System for providing automatic segmentation to non-contrast computed tomography imaging data and method thereof

Abstract

The present invention relates to a system for providing automatic segmentation to non-contrast computed tomography imaging data and a method thereof, the system comprising: an input data interface configured to obtain imaging data comprising at least non-contrast computed tomography imaging data; a segmentation module configured to implement a segmented artificial intelligence model adapted to generate segments for the obtained non-contrast computed tomography imaging data, where the segmented artificial intelligence model is trained with at least imaging data based on segmented contrast computed tomography imaging data; and an output data interface configured to output the generated segments.

Description

Technical Field

[0001] The present invention relates to a system for providing automatic segmentation of non-contrast computed tomography imaging data. The present invention also relates to a corresponding method, a computed tomography system, a computer program product and a non-transitory computer readable data storage medium. Background Art

[0002] The invention is in the field of medical imaging, where it is used to create visual representations of the inside of the body based on detection of physical signals and analysis of the resulting image data to support clinical analysis and medical intervention, and to create visual representations of organ or tissue function.

[0003] The present invention is primarily described with respect to vessel segmentation of non-contrast computed tomography imaging data, but the principles of the present invention are broader in scope and are equally applicable to other segmentation tasks associated with non-contrast computed tomography imaging data.

[0004] In this and upcoming articles, individuals with both male and female identities are included within the term, regardless of grammatical term usage.

[0005] Detecting changes in the lumen of a vessel is a key step in vascular imaging. Such changes may be pathological in nature, such as vascular stenosis or calcification, and may impede blood flow and cause inadequate supply to the corresponding downstream tissues. A commonly used form of assessment is computed tomography (CT).

[0006] To support clinical diagnosis and provide better patient-specific treatment, automated algorithms for blood flow assessment and stenosis detection are considered a valuable tool. In particular, deep learning-based methods for segmenting vascular structures from CT images are one of the most promising approaches to achieve this goal.

[0007] However, implementation of these methods typically requires the existence and availability of large amounts of high-quality image data, which must contain highly accurate annotations in order to achieve diagnostic performance.

[0008] Acquiring the necessary amount of data poses a well-known challenge. First, acquiring medical images is difficult due to various data privacy and data security regulations. Second, the data must be manually annotated with the highest possible accuracy in order to produce the consistent and high-quality datasets required for the training process of deep learning models.

[0009] The current state of the art in automated segmentation of CT imaging data is primarily based on computed tomography angiography (CTA), in which an intravenous contrast agent is administered to the patient to enhance the visibility and contrast of blood vessels, leading to better differentiation of blood vessels from surrounding tissue for both human and algorithmic assessment.

[0010] Contrast CT or contrast-enhanced computed tomography (CECT) is an X-ray computed tomography (CT) that uses a contrast agent. The contrast agent for X-ray CT is usually an iodine-based contrast agent. This helps to highlight structures such as blood vessels that would otherwise be difficult to delineate with their surroundings. The use of contrast material may also help to gain functional information about tissues. Typically, images are obtained with and without radioactive contrast.

[0011] However, in certain circumstances and patients, CTA may be contraindicated. This may occur when the patient is allergic or hypersensitive to the contrast agent used, or when the patient has renal failure or has received a kidney transplant. Contrast agents may also cause side effects such as nausea, cough, paresthesia, itching, swelling of the mucous membranes, or shortness of breath. Others have been shown to accumulate in the body, causing negative long-term effects such as contrast-induced nephropathy. In some of these cases, non-contrast computed tomography (NCCT) is preferred, and in some cases even necessary.

[0012] However, the implementation of NCCT algorithms is extremely challenging. In contrast to CTA, automated segmentation algorithms for NCCT are virtually non-existent, and clinicians must rely on tedious manual annotation. This is also made more difficult by the lack of contrast enhancement, which makes manual annotation more prone to inaccuracies. As a result, the available data for training customized algorithms for NCCT is too low to meet clinical standards. Especially for cardiovascular disease, this poses a significant obstacle to clinical assessment.

[0013] In this context, the problem addressed by the present invention is to provide better automatic segmentation of non-contrast computed tomography imaging data.

[0014] According to the invention, this problem is solved by a system having the features of claim 1 and / or a computer-implemented method having the features of claim 12 and / or a computer program product having the features of claim 14 and / or a non-transitory computer-readable data storage medium having the features of claim 15. Summary of the invention

[0015] A first aspect of the present invention provides a system for providing automatic segmentation for non-contrast computed tomography imaging data, the system comprising: an input data interface, configured to obtain imaging data comprising at least non-contrast computed tomography imaging data; a segmentation module, configured to implement a segmentation artificial intelligence model, which is suitable for generating segmentations for the obtained non-contrast computed tomography imaging data, wherein the segmentation artificial intelligence model is trained based on segmented contrast computed tomography imaging data at least with the imaging data; and an output data interface, configured to output the generated segmentations.

