Automated coronary angiography analysis

Abstract

A method and apparatus for analyzing diagnostic image data are provided, wherein multiple acquired images of a vessel of interest, acquired using a predefined acquisition method, are received at a trained classification device and classified by the classification device to extract at least one quantitative feature of the vessel of interest from at least one of the multiple acquired images. The at least one quantitative feature associated with the at least one acquired image is then output while the acquisition of the diagnostic image data continues, and one or more adjustable image acquisition settings are adjusted based on the at least one quantitative feature to optimize the acquisition of the diagnostic image data.

Description

Technical Field

[0001] This invention relates to methods, apparatus, and computer programs for analyzing diagnostic image data (particularly X-ray angiography image data). In particular, this invention relates to improved methods and apparatus that allow the automatic extraction of quantitative feature information from diagnostic image data acquired using predefined acquisition settings, and the use of this extracted quantitative feature information to adjust the predefined acquisition settings accordingly, thereby improving the quality of the acquisition process. Background Technology

[0002] Coronary artery disease is now widespread in many societies. Various types of coronary artery disease are known. In order to develop appropriate treatment plans for each corresponding coronary artery disease, it is crucial to accurately assess the disease, i.e., to have high-quality and reliable medical data to evaluate the coronary artery disease.

[0003] In clinical practice, coronary angiography is a widely accepted method for assessing coronary artery disease. Coronary angiography is typically performed by injecting a contrast agent into the blood vessels and subsequently irradiating the contrast-filled coronary arteries with X-rays to acquire a sequence of angiographic images of these vessels, thus making the coronary vascular system clearly visible. Therefore, the number and orientation of these angiographic image sequences, the contrast agent dosage, and the corresponding analysis of the image data can vary from patient to patient, making objective, comparable analyses for different patients very difficult.

[0004] To overcome these problems, improved acquisition methods have been developed in which predefined acquisition settings are used to acquire angiographic image sequences. Using these predefined acquisition settings reduces the variability of the acquired data, as some acquisition settings are known for each patient.

[0005] One such acquisition method is Xper Swing acquisition, in which angiographic image data is acquired along a predefined repeatable trajectory with a predefined dose of contrast agent in different orientations. Xper Swing acquisition thus provides angiographic image data for analysis as a single image sequence for evaluating a specific coronary artery. However, the quality of angiographic image data acquired using Xper Swing still varies due to the patient-dependent optimization of certain acquisition settings, patient-to-patient variability in anatomy, and (other) variability in acquisition settings.

[0006] As a result, automating data analysis is challenging. That is, automating analysis will require complex calculations that take into account all the variability that may arise in the data of different patients.

[0007] To date, no reliable, repeatable, objective, and efficient method has been developed for the acquisition and analysis of diagnostic image data, such as angiographic image data, especially diagnostic image data acquired using certain predefined acquisition settings. Summary of the Invention

[0008] Therefore, one object of the present invention is to provide a method and apparatus that enables the acquisition and analysis of diagnostic image data in a reliable, repeatable, objective and effective manner.

[0009] More specifically, one object is to provide methods and apparatus for efficiently acquiring diagnostic image data using medical imaging modalities (such as X-ray angiography) in a predetermined acquisition setting to obtain reliable results that allow for objective and patient-independent assessment of coronary artery disease. Even more specifically, an object of the present invention is to provide methods and apparatus that allow for reduced radiation and contrast agent doses delivered to patients during image-based assessment of coronary artery disease.

[0010] This objective is achieved by a method for analyzing diagnostic image data, the method comprising the steps of: receiving diagnostic image data, including multiple acquired images of a vessel of interest, at a trained classification device, the diagnostic image data having been acquired using a predefined acquisition method; classifying the diagnostic image data to extract at least one quantitative feature of the vessel of interest from at least one of the multiple acquired images; outputting the at least one quantitative feature of the vessel of interest associated with the at least one acquired image while the acquisition of the diagnostic image data is still in progress; and adjusting one or more adjustable image acquisition settings based on the at least one quantitative feature to optimize the acquisition of the diagnostic image data.

[0011] In other words, this objective is achieved by employing a trained classification device (such as a convolutional neural network) to automatically analyze the diagnostic image data during its acquisition so as to adjust a set of adjustable acquisition settings (such as certain acquisition parameters) during the ongoing acquisition to optimize data acquisition for specific vascular properties (i.e., for a specific patient).

[0012] The term diagnostic image data can therefore refer to a set of acquired images representing a patient's vascular system. Thus, the term vascular system can refer to a vascular tree or a single vessel. The term vascular system can specifically refer to one or more vessels of interest and / or segments thereof. The term vessel of interest can therefore refer to a vessel in a patient that should be evaluated using diagnostic image data—with regard to potential lesions and / or other diseases. In some embodiments, all acquired images of the diagnostic image data can represent vessels of interest within the coronary vascular system.

