through the use of a trained disease classification model.

[0023] In some embodiments, a system for processing medical CT data comprises a memory storing a plurality of instructions; and a processor coupled to the memory and configured to execute the plurality of instructions to retrieve projection data acquired by scanning a subject, process the projection data to reconstruct a three-dimensional image; generate a two-dimensional image based on the three-dimensional image; identify an uncertainty in the identification of a disease or uncertainty in the two-dimensional image; re-process the projection data or the three-dimensional image into an updated three-dimensional image, wherein at least one parameter of the re-processing is based on the uncertainty in the identification of the disease or uncertainty; and generate an updated two-dimensional image based on the updated three-dimensional image.

[0024] In some embodiments, the generation of the two-dimensional images and the reprocessing of the projection data or three-dimensional images is performed by a neural network.

[0025] In some embodiments, the three-dimensional image is denoised using a first value of at least one parameter. In some such embodiments, the updated three-dimensional image is denoised using a second value of the at least one parameter. The second value is based on an uncertainty in the identification of the disease or uncertainty.

[0026] In some embodiments, the 3D image and / or the updated 3D image are denoised by a trained convolutional neural network (CNN).

[0027] In some embodiments, the two-dimensional image is generated by the DRR network using a first value of the at least one parameter. In some such embodiments, an updated two-dimensional image is generated by the DRR network using a second value of the at least one parameter. The second value is based on an uncertainty in the identification of the disease or uncertainty.

[0028] In some embodiments, AI-based super-resolution using a first value for at least one parameter is applied to the three-dimensional image such that the three-dimensional image results in a high-resolution image. In some such embodiments, AI-based super-resolution using a second value for the at least one parameter is applied to the updated three-dimensional image such that the updated three-dimensional image results in another high-resolution image. The second value is based on an uncertainty in the identification of the disease or uncertainty. [Brief explanation of the drawings]

[0029] [Figure 1] 1 is a schematic diagram of a system according to one embodiment of the present invention. [Figure 2] 1 illustrates an exemplary imaging device according to one embodiment of the present invention. [Figure 3] 1 illustrates an imaging chain for use in a method according to one embodiment of the present invention. [Figure 4] 1 shows a schematic diagram of a model structure for implementing the method of one embodiment of the present invention; [Figure 5] 1 is a flowchart illustrating a method according to one embodiment of the present invention. [Figure 6] 1 illustrates schematically a ray tracing process applied to a three-dimensional image that can be used in the context of an embodiment of the present invention; [Figure 7] 1 is a flowchart illustrating a method for training an imaging chain according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0030] The description of illustrative embodiments according to the principles of the present invention is intended to be read in connection with the accompanying drawings, which are to be considered part of the entire written description. In describing the embodiments of the present invention disclosed herein, any reference to direction or orientation is intended for convenience of description only and is in no way intended to limit the scope of the invention. Relative meanings such as "bottom," "top," "horizontal," "vertical," "up," "bottom," "above," "bottom," "up," "top," "bottom," and derivatives thereof (e.g., "horizontal," "bottom," "above," etc.) should be interpreted to refer to the orientation as then described or as shown in the drawings under discussion. These relative meanings are for convenience of description only and do not require that the device be constructed or operated in a particular orientation unless expressly so indicated. "Mounted," "attached," "connected," "coupled," "interconnected," and similar terms refer to a relationship in which structures are fixed or attached to one another, both directly or indirectly through intervening structures, and through movable or rigid attachments or relationships, unless expressly stated otherwise. Furthermore, the features and advantages of the present invention are illustrated by reference to the illustrated embodiments. Accordingly, the present invention should not be limited to such exemplary embodiments, which illustrate some possible non-limiting combinations of features that may exist alone or in combination of features, and the scope of the present invention is defined by the claims appended hereto.

[0031] The present invention describes the best mode of carrying out the invention currently contemplated. This description is not intended to be understood in a limiting sense, but provides examples of the invention presented solely for illustrative purposes, with reference to the accompanying drawings, to advise those skilled in the art of the advantages and structure of the present invention. In the various views of the drawings, like reference numerals indicate like or similar parts.

[0032] It is important to note that the disclosed embodiments are merely examples of the many advantageous uses of the innovative teachings herein. In general, statements made in the specification of this application do not necessarily limit any of the various claimed embodiments. Moreover, some statements may apply to some inventive features, but not to others. In general, unless otherwise specified, singular elements may be in the plural and vice versa, without loss of generality.