[0016] In the present invention, CT imaging data broadly describes any information generated from a CT device and can be used for further processing and analysis in medical applications of digital medicine. CT imaging data may include two-dimensional imaging data generated from CT scan images recorded from a CT device using any conventional procedure. The CT imaging data may consist of one or more scanned images of soft tissue samples of an organ of a mammal, wherein some of these images may include selected sub-image portions or crops of larger images.

[0017] Under contrast CT imaging data is understood CT imaging data recorded after administration of a contrast agent to a patient. Non-contrast CT imaging data correspond to those CT techniques that do not involve a contrast agent.

[0018] Segmented (or annotated) CT imaging data corresponds to CT imaging data in which metadata of pixels or parts of a CT image are introduced with the purpose of identifying or recognizing parts of the CT image belonging to a certain category and being able to distinguish them from other parts of the CT image. The segmented CT imaging data may, for example, include information for distinguishing blood vessels from other body structures of a patient. The segmented CT imaging data may in particular be semantically segmented CT imaging data.

[0019] In this and upcoming articles, an artificial intelligence model will be understood as a model that includes one or more artificial neural networks (or encoders). An artificial neural network is a computing system based on a collection of connected units or nodes that are grouped into layers that can send signals to each other. The artificial neural network involved in the present invention may preferably include convolutional neural networks and diffusion models, but other neural networks may also be used.

[0020] The different modules, units and interfaces mentioned in this application are broadly understood as entities capable of acquiring, obtaining, receiving or retrieving general data and / or instructions through a user interface and / or programming code and / or executable program or any combination thereof. In particular, the image domain transfer module and the segmentation module are adapted to run the programming code and executable program and deliver the results for further processing.

[0021] Therefore, different modules, units and interfaces or parts thereof may respectively include at least a central processing unit CPU and / or at least one graphics processing unit GPU and / or at least one field programmable gate array FPGA and / or at least two application specific integrated circuits ASIC and / or any combination of the foregoing. Each of them may also include a working memory operably connected to at least one CPU and / or a non-transient memory operably connected to at least one CPU and / or working memory. Each of them may include an application programming interface (API) or be composed of it.

[0022] All parts of the system of the present invention can be implemented in hardware and / or software, wired and / or wireless and any combination thereof. Any part of the system may include an interface to an intranet or the Internet, to a cloud computing service, to a remote server, etc.

[0023] In particular, each module, unit and interface of the system or the entire system can be partially and / or completely implemented in a local device (such as a computer), in a computer system and / or partially and / or completely in a remote system, in particular in an edge or cloud computing platform.

[0024] In a system based on cloud computing technology, a large number of devices are connected to the cloud computing system via the Internet. These devices can be located in a remote facility connected to the cloud computing system. For example, these devices may include or consist of equipment, sensors, actuators, robots and / or machinery in (multiple) industrial settings. These devices may be equipment in medical devices and health care units. These devices may be household appliances or office appliances in residential / commercial institutions.

[0025] Cloud computing systems can enable remote configuration, monitoring, control, and maintenance of connected devices (also commonly referred to as 'assets'). Moreover, cloud computing systems can facilitate storage of large amounts of data periodically collected from devices, analysis of large amounts of data, and provision of insights (e.g., key performance indicators, outliers) and alerts to operators, field engineers, or device owners via a graphical user interface (e.g., a web application). The insights and alerts can enable control and maintenance of the devices, thereby achieving efficient and fail-safe operation of the devices. Cloud computing systems can also enable modification of parameters associated with the devices, and issuance of control commands via a graphical user interface based on the insights and alerts.

[0026] A cloud computing system may include multiple servers or processors (also referred to as 'cloud infrastructure') that are geographically distributed and connected to each other via a network. A dedicated platform (hereinafter referred to as 'cloud computing platform') is installed on the servers / processors for providing the above functionality as a service (hereinafter referred to as 'cloud service'). A cloud computing platform may include multiple software programs executed on one or more servers or processors of a cloud computing system to enable the delivery of requested services to devices and their users.

[0027] One or more application programming interfaces (APIs) are deployed in a cloud computing system to deliver various cloud services to users.

[0028] A second aspect of the present invention provides a computer-implemented method for providing automatic segmentation for non-contrast CT imaging data, comprising the following steps: obtaining imaging data including at least non-contrast computed tomography imaging data; generating segmentations for the obtained non-contrast computed tomography imaging data using a segmented artificial intelligence model, wherein the segmented artificial intelligence model is trained at least with the imaging data based on segmented contrast computed tomography imaging data; and outputting the generated segmentations.