[0013] Diagnostic image data may specifically include one or more acquired images of the one or more vessels of interest. Therefore, the term acquired image can generally be understood to refer to a single image acquired for the vessel of interest, whereby multiple acquired images can be included in the diagnostic image data.

[0014] Multiple images can be acquired specifically via medical imaging modalities such as computed tomography (CT), ultrasound (US) imaging, or magnetic resonance (MR) imaging. In some embodiments, the medical imaging modality may specifically correspond to X-ray angiography, or even more specifically to X-ray angiography performed with a set of predefined acquisition settings such as predefined imaging trajectories and predefined doses of contrast agents to be used.

[0015] In some embodiments, the medical imaging modality can be gated. Therefore, gated medical imaging modalities can typically employ gated reconstruction, where the acquisition of images is performed in parallel with the acquisition of data providing information within the cardiac cycle, such as electrocardiogram (ECG) or photoplethysmography (PPG) data. This data can then be used to gate image acquisition and reconstruction using appropriately selected phase points within the cardiac cycle.

[0016] Diagnostic image data is received at a trained classification device. The term "classification device" can specifically refer to a classifier or classification unit integrated into a corresponding apparatus for analyzing diagnostic image data. In some embodiments, the term "classification device" can also refer to a classifier provided separately from the apparatus. In some embodiments, the classification device can specifically be implemented as a convolutional neural network (CNN).

[0017] The classification device is a trained classification device. That is, the classification device has been previously trained using a training dataset that indicates the correlation between diagnostic image data and one or more quantitative features (such as vessel length, vessel location, lesion severity, etc.). Specifically, training is performed using a training dataset that includes diagnostic image data annotated with corresponding quantitative features that can be compared with the diagnostic image data to be classified. Thus, the annotations may have been obtained by manually annotating the diagnostic image data by clinical experts, or, in the case of using a simulated training dataset, may be inherently known standard data.

[0018] The weights and parameters of the classification device (especially a convolutional neural network) are then optimized during training so that, for input to the training dataset, the resulting neural network output numerically approximates the corresponding annotated feature values. In other words, the optimization of the neural network minimizes the difference between the neural network output and the annotated feature values ​​across all training datasets. Therefore, the comparison between the neural network output and the annotated feature values ​​can be achieved using various suitable metrics, such as the L2 norm or generalized dice loss. In some examples, the optimization can use the Adam optimizer.

[0019] During the training of a classification device, commonly known forms of data augmentation, such as image scaling, translation, or contrast alteration, can be used. An exemplary network architecture for such a task could be an encoder-decoder neural network architecture.

[0020] A classification device is used to classify diagnostic image data in order to extract at least one quantitative feature from the diagnostic image data. That is, based on training, the classification device is able to derive the value of at least one quantitative feature of a vessel of interest for one or more acquired images in the diagnostic image data. In some embodiments, a corresponding value for a specific quantitative feature can be derived for each acquired image. Therefore, multiple values ​​for a specific quantitative feature can be derived for multiple acquired images.

[0021] In some embodiments, quantitative features may therefore specifically correspond to features such as vessel length, vessel location, vessel diameter, lesion severity, myocardial imaging value, and visibility score of lesions and / or vessels in images acquired for an individual, i.e., features that can be derived on a per-image basis.

[0022] In some embodiments, quantitative features may alternatively or additionally include values ​​indicating the hydrodynamics through the vessel of interest, such as fractional flow reserve (FFR) values, instantaneous waveless ratio (iFR) values, or coronary flow reserve (CFR) values. Typically, these parameters can be derived from a hydrodynamic model capable of modeling the hydrodynamics through the vessel of interest, as described, for example, in international applications WO2016 / 087396, WO2020 / 053099A1, and WO2019 / 101630A1. In this embodiment, it may be possible to derive the values ​​of these parameters directly from the classification device. That is, by training on a training dataset, the fractional classification device can implicitly learn the fluid properties of (one or more) vessels of interest and therefore the fluid parameters associated with them, without having to simulate or model the fluid flow through (one or more) vessels of interest. This allows for the avoidance of using a hydrodynamic model, and instead, obtaining the fluid parameters directly from the trained classification device.

[0023] On the other hand, quantitative features can also correspond to features related to the overall diagnostic image data, such as an integrity score indicating whether sufficient angular information about the vessel of interest is available for obtaining a reliable analysis, a reference deviation index indicating whether the visible vascular system is similar to the patient's average reference, or an obstruction score indicating whether future tomographic reconstruction is likely to show strong artifacts if the current trajectory continues. Therefore, the obstruction score can be particularly useful when the implant of a particular external device is in the field of view. That is, if a particular trajectory causes the device to obstruct the field of view in the future projection of the planned trajectory, it can be beneficial to change the trajectory to avoid such obstruction.