[0033] Both computed tomography (CT) and conventional planar X-ray (CXR) are used in medical imaging. However, CT imaging, particularly ultra-low-dose CT imaging (ULDCT), aims to replace CXR in many clinical settings, such as chest imaging in routine outpatient settings.

[0034] Some of the key advantages of ULDCT imaging are immediately apparent. CT images, including ULDCT, provide three-dimensional spatial information that enables advanced analytical techniques. Furthermore, ULDCT avoids the relatively low sensitivity and high false-negative rate associated with CXR in many clinical scenarios. However, ULDCT has a slower reading time than CXR, and radiologists are less familiar with and comfortable with ULDCT. Therefore, radiologists prefer to be presented with more familiar CXR images and make diagnoses based on them. Accordingly, embodiments of the present invention provide a workflow for generating artificial CXR images, or images stylized to have the appearance of CXR images, from ULDCT data. Such embodiments may be implemented or enhanced using artificial intelligence (AI) techniques, including the use of learning algorithms in the form of neural networks, such as convolutional neural networks (CNNs).

[0035] Thus, a method is provided for converting three-dimensional CT data into two-dimensional images. In this manner, CXR-style images can be generated from ULDCT data and presented to a radiologist. Such presentations can be in either raw image format or three-dimensional image format, subject to implementation of analytical techniques on the underlying ULDCT data, and presented to the radiologist as a surrogate for a CXR image or in the context of a corresponding ULDCT-based image interface.

[0036] Thus, ULDCT image data may be generated as 3D CT image data using a system such as that shown in FIG. 1 and by an imaging device such as that shown in FIG. 2, and the extracted data may then be processed using the processing device of the system of FIG. 1.

[0037] Once the ULDCT imaging data is generated, the CT imaging data can be processed into a three-dimensional image by passing it through an imaging chain. Such an imaging chain can include noise reduction, super-resolution processing, and DRR simulation to convert the three-dimensional image into a two-dimensional image. The imaging chain can include numerous parameters that can be adjusted to affect the resulting DRR image. Iterating through an infinite number of possible images to identify which image is optimal for a given case is practically infeasible. Additionally, it is not always clear what image is best in a given situation.

[0038] In the setting of evaluating medical images, this is particularly relevant, as different radiologists may have different opinions about which aspects of the images they find relevant. Moreover, such expertise is not always readily available. The problem of "expensive" medical expertise is well known in the use of AI in disease classification.

[0039] Thus, one embodiment of the present invention uses an AI model that links the imaging chain to a pre-trained disease classification model. As a result, the effect of modifying adjustable parameters in the imaging chain can be evaluated and optimized. Thus, the pre-trained disease classification model can evaluate a first attempt to convert a 3D CT image into a 2D image and classify the image accordingly. The method can then modify adjustable parameters in the imaging chain and reprocess the projection data or 3D image using the adjusted parameters associated with the identified disease.

[0040] 1 is a schematic diagram of a system 100 according to one embodiment of the present invention. As shown, the system 100 typically includes a processing device 110 and an imaging device 120.

[0041] Processing device 110 may apply processing routines to image or measurement data, such as projection data, received from imaging device 120. Processing device 110 may include memory 113 and processor circuitry 111. Memory 113 may store a plurality of instructions. Processor circuitry 111 may be coupled to memory 113 and configured to execute the instructions. The instructions stored in memory 113 may comprise processing routines and data associated with the processing routines, such as machine learning algorithms, and various filters for processing images.

[0042] Processing device 110 may further include an input 115 and an output 117. Input 115 may receive information, such as 3D images or projection data, from imaging device 120. Output 117 may output information, such as a filtered image or a transformed two-dimensional image, to a user or user interface device. The output may include a monitor or display.

[0043] In some embodiments, processing device 110 may be directly associated with imaging device 120. In alternative embodiments, processing device 110 may be separate from imaging device 120, such that processing device 110 receives image or measurement data for processing at input 115 via a network or other interface.

[0044] In some embodiments, the imaging device 120 may include an image data processing device, i.e., a spectral or conventional CT scanning unit, for generating projection data when scanning a subject (e.g., a patient). In some embodiments, the imaging device 120 may be a conventional CT scanning unit configured to generate helical scans.