[0029] In particular, the method according to the second aspect of the invention may be performed by the system according to the first aspect of the invention. Therefore, the features and advantages disclosed herein in conjunction with the computing device are also disclosed for the method, and vice versa. In other words, the claims of the system according to the invention may be improved by features described or claimed in the context of the method, and vice versa. In this case, the functional features of the method are implemented by the target unit or module of the system.

[0030] According to a third aspect, the present invention provides a computer program product comprising executable program code which, when executed, is configured to perform the method according to the second aspect of the present invention.

[0031] According to a fourth aspect, the present invention provides a non-transitory computer-readable data storage medium comprising executable program code which, when executed, is configured to perform the method according to the second aspect of the present invention.

[0032] The non-transitory computer-readable data storage medium may include or consist of any type of computer memory, in particular semiconductor memory, such as solid-state memory. The data storage medium may also include or consist of a hard drive, a CD, a DVD, a Blu-ray disc, a USB memory stick, etc.

[0033] According to a fifth aspect, the present invention provides a data stream comprising or configured to generate executable program code, which when executed is configured to perform the method according to the second aspect of the present invention.

[0034] According to a sixth aspect, the present invention provides a computer tomography system having a computer tomography apparatus and a system according to the first aspect of the present invention. The computer tomography apparatus is configured to perform a computer tomography scan and acquire imaging data, preferably using photon counting technology, while the system according to the first aspect of the present invention is configured to obtain imaging data acquired by the computer tomography apparatus and provide segmentation thereof.

[0035] A computer program product according to a third aspect of the present invention can be provided as an application programming interface (API) to a computed tomography device according to a sixth aspect of the present invention.

[0036] One of the main ideas underlying the present invention is to provide a system capable of automatically segmenting non-contrast CT imaging data. This is achieved by training an artificial intelligence model that is adapted to implement a segmentation algorithm. The artificial intelligence model is trained with a data set based on segmented (or annotated) contrast imaging data. Thus, the artificial intelligence model is trained to provide annotations to non-contrast imaging data based on the annotations belonging to the contrast imaging data. Thereby, the system of the present invention allows the deployment of existing segmentation algorithms for contrast CT imaging data to non-contrast CT imaging data.

[0037] The above system allows for a simple implementation of a computer-implemented method comprising multiple steps. Initially, imaging data including at least non-contrast CT imaging data is obtained. The segmentation of such non-contrast imaging data is generated by the operation of an artificial intelligence model that is trained with imaging data based on segmented contrast CT imaging data. Through this learning, an image domain transformation of the annotations is achieved and automatic segmentation of the non-contrast imaging data can be performed.

[0038] An advantage of the present invention is that it provides a system and method capable of segmenting non-contrast CT imaging data using training algorithms conventionally used for segmenting contrast CT imaging data. Thus, the segmentation of non-contrast CT imaging data can be automated and existing algorithms can be used.

[0039] Another advantage of the present invention is that the system and method eliminate the need for manual annotation, thereby improving the accuracy of segmentation and reducing the annotation time, especially since the annotations already exist in large data sets of contrast CT imaging data.

[0040] Advantageous embodiments and further developments result from the description of the dependent claims and the different preferred embodiments illustrated in the figures.

[0041] According to some embodiments, refinements or variants of the embodiments, the segmentation artificial intelligence model is also trained with photon-counting-based virtual non-contrast computed tomography imaging data.

[0042] Generating training datasets for segmentation AI models is challenging due to the difficulty in obtaining annotated non-contrast datasets. CT devices endowed with photon counting technology generate scans from which spectral information at different energy thresholds can be obtained. This wealth of information along with the development and improvement of existing algorithms can be used to generate virtual non-contrast (VNC) imaging data from contrast imaging data. Using the combination of imaging data from at least two energy thresholds, VNC imaging data can be generated, allowing the contrast to be removed.

[0043] One of the advantages of VNC is that it involves the processing of contrast imaging data acquired at a single time point, and therefore, no registration of the resulting virtual images is required.

[0044] VNC images can be specifically generated from multi-energy (or spectral) computed tomography angiography (CTA). CTA is one of the gold standard techniques for contrast computed tomography. It uses intravenous contrast agents to enhance the visibility and differentiation of blood vessels from surrounding tissue.

[0045] According to some embodiments, refinements or variations of the embodiments, the segmentation artificial intelligence model is adapted to implement a segmentation algorithm, in particular a segmentation algorithm capable of processing imaging data from contrast CTA. In accordance with the principles of the present invention, a segmentation algorithm designed for segmentation of contrast CT imaging data can also be deployed for segmentation of non-contrast CT imaging data. Thus, all algorithms developed and tested for contrast CT imaging data can be reused or re-adapted, with all the benefits that come with it in terms of efficiency and time savings.

[0046] According to some embodiments, refinements or variations of the embodiments, the system also includes an image domain transfer module configured to implement an image domain transfer artificial intelligence model, which is trained and suitable for generating synthetic segmented non-contrast CT imaging data based on segmented contrast CT imaging data, wherein the generated segmented non-contrast CT imaging data is used as a training data set for training the segmented artificial intelligence model.