[0024] Then, quantitative features associated with an acquired image are output. That is, the values ​​of quantitative features already derived based on the corresponding acquired image are associated with said acquired image and then output for further evaluation and / or further processing. This output is performed specifically while image acquisition via the medical imaging modality is still in progress.

[0025] Based on this output, the computing unit or other processing device then evaluates at least one quantitative feature associated with the corresponding acquired image, and its value accordingly, to determine whether the current acquisition settings used result in sufficient image quality. In some embodiments, the computing unit may therefore specifically use quantitative features such as visibility scores, integrity scores, etc. If the evaluation indicates that the current acquisition settings do not produce sufficient acquired images, one or more of the adjustable acquisition settings are adjusted. Thus, adjustment can be performed automatically, particularly based on previous classification.

[0026] The term "adjustable acquisition settings" can therefore specifically refer to acquisition settings that are not predefined due to the medical imaging modality used. Thus, in the current context, a distinction is made between predefined acquisition settings that should not be changed (i.e., kept the same to reduce variability) and adjustable acquisition settings that can be changed according to each patient's individual requirements.

[0027] It is possible to obtain diagnostic image data in a reliable, efficient, and reproducible manner by keeping certain acquisition settings constant and predefined while adjusting other acquisition settings based on automatic analysis of previously acquired diagnostic image data, while reducing the variability of different datasets acquired for different patients at different measurement times.

[0028] More specifically, it is possible to perform "live" adjustments to the acquisition settings by analyzing a subset of the acquired diagnostic image data during acquisition. This allows for individual optimization of these acquisition settings to obtain diagnostic image data in the most efficient manner. This reduces the amount of radiation and contrast agent doses delivered to patients, as sufficient integrity can be achieved more quickly for some patients than for others.

[0029] In some embodiments, the method can be implemented to perform out-of-distribution detection. That is, the method can be implemented to determine whether diagnostic image data input to a computing unit or other processing device falls within a distribution expected based on the training of the classification device. This allows detection of whether diagnostic image data input to the computing unit or other processing device is not placed in relation to the types of diagnostic data that the classification device has already trained on.

[0030] In this case, an indication can be output to the user that the acquired diagnostic image data cannot be properly evaluated, since it is unrelated to the type of diagnostic image data expected to be evaluated. This indication can be a simple warning that the diagnostic image data cannot be evaluated or can only be evaluated incorrectly. Alternatively or additionally, the indication may include a suggestion to perform new or additional diagnostic image data acquisition. In some embodiments, alternatively or additionally, the method can be implemented to perform an evaluation of the diagnostic image data, whereby the output of the evaluation can be accomplished using large error bars.

[0031] In some embodiments, adjusting the one or more adjustable image acquisition settings includes prematurely terminating the acquisition of the diagnostic image data if it is determined that the acquired portion of the diagnostic image data meets at least one predefined reliability criterion.

[0032] In some embodiments, adjusting the adjustable acquisition settings may specifically include terminating acquisition before its planned end if it is determined that sufficient diagnostic information has been obtained. That is, the acquired diagnostic image data is divided into two or more subsets of diagnostic image data, thereby evaluating the first subset while currently acquiring the second subset. The size of each subset can depend considerably on the given medical imaging modality and acquisition quality. A single acquired image can form a subset. In other embodiments, more acquired images can form a subset of diagnostic image data.

[0033] The evaluation of the first subset includes determining whether the diagnostic information derived from it meets a predefined reliability criterion, i.e., whether there is sufficient angular information to provide a reliable assessment of the vessel of interest. In some embodiments, the reliability criterion may be quantified specifically based on an integrity score. That is, a threshold for the integrity score may be determined, and sufficient angular information is considered available for a reliable diagnosis as long as the integrity score is above the threshold. Alternatively or additionally, the reliability criterion may include other scores and / or criteria.

[0034] If sufficient information is determined to be available, the acquisition settings to be adjusted can specifically be the acquisition end time. Even more specifically, the acquisition end time can be set, for example, by means of a termination signal, such that acquisition is terminated immediately. By terminating acquisition as soon as sufficient information is available, it is possible to keep the radiation dose experienced by the patient as low as possible. On the other hand, if it is determined that the reliability criterion is not met, i.e., insufficient information is available, the measurement can continue, i.e., the adjustment of the adjustable acquisition settings is not performed. This feedback loop of allowing the adjustment of the acquisition time can be repeated frequently until acquisition stops due to sufficient information being available.