[0045] 2 illustrates an exemplary imaging device 200 in accordance with one embodiment of the present invention. Although a CT imaging device 200 is shown and the following description is generally in the context of CT images, it will be understood that similar methods may be applied in the context of other imaging devices, and that images to which these methods may be applied may be acquired in a wide variety of ways.

[0046] In the imaging device 200 according to an embodiment of the present invention, the CT scanning unit may be adapted to perform one or more axial and / or helical scans of the subject to generate CT projection data. In the imaging device 200 according to an embodiment of the present invention, the CT scanning unit may include an energy-resolved photon counting and / or spectral dual-layer image detector. Spectral content may also be acquired using other detector setups. The CT scanning unit may include a radiation source that emits radiation to traverse the subject when acquiring the projection data.

[0047] 2, a CT scanning unit 200, e.g., a CT scanner, may include a stationary gantry 202 and a rotating gantry 204 that may be rotatably supported by the stationary gantry 202. The rotating gantry 204 may rotate about a longitudinal axis around an examination region 206 of a subject (patient) when acquiring projection data. The CT scanning unit 200 may include a table or support 207 for supporting the patient in the examination region 206 and may be configured to move the patient through the examination region during the imaging process.

[0048] The CT scanning unit 200 may include a radiation source 208, such as an X-ray tube, that may be supported by and configured to rotate with a rotating gantry 204. The radiation source 208 may include an anode and a cathode. A power supply voltage applied between the anode and the cathode may accelerate electrons from the cathode to the anode. A current flow may be provided from the cathode to the anode such that the electron flow generates radiation to traverse the examination region 206.

[0049] The CT scanning unit 200 may include a detector 210. The detector 210 may subtend an angled arc on an opposite side of the examination region 206 relative to the radiation source 208. The detector 210 may include, for example, a one-dimensional or two-dimensional array of pixels, such as direct conversion detector pixels. The detector 210 may be adapted to detect radiation traversing the examination region 206 and generate a signal indicative of its energy.

[0050] The CT scanning unit 200 may include generators 211 and 213. The generator 211 may generate projection data 209 based on signals from the detector 210. The generator 213 receives the projection data 209 and, in some embodiments, may generate a three-dimensional image 311 of the subject based on the projection data 209. In some embodiments, the projection data 209 may be provided to an input 115 of the processing unit 110, while in other embodiments, the 3D image 311 is provided to an input of the processing unit.

[0051] Figure 3 shows an imaging chain for use in a method / apparatus according to an embodiment of the present invention, Figure 4 shows a schematic diagram of a model structure for implementing a method according to an embodiment of the present invention, and Figure 5 is a flow chart illustrating a method according to an embodiment of the present invention.

[0052] As shown, according to one embodiment of the method, CT data is processed into an image, typically including retrieving projection data (500) acquired from multiple angles around a central axis.

[0053] 2, the subject may be a patient on a support 207, and the central axis may be an axis through the examination region. When projection data is acquired from the imaging device 200, the rotating gantry 204 may then rotate about the central axis of the object, thereby acquiring projection data from various angles.

[0054] Once acquired, the projection data is processed (510) and then processed (520) to reconstruct the 3D image 300.

[0055] Although reconstruction (510) and processing (520) are shown as separate processes, it is understood that reconstruction itself may be the actual processing of the projection data into a 3D image. Likewise, reconstruction may be part of such processing. Such reconstruction (510) may be by using standard reconstruction techniques, such as by filtered backprojection.

[0056] As shown in Figures 3 and 5, processing (520) can include denoising 310 (523), which can be performed, for example, by a neural network or other algorithm, e.g., an AI-based learning algorithm. In the illustrated example, denoising 310 (523) is by a convolutional neural network (CNN) previously trained on suitable images. Such denoising processing 310 (523) can be utilized, for example, when the CT imaging is noisy, such as in the case of ULDCT images. The denoising process (523) can then result in a denoised or partially denoised 3D image 320.

[0057] The denoising process 310 (523) may incorporate features that allow it to generalise well to different contrasts, anatomical structures, reconstruction filters, and noise levels. Such a denoising process 310 can compensate for the high noise levels inherent in ULDCT images. The denoising process 310 (523) involves denoising the imaging data using a denoising algorithm with at least one adjustable parameter. During initial processing (520), the adjustable parameter is typically set to a first value. The first value may be based on generic or non-diseased images, or may be based on expected diseases associated with CT imaging.