[0047] The image domain transfer AI model provides a mechanism to generate non-contrast training data for the segmentation AI model from annotated contrast imaging data. This ensures the existence of a sufficient amount of training data, as annotated contrast CT imaging data is quite abundant.

[0048] Therefore, the trained image domain transfer artificial intelligence model is suitable for performing domain transfer from contrast computed tomography (CT) imaging data to non-contrast CT imaging data. Since the domain transfer preserves the annotations of the contrast CT imaging data, the training data is also annotated.

[0049] By utilizing the image domain transfer AI model to generate synthetic imaging data, the segmentation algorithm of the segmentation AI model is provided with the generation of arbitrarily large training data sets, which means that the quality of the segmentation algorithm is correspondingly improved.

[0050] According to some embodiments, refinements or variations of embodiments, the image domain transfer AI model is based on a denoising diffusion probabilistic model. Denoising diffusion probabilistic models (DDPMs) are deep generative AI models that are trained and adapted to generate synthetic data, particularly suitable for performing transitions between image domains. DDPMs are particularly advantageous because they are very efficient in terms of training, i.e., very good results can be obtained with only a relatively small amount of training data. In the present case, this means that they can be successfully used even if only a few paired imaging data are available.

[0051] According to some embodiments, refinements or variations of embodiments, the image domain transfer artificial intelligence model is trained using paired imaging data corresponding to segmented computed tomography angiography and non-contrast computed tomography and / or using photon counting-based VNC imaging data.

[0052] The image domain transfer artificial intelligence model can be trained with annotated contrast CT imaging data and corresponding non-contrast CT (NCCT) imaging data, such as NCCT obtained from a CT device, which must be paired with the annotated contrast CT imaging data. Alternatively or additionally, if the CT device generating the contrast CT imaging data is endowed with photon counting technology, VNC imaging data can be generated. In these embodiments of the present invention, the image domain transfer artificial intelligence model can advantageously be trained with non-contrast CT imaging data, thereby reducing the impact of potential biases of the pairing algorithm or VNC algorithm used to pre-process the training imaging data.

[0053] According to some embodiments, refinements or variations of embodiments, the image domain transfer module includes a registration unit, which is configured to implement a registration algorithm suitable for pairing segmented CTA imaging data with NCCT imaging data. The different acquisition times of contrast and non-contrast imaging data require that the two data sets be paired before they can be used to train the image domain transfer artificial intelligence model. A transformation is required to associate, for example, pixels of an image of a contrast imaging data set with corresponding images of a non-contrast imaging data set. This can be achieved by a non-rigid registration algorithm. The registration algorithm can be implemented in a computer, server or cloud platform.

[0054] According to some embodiments, refinements or variations of embodiments, the image domain transfer module comprises a VNC generation unit configured to generate virtual non-contrast computed tomography imaging data based on contrast multi-energy computed tomography angiography.

[0055] As mentioned above, VNC imaging data is a post-processing step that can be applied to data obtained from a CT device using photon counting technology. Using photon counting technology, spectral information at different energy thresholds can be obtained. Using the combination of data from at least two energy thresholds, VNC imaging data can be generated, so that the contrast of the original CTA imaging data can be removed. Due to the nature of VNC imaging data as a post-processing of CTA imaging data, the VNC imaging data is paired and does not need to be aligned. In some cases, it may be advantageous to generate the VNC imaging data by the system of the present invention itself rather than having to import it, especially because photon counting technology is not currently common.

[0056] According to some embodiments, refinements or variations of embodiments, further comprising a database module configured to store synthetic segmented NCCT imaging data generated using the image domain transfer artificial intelligence model.

[0057] Generating synthetic high quality non-contrast CT imaging data is one of the primary mechanisms for training the segmentation artificial intelligence model. Additionally, if the segmentation algorithm changes or is upgraded, the segmentation artificial intelligence model may need to be retrained. The training data set stored in the database module ensures that this can be done at any time. Training data sets can also be generated for training additional artificial intelligence models that are not necessarily part of the present invention. Therefore, the imaging data stored in the database module can also be exported for training other segmentation algorithms.

[0058] According to some embodiments, refinements or variations of embodiments, the obtained NCCT imaging data and segmented contrast CT imaging data are derived from a computed tomography scan of a vessel of the patient.

[0059] The present invention is primarily described with respect to vascular imaging, where detecting changes in the vessel lumen is a key aspect. In particular, segmentation of blood vessels is essential for correctly assessing blood flow in patients, identifying abnormalities and preventing disease.