[0035] In some embodiments, adjusting the one or more adjustable acquisition settings includes adjusting the image acquisition trajectory to improve the visibility of the vessel of interest in the diagnostic image data. In some modifications, adjusting the one or more adjustable acquisition settings includes adjusting the contrast agent injection rate into the vessel of interest during image acquisition.

[0036] In some embodiments, adjustment of the adjustable image acquisition settings may additionally or alternatively include adjustment of the imaging trajectory used for image acquisition. In this case, a visibility score of the vessel of interest and / or lesions therein is determined for a first subset of diagnostic image data. For this purpose, the entire diagnostic image data is considered, not just individual acquired images. The visibility score can be compared to a preset reference value or threshold, whereby visibility is considered adequate if the score is above (or below) the value, and poor if the score is below (or above) the value: if the visibility score indicates insufficient visibility, i.e., poor visibility, then adjustment of the adjustable imaging settings may specifically include adjustment of the image acquisition trajectory used to acquire the acquired images. This allows for improved image quality, meaning fewer acquired images are needed to obtain sufficient diagnostic information. This effectively reduces the radiation dose delivered to the patient. The adjusted trajectory also avoids the need to diagnose images with non-ideal visibility.

[0037] In some embodiments, the adjustment of one or more adjustable acquisition settings can also encompass the adjustment of the contrast agent injection rate in the vessel of interest. That is, the contrast of the vessel of interest can be determined using a classification device for a first subset of diagnostic image data. By examining the contrast, it can be determined whether sufficient contrast agent has been injected into the vessel of interest. Therefore, the amount of contrast agent can vary from patient to patient, as patients with narrower vessels may require less contrast agent than those with wider vessels to achieve similar visibility. Thus, based on the contrast of the vessel of interest, it can be evaluated whether sufficient contrast agent is present in the vessel of interest, and consequently whether the contrast agent injection rate is adequate or should be adjusted due to too little or too much contrast agent currently being injected into the vessel of interest.

[0038] In this context, the adjustment of the adjustable acquisition settings therefore includes adjusting the contrast agent injection rate based on the properties of the vessel of interest. This adjustment allows for optimization of the contrast agent dose delivered to each patient.

[0039] These examples can also be used in feedback loops, which allow for frequent evaluation of individual subsets of diagnostic image data during acquisition in order to perform real-time adjustments to the acquisition settings.

[0040] In some embodiments, the method further includes obtaining training image data of the blood vessel of interest according to the predefined acquisition method; extracting the at least one quantitative feature from the training image data; generating at least one training dataset for the classification device, the training dataset including the training image data associated with the at least one quantitative feature; and using the at least one training dataset to train the classification device.

[0041] The classification device can be trained using a corresponding training dataset. In some embodiments, these training datasets can be derived from training image data. The term training image data can therefore specifically refer to multiple training image datasets (i.e., measurement data) that have been acquired in a clinical setting, or multiple training images that have been generated through simulation.

[0042] Subsequently, one or more quantitative features can be extracted from individual training images and the overall training image data. Therefore, whether to use individual images or the overall data depends on the corresponding quantitative values. Thus, feature extraction can be performed manually by one or more users, automatically by a corresponding algorithm, or it can correspond to quantitative features readily available from simulations of the data.

[0043] The extracted quantitative features and training image data are used to generate a corresponding training dataset. That is, the quantitative feature values ​​are correlated with the corresponding training images and / or training image data to derive the correlation between the quantitative feature values ​​and the corresponding image data. The resulting training dataset can then be used to train a classification device.

[0044] In another embodiment, the training image data includes simulated training image data generated by simulating image acquisition according to the predefined acquisition method, wherein the simulation includes the following steps: obtaining at least one three-dimensional geometric model of the vessel of interest; obtaining at least one two-dimensional background image of the vessel of interest; and simulating contrast agent hydrodynamics through the patient's vascular system based on at least one contrast agent fluid parameter. In some modifications, the simulation further includes obtaining deformation translation and rotation data, and enhancing the simulated training image data based on the translation and rotation data. In some modifications, generating the at least one training dataset further includes the steps of receiving additional patient data and adjusting the at least one training dataset based on the additional patient data.

[0045] In some embodiments, training image data is generated by means of simulation. For this purpose, at least one three-dimensional geometric model of the vascular system of a patient, including the vessels of interest, is obtained. Thus, the geometric model can be obtained from medical images, which can have been acquired by any medical imaging modality that allows the acquisition of three-dimensional medical images.

[0046] In some embodiments, the medical imaging modality may correspond to a medical imaging modality for which reality-based adjustments should be performed. In some embodiments, the medical imaging modality may be a different imaging modality. The geometric model may also be purely virtual and defined by common anatomical knowledge. Furthermore, at least one two-dimensional background image of the vascular system of a patient, including the vessels of interest, is acquired. With the aid of the background image, it is possible to distinguish the background and the vascular system in the medical image in order to appropriately perform vessel identification of the vessels in the vascular system.