[0058] In the illustrated example, processing of the 3D image (520) may further include implementation of a super-resolution process 330 (526). As with the denoising process 310, the super-resolution process 330 (526) may be via other algorithms, such as neural networks or AI-based learning algorithms, e.g., CNNs. In some embodiments, the super-resolution process 330 may include deblurring the image. The super-resolution process 330 results in a higher resolution 3D image 340.

[0059] The super-resolution process 330 typically interpolates the image to a smaller voxel size while maintaining or improving perceived image sharpness. Alternatively, the super-resolution process 330 may operate on a fixed voxel size and perform operations such as deconvolution to restore higher resolution from blurred images. The AI-based super-resolution process 330 may be trained on either actual CT images, including ULDCT images, or more general image material, such as natural high-resolution photographs. The super-resolution process 330 (526) may include performing the AI-based super-resolution process using a super-resolution algorithm having at least one adjustable parameter. During the initial process (520), the adjustable parameter is typically set to a first value. The first value may be based on typical or non-diseased images or on expected diseases associated with CT imaging.

[0060] In the illustrated embodiment, both the denoising process 310 (523) and the super-resolution process 330 (526) are applied sequentially. However, it will be understood that both processes may be incorporated into a single neural network, such as a CNN. Furthermore, while both processes 310, 330 are shown as being applied to the 3D image 300, in some embodiments the processes may be applied directly to the projection data prior to reconstruction (510). Furthermore, in some embodiments, one or both of the processes 310, 330 may be applied on a two-dimensional plane within the 3D CT imaging dataset that is perpendicular to the projection direction used to generate the two-dimensional image described below.

[0061] In some embodiments, processing can further include identifying 530 at least one physical element in the 3D image. Once identified 530, the physical element can be removed or masked from the 3D image 535. By removing or masking 535 the element prior to generation of the 2D image, such physical element can be removed from simulated x-rays generated from the projection data.

[0062] The identified physical element (530) may be an anatomical element, such as one or more ribs or a heart. By removing such an anatomical element from the simulated x-ray, other anatomical elements of interest to a radiologist viewing the image may be more easily seen; for example, the simulated x-ray may show an incision in the patient's chest cavity without obstructing ribs.

[0063] Alternatively, the identified physical element (530) may be a table or support 207 or an implant. CT image data is typically acquired with a patient lying on a table or support 207, as in the imaging device 200 described above. In contrast, conventional planar x-rays are often acquired with the patient in an upright position. Thus, by removing the support 207, the simulated x-ray may appear more natural to the radiologist viewing the image. Similarly, removing an implant may provide a better view of the patient's anatomy.

[0064] In some embodiments, rather than removing the identified physical elements (530), the physical elements may be weighted. Similarly, different sections of the three-dimensional image 300 may be weighted differently. The removal, masking out, or weighting (530) of the identified physical elements may be based on an algorithm having adjustable parameters. Thus, as described in the context of the denoising (523) and super-resolution (526) processes, the adjustable parameters may be given first values ​​prior to initial processing.

[0065] Following processing of the 3D image (520), the method proceeds to generating a two-dimensional image 350 (540) based on the three-dimensional image. This can be performed in several ways, but in some embodiments, such generation of the two-dimensional image 350 is by tracing x-rays from a simulated radiation source outside the subject of the three-dimensional image. Such an x-ray tracing process can be performed, for example, by implementing a Siddons Jacobs ray tracing algorithm.

[0066] Once the two-dimensional image 350 is generated, the method proceeds to identify diseases represented in the two-dimensional image (550). Such identification may be through the use of a pre-trained disease classification network 360. Such a network may be configured to recognize a specific set of diseases associated with specific anatomical structures appearing in the two-dimensional image 350. For example, chest imaging may be processed using a disease classification network 360 trained to identify cardiac hypertrophy 370, emphysema 380, and individual nodules 380. Thus, a disease classification model may be pre-trained based on potential diseases. In addition to identifying specific diseases, the network 360 may be designed to output additional parameters of the identification, such as the degree of uncertainty in the results. Thus, in embodiments in which portions of the three-dimensional image are masked out before generating the updated two-dimensional image 540, nodules may be identified but may be partially hidden, making parameters based on such uncertainty in the results meaningful.