[0060] Although here, in the foregoing and below, some functions are described as being performed by modules or units, it should be understood that this does not necessarily mean that such modules or units are provided as separate entities from each other. In the case where one or more modules or units are provided as software, these modules or units can be implemented by program code segments or program code fragments, which can be different from each other, but can also be intertwined or integrated with each other.

[0061] Similarly, where one or more modules or units are provided as hardware, the functionality of one or more modules or units may be provided by the same hardware component, or the functionality of several modules or units may be distributed across several hardware components, which do not necessarily need to correspond to the modules or units. Therefore, any device, system, method, etc. that exhibits all features and functions attributed to a specific module or unit should be understood to include or implement the module or unit. In particular, all modules or units may be implemented by program code executed by a computing device (e.g., a server or cloud computing platform).

[0062] Where appropriate, the above configurations and developments may be combined together, and as long as it makes sense, the implementations may be combined with each other as needed.

[0063] Other possible configurations, developments and implementations of the present invention also include combinations of features of the present invention previously described or described below with reference to the embodiments, which combinations are not explicitly mentioned. In particular, in this case, those skilled in the art will also add various aspects as improvements or supplements to the basic form of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] The invention is described in more detail below based on an embodiment shown in the schematic diagram of the accompanying drawings, in which:

[0065] Figure 1 is a schematic depiction of a system for providing automatic segmentation of non-contrast computed tomography imaging data in accordance with an embodiment of the present invention;

[0066] Figure 2 is a block diagram illustrating an exemplary embodiment of a computer-implemented method for providing automatic segmentation for non-contrast computed tomography imaging data in accordance with an embodiment of the present invention;

[0067] Figure 3 is a schematic depiction of different computed tomography imaging data involved in using the system and / or performing the method of the present invention according to a preferred embodiment;

[0068] Figure 4 is a schematic illustration of a computed tomography system comprising a computed tomography apparatus and the system of the first aspect of the present invention;

[0069] Figure 5 is a schematic block diagram illustrating a computer program product according to an embodiment of the third aspect of the present invention; and

[0070] Figure 6 is a schematic block diagram illustrating a non-transitory computer-readable data storage medium according to an embodiment of the fourth aspect of the present invention.

[0071] The attached drawings are intended to provide a further understanding of the embodiments of the present invention. They illustrate the embodiments and, in conjunction with the description, help explain the principles and concepts of the present invention. Other advantages and many of the advantages mentioned become apparent in view of the drawings. The elements in the drawings are not necessarily drawn to scale.

[0072] In the drawings, similar, functionally equivalent and identical operating elements, features and components are provided with like reference numerals in each case, unless stated otherwise.

[0073] The numbering of the steps in the method is for convenience of description. They do not necessarily imply a certain order of the steps. In particular, several steps may be performed concurrently. DETAILED DESCRIPTION

[0074] The detailed description includes specific details in order to provide a thorough understanding of the present invention. However, it will be apparent to one skilled in the art that the present invention may be practiced without these specific details.

[0075] Figure 1 is a schematic depiction of a system 100 for providing automatic segmentation of non-contrast CT imaging data, according to an embodiment of the present invention.

[0076] Figure 1 The depicted system 100 includes an input data interface 10 , an image domain transfer module 20 , a segmentation module 30 , an output data interface 40 , and a database module 50 .

[0077] The input data interface 10 is configured to obtain imaging data, for example, from a plurality of blood vessels of a patient, the imaging data including at least non-contrast computed tomography imaging data D2 of computed tomography imaging data. Figure 1 In the embodiment, the input data interface 10 is configured to additionally obtain segmented contrast CT imaging data D1. According to some embodiments, the segmented contrast CT imaging data D1 may include imaging data from single-energy computed tomography angiography (CTA).

[0078] The input data interface 10 may be any device capable of receiving and processing different data types, such as those detailed above for segmented contrast CT imaging data D1 and non-contrast CT imaging data D2. In a preferred embodiment, the input data interface 10 may be connected to the CT device by wire or wireless or a combination thereof.

[0079] The image domain transfer module 20 is configured to implement an image domain transfer artificial intelligence model that is trained and adapted to generate synthetic segmented non-contrast CT imaging data D3 based on the acquired imaging data.

[0080] The image domain transfer artificial intelligence model can be trained with at least a portion of the imaging data obtained by the input data interface 10. For efficient training, the segmented contrast CT imaging data D1 and the non-contrast computed tomography imaging data D2 generated for this purpose using different CT imaging techniques preferably correspond to the same region of interest. Typically, since the imaging data are acquired at different time points, the imaging data of the segmented contrast CT imaging data D1 and the corresponding imaging data of the non-contrast CT imaging data D2 must be paired, aligned or registered before being used as a training data set so that a meaningful comparison, such as a pixel-based comparison, can be made.