[0047] Background images can also provide a realistic look to the simulated data. Two-dimensional background images can therefore be obtained from actual clinical acquisitions, and / or can be constructed from forward projections of three-dimensional medical images, and / or can be virtual images designed to mimic the typical backgrounds seen in the diagnostic data to be simulated.

[0048] Furthermore, three-dimensional medical images and / or two-dimensional background images can be used to generate a fluid dynamics model representing the fluid dynamics of the patient's vascular system. In some embodiments, the fluid dynamics model may specifically include a lumped parameter model.

[0049] The term lumped parameter model can specifically refer to a model in which the hydrodynamics of blood vessels is approximated by the topology of discrete entities. As an example, a vascular system such as a vascular tree can be topologically represented by resistive elements, each with a specific resistance. Thus, the outlet at the distal end of the blood vessel is also represented by a specific resistive element. This resistive element is then connected to ground, such as to represent the terminal end of the blood vessel. Similarly, corresponding resistive elements can be connected to a series of resistive elements representing the blood vessel of interest, such as to represent the outflow from the blood vessel of interest at certain bifurcations. These resistive elements can also typically be connected to ground as well.

[0050] Compared to other methods such as Navier-Stokes, these lumped parameter models reduce the number of dimensions. Therefore, using lumped parameter models allows for simplified calculations of the fluid dynamics within blood vessels, and can ultimately lead to reduced processing time. For example, the adoption of such lumped parameter models is described in international application WO2016 / 087396.

[0051] The resulting hydrodynamic model can then be used to simulate the flow of contrast agent fluid through the patient's vascular system, and particularly through one or more vessels of interest. This allows for the generation of training image data representing the vascular system and the corresponding hydrodynamics through it. In some embodiments, to enhance the training image data, deformation translation and rotation data can be added as additional information to the simulation. The resulting training image data can then be provided to a classification device for training.

[0052] In some embodiments, at least one quantitative feature includes one or more of the following: vascular label of the blood vessels in the patient's vascular system and / or vascular length of the blood vessels in the patient's vascular system and / or severity of lesions in the blood vessels in the patient's vascular system and / or vascular diameter of the blood vessels in the patient's vascular system and / or visibility score of lesions and / or integrity score of at least one of the multiple acquired images and / or myocardial contrast value.

[0053] In some embodiments, additional patient information, such as ECG data, aortic pressure values, or historical data for a specific patient, may be added to the training dataset and / or classification. This can have the further benefit of detecting additional patient abnormalities, such as strongly elevated aortic pressure, in which case the injection of contrast agent may also need to be adjusted.

[0054] According to some embodiments, outputting the at least one quantitative feature for further evaluation includes the steps of: displaying the at least one quantitative feature to a user, and / or outputting the at least one quantitative feature in a predefined format for automatic reporting to a reporting entity. In some embodiments, the user may input additional data in response to the output, whereby the additional data may also be used to train a classification device and / or evaluate diagnostic image data.

[0055] According to another aspect, an apparatus for analyzing diagnostic image data is provided, the apparatus comprising a trained classification device and a computing unit, the trained classification device being configured to receive diagnostic image data including multiple acquired images of a vessel of interest, the diagnostic image data having been acquired using a predefined acquisition method; classifying the diagnostic image data to extract at least one quantitative feature of the vessel of interest from at least one of the multiple acquired images; and outputting the at least one quantitative feature of the vessel of interest associated with the at least one acquired image, while the acquisition of the diagnostic image data is still in progress, the computing unit being configured to adjust one or more adjustable image acquisition settings based on the at least one quantitative feature to optimize the acquisition of the diagnostic image data.

[0056] In some embodiments, the apparatus further includes an input unit and a training dataset generation unit. The input unit is configured to obtain training image data of the vessel of interest according to the predefined acquisition method. The training dataset generation unit is configured to: extract at least one quantitative feature of the vessel of interest from the training image data; generate at least one training dataset for the classification device, the training dataset including the training image data associated with the at least one quantitative feature; and provide the at least one training dataset to the classification device for training. In some embodiments, the apparatus may further include a display unit and a user interface. The display unit is configured to generate a graphical representation of at least one of the multiple acquired images and / or the at least one quantitative feature. The user interface is configured to receive user input in response to the graphical representation.

[0057] In another aspect, a computer program for controlling an apparatus according to the invention is provided, which, when executed by a processing unit, is adapted to perform the method steps according to the invention. In yet another aspect, a computer-readable medium is provided on which the aforementioned computer program is stored.