[0067] Upon identifying a disease represented in the two-dimensional image 350, the method recalibrates (560) the imaging chain by either 1) reconstructing the projection data as a 3D image or 2) processing the original reconstructed 3D image in the image domain without requiring reconstruction in the projection domain. This results in an updated 3D image. Such recalibration is based on the identified disease (550) and / or the output of the pre-trained disease classification network 360. Such recalibration includes modifying (520) a portion of the processing of the image such that at least one parameter of the reprocessing is based on the identified disease (550) and / or the output of the network 360. For example, in some embodiments, the recalibration may take into account the reported uncertainty of the disease classification network in addition to the actual identified disease.

[0068] As described above, processing of the image (520) can include denoising (523) and / or super-resolution processing (526). As described above, denoising (523) can utilize a denoising algorithm having adjustable parameters. Recalibrating (560) the imaging chain can then include providing (523) a second value for an adjustable parameter of the denoising algorithm that is different from the first value. The second value can then be based on the identified disease (550). Similarly, super-resolution processing (523) can utilize a super-resolution algorithm having adjustable parameters. Recalibrating (560) the imaging chain can then include providing (560) a second value for an adjustable parameter of the super-resolution algorithm (526) that is different from the first value. As with the denoising algorithm, the second value can be based on the identified disease (550).

[0069] Similarly, in embodiments in which physical elements are identified (530) and removed, masked, or weighted (535), calibration of the imaging chain (560) can set adjustable parameters of the algorithms used to control the identification, removal, masking, or weighting. Thus, the adjustable parameter may initially be given a first value, but calibration (560) may set it to a second value different from the first value. Thus, removal or masking out of at least one physical element may be based on identification (550) of a disease represented in the two-dimensional image.

[0070] Once the 2D image has been generated (540) based on the at least one updated parameter calibrated through the imaging chain, the method proceeds with presenting the updated 2D image to the user (570).

[0071] FIG. 6 illustrates an implementation of a ray tracing process applied to the three-dimensional image 300 (540) to generate a two-dimensional image 350. The ray tracing process can then proceed by simulating the processing of X-rays 345 by propagating incident X-ray photons from a simulated radiation source 600 through the reconstructed three-dimensional image 300. The generation of the two-dimensional image 350 may be via a neural network such as a CNN, in which case the CNN may incorporate one or more of the denoising and super-resolution processes described herein. Such a neural network may be a generative adversarial network (GAN). In such an embodiment, many or all of the steps described herein may be combined into a single network, such that CT volume data is provided to the network and simulated CXR projections are output.

[0072] In some embodiments, generation of the two-dimensional image (540) may be based on an algorithm having tunable parameters, such as an algorithm utilizing a CNN. Such a CNN may be a network configured to generate digitally reconstructed radiographs (DRRs). As in the case of the adjustable parameters of the denoising algorithm (523) and the super-resolution algorithm (526), ​​such adjustable parameters may be assigned first values ​​during the initial generation of the two-dimensional image (540). In such embodiments, after identifying the disease (550) and calibrating the imaging chain (560), and during reprocessing of the image, the method may generate an updated two-dimensional image based on the identification of the disease represented in the two-dimensional image and on a second value for the adjustable parameter.

[0073] In some embodiments, the projection angle or orientation of the ray tracing process (540) can be adjusted to improve the resulting two-dimensional image 350. Similarly, the weighting of physical elements in the three-dimensional image 300 can be adjusted to improve the resulting two-dimensional image 350. Such characteristics can be controllable by at least one adjustable parameter that can be set by a calibration process (560).

[0074] Figure 7 is a flow chart illustrating a method for training an imaging chain in accordance with the present invention. The method begins by retrieving projection data acquired from multiple angles about a central axis (700). Thus, the acquisition of such training data is similar to that described above with respect to Figure 5 (500).

[0075] The training data further includes an indicator of a disease state associated therewith, which may indicate a particular disease, such as cardiac hypertrophy 370, emphysema 380, and the presence of individual nodules 380. Alternatively, the indicator of the disease state may indicate that the training data corresponds to healthy anatomy.

[0076] The method then proceeds with reconstruction (710) and processing (720) the reconstructed 3D image. The processing of the image (720) is similar to that described above with respect to Figure 5 and proceeds along a similar image formation chain. Thus, the processing of the image (720) includes denoising (723) and super-resolution processing (726). After such processing, a two-dimensional image is generated (730) based on the training image data.

[0077] As discussed above with respect to Figure 5, the processes incorporated into the imaging chain, including the denoising process (723), the super-resolution process (726), and the generation of two-dimensional images (730), can each incorporate adjustable parameters, which can be provided with default or first values. The methods described herein for training the imaging chain can be used to derive disease-specific second values ​​for these adjustable parameters to enhance the output of the imaging chain after disease identification.