[0081] In some embodiments of the present invention, such as Figure 1 In the embodiment described in the figure, the image domain transfer module 20 comprises a registration unit 210, which is configured to implement a registration algorithm suitable for aligning at least some imaging data of the segmented contrast CT imaging data D1 and the non-contrast CT imaging data D2. The registration algorithm may be a non-rigid registration algorithm.

[0082] The image domain transfer module 20 may also include a VNC generation unit 220, which is configured to generate virtual non-contrast (VNC) computed tomography imaging data based on contrast multi-energy computed tomography angiography (CTA). VNC imaging data is a post-processing step that can be applied to data obtained from a computed tomography device incorporating photon counting technology. Using photon counting technology, spectral information at different energy thresholds can be obtained. Using a combination of data from at least two energy thresholds, VNC imaging data can be generated, whereby contrast can be removed. Therefore, for example, non-contrast imaging data can be obtained from CTA imaging data. The advantage of VNC is that it involves the processing of contrast images to obtain non-contrast images, so there is no need to align the resulting virtual images.

[0083] Therefore, the image domain transfer artificial intelligence model can be trained with the CTA data belonging to the segmented contrast CT imaging data D1 and the non-contrast CT data from the non-contrast CT imaging data D2 once they are registered by the registration unit 210. Alternatively or additionally, the image domain transfer artificial intelligence model can also be trained with the CTA data from the segmented contrast CT imaging data D1 and the VNC data generated thereby using, for example, the VNC generation unit 220.

[0084] The image domain transfer AI model may preferably be based on a denoising diffusion probabilistic model (DDPM), which can be trained even with relatively small amounts of training data.

[0085] The segmentation module 30 is configured to implement a segmentation artificial intelligence model suitable for generating segments D4 to the non-contrast CT imaging data D2.

[0086] The segmented artificial intelligence model can be trained with the synthetic non-contrast CT imaging data D3 produced by the image domain transfer module 20. In some advantageous embodiments of the present invention, it can also be trained with photon counting-based VNC CT imaging data. These VNC images can be used in parallel with the synthetic non-contrast CT imaging data. If the quality of these VNC images is good enough, they may even represent a large part of the training data set.

[0087] The segmentation artificial intelligence model is suitable for implementing segmentation algorithms, in particular segmentation algorithms deployed for segmentation of CTA images. According to the principles of the present invention, segmentation algorithms designed for segmentation of contrast CT imaging data can also be deployed for segmentation of non-contrast CT imaging data. Therefore, all algorithms developed and tested for contrast CT imaging data can be reused, bringing all benefits in terms of efficiency and time saving. The segmentation algorithm can preferably be based on U-Net and U-Net class architectures.

[0088] The output data interface 40 is configured to output the segmentation D4 provided by the segmentation module 30. This can then be evaluated by a medical expert and used as support, for example, for clinical diagnosis or surgical intervention.

[0089] Figure 1 Also shown is a database module 50 configured to store synthetic segmented non-contrast CT imaging data D3 generated using an image domain transfer artificial intelligence model. Thus, the database module 50 can be deployed as a repository for training data sets, such as for training the segmented artificial intelligence model of the segmentation module 30. Additionally, if the segmentation algorithm changes or is upgraded, the segmented artificial intelligence model may need to be retrained. The training data sets stored in the database module 50 ensure that this can be done at any time. Training data sets can also be generated for training additional artificial intelligence models that are not necessarily part of the present invention. Therefore, the imaging data stored in the database module 50 can also be exported for training other segmentation algorithms.

[0090] Figure 2 is a block diagram illustrating an exemplary embodiment of a computer-implemented method for providing automatic segmentation for non-contrast CT imaging data according to an embodiment of the present invention. The method may preferably be used with respect to Figure 1 The method is implemented by the computing system 100 described above. The method includes multiple steps.

[0091] In step S1, imaging data from computed tomography (CT) is obtained. The imaging data includes segmented contrast CT imaging data D1 and non-contrast CT imaging data D2 of a specific region of interest (e.g., multiple blood vessels from a patient). According to some embodiments of the present invention, the segmented contrast CT imaging data D1 may include imaging data from single-energy computed tomography angiography (CTA). Both the segmented contrast CT imaging data D1 and the non-contrast CT imaging data D2 can be obtained using a CT device, such as Figure 4 the computed tomography device 200 described in

[0092] Prior to the system 100 obtaining the segmented contrast CT imaging data D1, it can be automatically annotated using, for example, conventional segmentation algorithms used in CTA. Figure 1 This can be achieved by deploying a registration algorithm (such as the registration algorithm implemented in the registration unit 210 described in Figure 2 ), which is adapted to register (or align) the segmented contrast CT imaging data D1 and the non-contrast CT imaging data D2.