[0058] It should be understood that the method according to claim 1, the apparatus according to claim 11, the computer program according to claim 14, and the computer-readable medium according to claim 15 have similar and / or identical preferred embodiments, particularly as defined in the dependent claims.

[0059] It should be understood that the preferred embodiments of the present invention can also be any combination of the independent claims or the above embodiments and the corresponding dependent claims.

[0060] These and other aspects of the invention will become apparent and will be explained with reference to the embodiments described below. Attached Figure Description

[0061] In the following figures:

[0062] Figure 1 An apparatus for analyzing diagnostic image data of blood vessels according to an embodiment is schematically illustrated.

[0063] Figure 2 A flowchart of a method for analyzing diagnostic image data according to an embodiment is shown.

[0064] Figure 3 A flowchart of a method for generating training data according to an embodiment is shown. Detailed Implementation

[0065] The illustrations in the accompanying drawings are schematic. In different drawings, similar or identical elements are provided with the same reference numerals.

[0066] Figure 1 An exemplary embodiment of a device 1 for analyzing diagnostic image data is schematically illustrated. Device 1 includes an input unit 100, a training dataset generation unit 200, a classification unit 300, a calculation unit 400, and a display unit 500. Furthermore, the classification unit 300 and the communication unit 400 are communicatively coupled to a medical imaging modality 2 in a feedback loop 600.

[0067] Input unit 100 is configured to receive training image data 10 of the patient's vascular system. The training image data 10 may specifically correspond to or include image data previously acquired using a predefined acquisition method (i.e., an acquisition method performed using one or more predefined (known) acquisition settings, such as known contrast agent doses and acquisition trajectories). In accordance with... Figure 1 In a specific embodiment, the training image data 10 may specifically correspond to clinical data already acquired using X-ray angiography with a C-arm. That is, in Figure 1 In a specific embodiment, the training image data 10 has been derived from the actual measurement data.

[0068] However, it should be understood that, alternatively or additionally, the training image data 10 may also have been generated using simulation, etc. For the simulated training image data 10, typically three-dimensional medical images acquired using CT and / or MR imaging modalities can be acquired to generate a three-dimensional model of the vessel of interest and combined with two-dimensional background data showing an image of the heart without any contrast agent filling and without arteries. The training image data and corresponding training dataset are generated based on contrast agent injection parameters for a predefined acquisition method and the corresponding hydrodynamic model. Thus, deformation translation and rotation can be added to the three-dimensional representation of the vessel of interest and the two-dimensional background projection to achieve data augmentation. The entire range of the acquisition trajectory is then typically covered by a cardiac motion model.

[0069] Input unit 100 provides training image data 10 to training dataset generation unit 200. Training dataset generation unit 200 is configured to extract one or more quantitative features of the patient's vascular system, and particularly the vessels of interest, from the training image data 10. In some embodiments, these quantitative features may specifically relate to vessel labels, vessel numbers, vessel locations and / or vessel lengths, severity of one or more lesions in one or more vessels of interest, myocardial contrast values, vessel diameter of the vessel of interest, visibility scores of lesions in the vessels of interest for each individual training image of the training image data, integrity scores indicating whether sufficient angular information for a given vessel is available to allow for reliable analysis, reference deviation indices indicating that the visible vascular system is similar to a reference, etc.

[0070] Using the extracted quantitative features, the training dataset generation unit 200 is configured to generate at least one training dataset comprising training image data 10 and corresponding predefined features associated with one or more training images in the training image data 10. Thus, the training dataset generation unit 200 obtains the correlation between the training image data 10 and the extracted predefined features, and generates a corresponding dataset including the relevant information. This corresponding dataset is then provided as training dataset 20 to the classification unit—or classification device—300.

[0071] The classification unit 300 includes an input port 301 configured to receive a training dataset from the training dataset generation unit 200. The classification device 300 uses the training dataset 20—or optionally multiple training datasets 20—to train quantized relationships between features and training images in the training image data 10. Figure 1 In an exemplary embodiment, the classification unit includes or corresponds to a convolutional neural network, and in some embodiments, a deep convolutional neural network. That is, the classification unit 400 implements multiple convolutional layers in combination with pooling layers.

[0072] Input to According to Figure 1 In a specific embodiment, the training dataset 20 in classification unit 300 corresponds to multiple training images acquired using X-ray angiography. Specifically, in Figure 1 In a specific embodiment, 10,000 individual angiographic images are used as corresponding training images. These training images are provided with corresponding feature data based on a pixel mask provided for each individual angiographic image, whereby each pixel is classified as belonging to an artery such as the left anterior descending artery (LAD), left circumflex artery (LCX), obtuse marginal branch (OM), right coronary artery, etc., or as belonging to the background. Alternatively or additionally, the training images may be provided with feature data including the minimum diameter of the indicator artery and / or a single value indicating that (parts) of the indicator artery are not visible for each angiographic image.