[0078] Following generation of the two-dimensional image (730), a previously trained disease classification network is applied to the two-dimensional image output (740) and used to generate a disease classification prediction associated with the training image data. A loss metric is then defined (750) based on the output of the previously trained disease classification network and the disease state indicator associated with the training image data.

[0079] The loss metric may correspond to the classification error 400 and may then be provided to the pre-trained disease classification net. Such classification error may be back-propagated through the imaging chain (410) to identify and refine values ​​for adjustable parameters of the imaging chain (FIG. 4). Accordingly, the loss metric (750) is then used to adjust (760) at least one parameter of the processing of the training images to minimize the loss metric. Thus, because the loss metric is defined (750) based on the disease state indicator, the values ​​of one or more of the adjustable parameters identified by the training method are optimized for the specified disease state.

[0080] The training process may be repeated multiple times on a large number of training images for various specified disease states. Furthermore, the adjustable parameter may be a single adjustable parameter in one of the denoising process 310, the super-resolution process 330, or the DRR simulation process 345, or may be, for example, a first adjustable parameter in the denoising process 310 and a second adjustable parameter in the super-resolution process, each of which is adjusted by the training method to find an ideal value.

[0081] In some embodiments, at least one parameter affects the suppression of anatomical structures in an output image generated by processing the three-dimensional image using a tuned value of the at least one parameter. Thus, the disease state indicator can indicate the presence of an identified disease, and the loss metric can optimize the visibility of the identified disease indicator. For example, the loss metric can optimize the adjustable parameter to increase the visibility of nodules.

[0082] In some such embodiments, the loss metric optimizes the visibility of nodules that would otherwise be obscured by adjacent anatomical structures. Thus, the tuned parameters can result in masking of blocking anatomical structures, such as the rib cage, for example.

[0083] For example, in the context of chest imaging, identification of pulmonary nodules is notoriously difficult in chest x-rays due to film resolution, nodule size, and the potential for nodule obscuration by adjacent anatomical structures. In this embodiment, the imaging chain is then optimized to suppress obscuring of adjacent anatomical structures in the DRR image simulation and to enhance nodule detail and granularity via the super-resolution and denoising modules, respectively.

[0084] Each model in the imaging chain can typically accept the output of other individual models.

[0085] A pre-trained disease classification network expects images (or DRRs in this case) from the domain of images it was trained on. Specifically, it means feeding DRR images from chest CT scans if it was trained on chest X-rays, etc. The added value of combining these features is the creation of a model that can simultaneously generate and optimize 2D representations from 3D images, optimizing for real-world applications such as disease classification.

[0086] In some embodiments, the imaging chain can be extended in both directions. In one embodiment, the imaging chain can include a disease classification model based on the underlying three-dimensional image. This can complement two-dimensional pre-trained classification models by incorporating contextual and spatial information into the classification of a given case. Similarly, the imaging chain can be extended to include CT image reconstruction, so that learnable parameters can be utilized in the reconstruction process. For example, visualization of individual CT slides can be improved for specific disease classes.

[0087] Various embodiments of the present invention may be implemented on a computer as a computer-implemented method, or on dedicated hardware, or a combination of both. Executable code for a method according to an embodiment of the present invention may be stored on a computer program product. Examples of computer program products include memory devices, optical storage devices, integrated circuits, servers, online software, etc. Preferably, the computer program product may include non-transitory program code stored on a computer-readable medium for performing a method according to the present invention when the program product is run on a computer. In one embodiment, a computer program may include computer program code adapted to perform a method according to various embodiments of the present invention when the computer program is run on a computer. The computer program may be embodied on a computer-readable medium.

[0088] While the present invention has been described at some length and with some particularity with respect to several described embodiments, it should not be limited to any such details or embodiments or to any particular embodiment, but should be construed with reference to the appended claims so as to provide the broadest possible interpretation of such claims in view of the prior art, and therefore so as to effectively encompass the intended scope of the invention.

[0089] All examples and conditional language recited herein are for educational purposes to aid the reader in understanding the principles of the present invention and concepts contributed by the inventors to further the art, and should not be construed as being limited to such specifically recited examples and conditions. Moreover, all statements herein reciting principles, aspects, and embodiments of the present invention, as well as specific examples thereof, are intended to encompass both structural and functional equivalents thereof. Furthermore, such equivalents are intended to include both currently known equivalents and equivalents developed in the future, i.e., any elements developed that perform the same function, regardless of structure.