[0093] This optional registration step is not shown in Figure 1 Figure 2 Alternatively or additionally, the training data for the image domain transfer artificial intelligence model can be based on virtual non-contrast (VNC) imaging data. When data acquisition is performed using photon counting technology, this can be generated from contrast imaging data (e.g., CTA imaging data). The advantage of this training option is that the generated VNC imaging data does not require a registration step. In some embodiments of the present invention, VNC generation is performed using a VNC generation unit 220, such as Figure 1 Figure 2 This optional step of VNC generation is not shown in

[0094]

[0095] In subsequent step S3, segmentation D4 of the imaging data of the non-contrast CT imaging data D2 is generated using the deployment of the segmentation artificial intelligence model. The segmentation artificial intelligence model can be trained using the synthetic segmented non-contrast CT imaging data D3 generated in step S2. The training data set can be obtained from the database module 50 described regarding Figure 1 as described.

[0096] In some embodiments of the present invention, the segmentation artificial intelligence model can be further trained using photon-counting based VNC imaging data. This training data is not generated by an image domain transfer artificial intelligence model (such as DDPM), and thus provides an additional training data source for a training method that does not rely on DDPM.

[0097] The segmentation artificial intelligence model can preferably be adapted to implement a segmentation algorithm. According to the principles of the present invention, this can in particular be a segmentation algorithm capable of processing imaging data from CTA, such as U-Net or a segmentation algorithm with a U-Net-like architecture.

[0098] In step S4, segmentation D4 of the non-contrast CT imaging data D2 is output. Then, this can be evaluated by a medical expert and used as support, for example, for clinical diagnosis or surgical intervention.

[0099] Figure 3 is a schematic depiction of different CT imaging data involved in using the system 100 and / or performing the method of the present invention according to a preferred embodiment.

[0100] Figure 3 Examples of operations of the registration unit 210, the VNC generation unit 220, the image domain transfer module 20, and the segmentation module 30 are shown in different rows.

[0101] In the first row, the operation of the registration unit 210 is illustrated. The registration unit 210 processes imaging data from the segmented contrast CT imaging data D1 (for example, corresponding to CTA imaging data) and the non-contrast CT imaging signal D2. This imaging data may be acquired using a CT device and is not necessarily paired. The registration unit 210 is configured to pair the imaging data, for example, to provide a correspondence between the imaging data at the pixel level. Thus, the resulting imaging data set D2R can be combined with the segmented contrast CT imaging data D1, and both are used as a training data set for the image domain transfer artificial intelligence model.

[0102] In the subsequent rows, the operation of the VNC generation unit 220 is illustrated. In this case, the imaging data of the segmented contrast CT imaging data D1 is acquired with a CT device that supports photon counting technology and can generate multi-energy or spectral CTA imaging data. The VNC generation unit 220 is configured to process the imaging data of the segmented contrast CT imaging data D1 and generate VNCCT imaging data D2VNC therefrom. The resulting imaging data set D2VNC can be combined with the segmented contrast CT imaging data D1, and both are used as training data sets for the image domain transfer artificial intelligence model.

[0103] In the subsequent row, the generation of a synthetic segmented non-contrast CT imaging database D3 from the imaging data of the segmented contrast CT imaging data D1 is illustrated. This is achieved through the operation of the image domain transfer module 20 and corresponds to Figure 2 Step S2 described in .

[0104] In a further row, the training of the segmentation module 30 is illustrated. The synthetic, annotated imaging dataset D3 generated by the image domain transfer module 20 is used as a training dataset in order to train the segmentation module 30 to generate segmentations of the non-contrast CT imaging data D2.

[0105] In the subsequent rows, an exemplary operation of the system 100 of the present invention for segmentation of non-contrast CT imaging data is illustrated. The system 100 obtains imaging data from non-contrast CT imaging data D2 and generates segmented non-contrast imaging data D4 therefor using a trained segmentation algorithm.

[0106] Figure 4 is a schematic illustration of a computed tomography system 500, comprising a computed tomography device 200 and a system 100 of the present invention. The computed tomography device 200 may be any device suitable for performing computed tomography in any variant thereof and preferably acquiring imaging data using photon counting technology. The computed tomography device 200 may in particular be a NAEOTOM Alpha photon counting CT scanner from Siemens Medical of Erlangen, Germany. The system 100 is implemented or connected (wired or wireless) to the computed tomography device 200. The system 100 of the present invention is configured to obtain imaging data acquired by the computed tomography device 200. In some embodiments, the system 100 is provided to the computed tomography device 200 as an application programming interface (API).

[0107] Figure 5 is a schematic block diagram illustrating a computer program product 300 according to an embodiment of the third aspect of the invention. The computer program product 300 comprises an executable program code 350 which, when executed, is configured to perform a method according to any embodiment of the second aspect of the invention, in particular as described with respect to the previous figures.

[0108] Figure 6 is a schematic block diagram illustrating a non-transitory computer-readable data storage medium 400 according to an embodiment of the fourth aspect of the present invention. The data storage medium 400 comprises an executable program code 450 which, when executed, is configured to perform a method according to any embodiment of the second aspect of the present invention, in particular as described with respect to the previous figures.