[0073] During training, 300 classification devices (based on...) Figure 1 In a specific embodiment, the weights and parameters of the neural network are optimized such that, for the input training dataset 20, the obtained neural network output is numerically close to the corresponding annotation feature value. That is, the optimization of the neural network minimizes the difference between the neural network output and the annotation feature value across all training datasets. Therefore, the comparison between the neural network output and the annotation feature value can be achieved using various suitable metrics, such as, for example, the L2 norm or generalized dice loss. According to... Figure 1 In a specific embodiment, optimization may specifically utilize the Adam optimizer.

[0074] During training, typical known forms of data augmentation can be used, such as image scaling, translation, or contrast alteration. An exemplary network architecture for such a task could be an encoder-decoder neural network architecture.

[0075] During training using training dataset 20, classification unit 300 is configured to receive a first subset of diagnostic image data 30 acquired for a specific patient from medical imaging modality 2 via input port 302. This first subset of diagnostic image data 30 may specifically include multiple acquired images 31 acquired using a predefined acquisition method, whereby the predefined acquisition method corresponds to a predefined acquisition method used for training image data, to ensure that classification unit 300 has been trained with an appropriate training dataset to accurately classify the diagnostic image data 30.

[0076] exist Figure 1In a specific embodiment, the input to the classification device corresponds to multiple acquired images 31 in the diagnostic image data 30, each acquired image 31 corresponding to a single two-dimensional X-ray angiography image. Alternatively, the multiple acquired images 31 may also correspond to a chronological stacking of multiple two-dimensional angiography images, such as corresponding C-arm angles. That is, the input to the classification device corresponds to the same diagnostic image data 30 presented to a user (such as a physician) for visual viewing.

[0077] After classifying the diagnostic image data, at least one quantitative feature suitable for analyzing the diagnostic image data 30 is extracted from the diagnostic image data 30.

[0078] The extracted quantitative feature values ​​and a first subset of diagnostic image data 30, including one or more acquired images 31, are then provided to the computing unit 400 for further processing. It should be understood that while the first subset of diagnostic image data 30 is provided to the computing unit 400 for further processing, the acquisition of a second subset of diagnostic image data 30 is still in progress. This allows for the adjustment of image acquisition as possible and / or necessary using the evaluation by the computing unit 400.

[0079] In other words, the computing unit 400 determines whether adjusting the acquisition parameters used for image acquisition might be beneficial based on a first subset of the diagnostic image data and the extracted quantitative features. Figure 1 In a specific example, the computing unit 400 derives for this purpose reliability criteria for the diagnostic information to be derived from the diagnostic image data 30 and quantitative features.

[0080] Furthermore, the computing unit 400 processes a first subset of the diagnostic image data 30 and the quantitative features derived therefrom, and determines whether a reliability criterion is met. In a specific embodiment, this is achieved by comparing the diagnostic information derived from the first subset of the diagnostic image data 30 and the quantitative features with a threshold value indicating the sufficiency of the diagnostic information.

[0081] If sufficient diagnostic information can be derived, the reliability criterion is considered met. In this case, the computing unit 400 is then configured to adjust the adjustable image acquisition settings by outputting a corresponding termination signal to medical imaging modality 2, i.e., to terminate the acquisition prematurely, before the termination point of its original settings. In other words, in response to the termination signal, medical imaging modality 2 terminates further image acquisition, thereby avoiding the delivery of unnecessary radiation and contrast agent doses to the patient.

[0082] On the other hand, if the computing unit 400 determines that the reliability criterion is not met, i.e., there is not enough information available, the computing unit 400 will not output any termination signal to the medical imaging modality 2, and the medical imaging modality 2 will continue to acquire a second subset of diagnostic image data.

[0083] It should be understood that the above evaluation process can be repeated for a second subset (and any subsequent subset) of the diagnostic image data 30 until the reliability criteria are met. Thus, the computing unit 400 can terminate the acquisition process once sufficient diagnostic information is determined to be available.

[0084] In some embodiments, adjusting the adjustable image acquisition settings may additionally or alternatively include adjusting the imaging trajectory used for image acquisition. In this case, evaluating a first subset of the diagnostic image data 30 may include determining a visibility score for the vessel of interest in the individual acquired images. If the calculation unit 400 records poor visibility, the calculation unit 400 may be configured to automatically adjust the imaging trajectory to improve the visibility of the vessel of interest. By adjusting the imaging trajectory to improve visibility, fewer acquired images 31 are needed to obtain sufficient diagnostic information, thereby optimizing the radiation dose delivered to the patient.