Claims

1. 1. A computer-implemented method for processing medical computed tomography data, comprising: retrieving projection data obtained by scanning the object; processing the projection data to reconstruct a three-dimensional image; generating a two-dimensional image based on the three-dimensional image; identifying a disease or uncertainty in identifying the disease in the two-dimensional image; reprocessing the projection data or the three-dimensional image into an updated three-dimensional image, wherein at least one parameter of the reprocessing is based on the disease or uncertainty in the identification of the disease; generating an updated two-dimensional image based on the updated three-dimensional image; A method comprising:

2. The method of claim 1 , wherein the step of generating the two-dimensional image is performed by a neural network.

3. The method of claim 1 , wherein the step of reprocessing the projection data or the three-dimensional image is performed by a neural network.

4. The method of claim 1 , further comprising denoising the three-dimensional image using the first value for the at least one parameter.

5. 5. The method of claim 4, further comprising denoising the updated three-dimensional image using a second value for the at least one parameter, the second value being based on uncertainty in the disease or in identifying the disease.

6. 2. The method of claim 1, further comprising applying artificial intelligence-based super-resolution to the three-dimensional image using a first value for the at least one parameter, such that the three-dimensional image results in a high-resolution image.

7. 7. The method of claim 6, further comprising applying the AI-based super-resolution to the updated three-dimensional image using a second value for the at least one parameter, such that the updated three-dimensional image results in another high-resolution image, the second value being based on uncertainty in the disease or identification of the disease.

8. The method of claim 1 , further comprising denoising the 3D image and / or the updated 3D image by a trained convolutional neural network.

9. The method of claim 1 , wherein generating the two-dimensional image comprises using a digitally reconstructed radiographic network using a first value for the at least one parameter.

10. 10. The method of claim 9, wherein generating the updated two-dimensional image comprises using the radiographic network with a second value for the at least one parameter, the second value being based on uncertainty in the disease or the identification of the disease.

11. 10. The method of claim 1, further comprising identifying at least one physical element in the three-dimensional image and / or the updated three-dimensional image, and removing or masking the at least one physical element before generating an output image.

12. The method of claim 11 , wherein the at least one physical element is a plurality of ribs or a heart.

13. 12. The method of claim 11, wherein the step of removing or masking the at least one physical element is based on uncertainty in the disease or in the identification of a disease in the two-dimensional image.

14. 10. The method of claim 1, wherein the step of identifying a disease comprises using a trained disease classification model.

15. 1. A system for medical computed tomography data, comprising: a memory for storing a plurality of instructions; a processor, coupled to the memory, configured to execute the plurality of instructions, the plurality of instructions comprising: retrieving projection data acquired by scanning the object; processing the projection data to reconstruct a three-dimensional image; generating a two-dimensional image based on the three-dimensional image; identifying a disease or uncertainty in identifying a disease in the two-dimensional image; reprocessing the projection data or the three-dimensional image into an updated three-dimensional image, wherein at least one parameter of the reprocessing is based on uncertainty in the disease or the identification of the disease; generating an updated two-dimensional image based on the updated three-dimensional image; Including, the system.

16. 16. The system of claim 15, wherein the steps of generating the two-dimensional image and reprocessing the projection data or the three-dimensional image are performed by a neural network.

17. 16. The system of claim 15, wherein the three-dimensional image is denoised using a first value for the at least one parameter and the updated three-dimensional image is denoised using a second value for the at least one parameter, the second value being based on uncertainty in the disease or identification of the disease.

18. 16. The system of claim 15, wherein the 3D image and / or the updated 3D image are denoised by a trained convolutional neural network.

19. 16. The system of claim 15, wherein the two-dimensional image is generated by a digitally reconstructed radiography network using a first value for the at least one parameter, and the updated two-dimensional image is generated by the radiography network using a second value for the at least one parameter, the second value being based on uncertainty in the disease or the identification of the disease.

20. 16. The system of claim 15, wherein artificial intelligence based super-resolution using a first value for the at least one parameter is applied to the three-dimensional image to result in a high-resolution image, and artificial intelligence based super-resolution using a second value for the at least one parameter is applied to the updated three-dimensional image to result in another high-resolution image, the second value being based on uncertainty in the disease or the identification of the disease.

Images