[0109] The non-transitory computer-readable data storage medium may include or consist of any type of computer memory, in particular semiconductor memory, such as solid-state memory. The data storage medium may also include or consist of a CD, DVD, Blu-ray disc, USB memory stick, etc.

[0110] The previous description of the disclosed embodiments is merely an example of possible implementations, which are provided to enable those skilled in the art to make or use the present invention. Various changes and modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments without departing from the spirit or scope of the present disclosure. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but should be given the widest scope consistent with the principles and novel features disclosed herein. Therefore, the present invention is not limited, except according to the following claims.

Claims

1. A system (100) for providing automatic segmentation of non-contrast computed tomography imaging data, the system (100) comprising: An input data interface (10) configured to obtain imaging data including at least non-contrast computed tomography imaging data (D2); a segmentation module (30) configured to implement a segmentation artificial intelligence model adapted to generate segments (D4) for the obtained non-contrast computed tomography imaging data (D2), wherein the segmentation artificial intelligence model is trained with at least imaging data based on the segmented contrast computed tomography imaging data (D1); as well as An output data interface (40) is configured to output the generated segment (D4).

2. The system of claim 1 , wherein the segmented artificial intelligence model is also trained with photon counting based virtual non-contrast computed tomography imaging data.

3. A system according to any of the preceding claims, wherein the segmented artificial intelligence model is suitable for implementing a segmentation algorithm, in particular a segmentation algorithm capable of processing imaging data from contrast computed tomography angiography.

4. The system according to any of the preceding claims further comprises an image domain transfer module (20), which is configured to implement an image domain transfer artificial intelligence model, which is trained and suitable for generating synthetic segmented non-contrast computed tomography imaging data (D3) based on segmented contrast computed tomography imaging data (D1), wherein the generated segmented non-contrast computed tomography imaging data (D3) is used as a training data set for the training of the segmented artificial intelligence model.

5. The system of claim 4, wherein the image domain transfer artificial intelligence model is based on a denoising diffusion probabilistic model.

6. A system according to claim 4 or 5, wherein the image domain transfer artificial intelligence model is trained using paired imaging data corresponding to segmented computed tomography angiography and non-contrast computed tomography and / or using virtual non-contrast imaging data (D2R) based on photon counting.

7. The system according to any one of claims 4 to 6, wherein the image domain transfer module (20) comprises a registration unit (210) configured to implement a registration algorithm suitable for pairing the segmented computed tomography angiography imaging data (D1) and the non-contrast computed tomography imaging data (D2).

8. The system according to any one of claims 4 to 7, wherein the image domain transfer module (20) includes a VNC generation unit (220), which is configured to generate virtual non-contrast computed tomography imaging data (D2VNC) based on contrast multi-energy computed tomography angiography.

9. The system according to any one of claims 4 to 8, further comprising a database module (50) configured to store the synthetic segmented non-contrast computed tomography imaging data (D3) generated using the image domain transfer artificial intelligence model.

10. The system according to any of the preceding claims, wherein the obtained non-contrast computed tomography imaging data (D2) and the segmented contrast computed tomography imaging data (D1) are derived from a computed tomography scan of a patient's blood vessels.

11. A computer tomography system (500), the computer tomography system (500) comprising: A computer tomography device (200) configured to perform computer tomography and acquire imaging data, preferably using photon counting technology; as well as The system (100) according to any one of claims 1 to 10, wherein the system (100) according to any one of the preceding claims is configured to obtain the imaging data acquired by the computer tomography device (200).

12. A computer-implemented method for providing automatic segmentation of non-contrast computed tomography imaging data, the method preferably being implemented using a system (100) according to any one of claims 1 to 11, the method comprising the following steps: obtaining (S1) imaging data comprising at least non-contrast computed tomography imaging data (D2); generating (S3) segments (D4) for the obtained non-contrast computed tomography imaging data (D2) using a segmented artificial intelligence model, wherein the segmented artificial intelligence model is trained with at least imaging data based on the segmented contrast computed tomography imaging data (D1); and The generated segment (D4) is output (S4).

13. The computer-implemented method of claim 12, further comprising the steps of: An image domain transfer artificial intelligence model is used to generate (S2) synthetic segmented non-contrast computed tomography imaging data (D3) based on segmented contrast computed tomography imaging data (D1), wherein the generated segmented non-contrast computed tomography imaging data (D3) is used as a training data set for the training of the segmented artificial intelligence model.

14. A computer program product (300) comprising an executable program code (350) which, when executed, is configured to perform the computer-implemented method according to claim 12.

15. A non-transitory computer-readable data storage medium (400) comprising executable program code (450) which, when executed, is configured to perform the computer-implemented method of claim 12.

Images