[0085] In other embodiments, the computing unit 400 may also evaluate a first subset of the diagnostic image data 30 and the extracted quantitative features to determine the contrast of the vessel of interest. This allows determination of whether sufficient contrast agent has been injected into the vessel of interest. The amount of contrast agent required to provide sufficient visibility of one or more vessels of interest may vary from patient to patient. Thus, patients with narrower vessels may require less contrast agent, while patients with wider vessels may require more contrast agent to achieve similar visibility. Therefore, the computing unit 400 may be configured to adjust the contrast agent injection rate based on the nature of the vessel of interest as an additional adjustable acquisition setting, whereby a lower rate is used for patients with narrower vessels (i.e., requiring less contrast agent) and a higher rate is used for patients with wider vessels (i.e., requiring more contrast agent). With this adjustment, the contrast agent dose delivered to each patient can be optimized.

[0086] It should be understood that, for these examples, the above evaluation process can also be repeated for a second subset (and any subsequent subset) of the diagnostic image data 30 until the reliability criterion is met, i.e. until sufficient diagnostic information is available.

[0087] It should also be understood that different adjustment procedures can be combined with each other. In this way, the computing unit 400 can be configured to adjust the contrast agent injection rate according to the patient's corresponding vascular characteristics, and terminate the acquisition procedure once sufficient diagnostic information is determined to be available.

[0088] This arrangement enables real-time adjustment of acquisition parameters to optimize the feedback loop for diagnostic image data acquisition.

[0089] exist Figure 1 In an exemplary embodiment, diagnostic image data 30 and extracted features are also provided to display unit 500. Display unit 500 may specifically include screen 501 for displaying information graphically and user interface 502 configured to allow user input and generally operate the device, such as a keyboard, touchpad, mouse, touch screen, etc.

[0090] Display unit 500 is configured to generate a graphical representation of image data 30 and extracted quantitative features and present this information to a user on screen 501. The user can then view the presented information and provide corresponding input via user interface 502. The user input can then be used for further evaluation of the data. In some embodiments, the user input can also be used to return to the trained classification unit 300 and be used by the classification unit for further training.

[0091] Figure 2 It shows the use of according to Figure 1 A flowchart of a method 1000 for analyzing diagnostic image data using device 1. In step S101, input unit 100 receives data that may have already been processed as described above. Figure 3 The training image data 10 was generated as explained. Alternatively or additionally, the training image data may have been generated by different means.

[0092] In step S102, input unit 100 provides training image data 10 to training dataset generation unit 200. In step S201, training dataset generation unit 200 receives training image data 10, and in step S202, extracts one or more quantitative features of the patient's vascular system from the training image data 10, particularly one or more quantitative features of one or more vessels of interest in the patient's vascular system. Optionally, in step S203, training dataset generation unit 200 correlates one or more quantitative features with training image data 10.

[0093] In step S204, the training dataset generation unit 200 generates at least one training dataset. This at least one training dataset 20 is provided to the classification unit 300 in step S205.

[0094] In step S301, classification unit 300 receives training dataset 20 from training dataset generation unit 200. In step S302, classification unit 300 then uses training dataset 20 to perform the following... Figure 1 Train as described.

[0095] Subsequently, in step S303, the classification unit 300 receives a first subset of diagnostic image data 30 acquired by the medical imaging modality 2. In step S304, the classification unit 300 classifies multiple acquired images in the first subset of the diagnostic image data 30 to extract at least one quantitative feature from at least one acquired image 31 of the diagnostic image data 30, specifically at least one value of at least one quantitative feature.

[0096] Therefore, classification is performed based on previous training using training dataset 20. In step S305, classification unit 300 provides diagnostic image data 30 and the extracted features to computing unit 400 for further processing. Thus, it should again be understood that a first subset of diagnostic image data 30 is provided to computing unit 400 for further processing, while the acquisition of a second subset of diagnostic image data 30 is still in progress, allowing image acquisition to be adjusted using the evaluation by computing unit 400.

[0097] For this purpose, the computing unit 400 receives a first subset of the diagnostic image data 30 and the extracted quantitative features in step S401, and evaluates the received data in step S402 to determine whether adjustment of one or more adjustable acquisition settings may be necessary. Figure 2 In a specific embodiment, this means that the computing unit 400 determines the visibility score of the vessel of interest in each of the individual acquisition images 31 in the diagnostic image data 30 based on a first subset of the diagnostic image data 30. In step S403, the computing unit 400 then compares the visibility score with a reference value to determine whether the visibility of the vessel of interest is sufficient or whether it needs to be improved.

[0098] In the first case (“Y”), i.e., when visibility is sufficient, the method proceeds to step S404. If the second case (“N”) is present in step S403, i.e., when improved visibility is required, the method proceeds to step S404'. In step S404', the computing unit 400 determines an optimized imaging trajectory to improve visibility, and in step S407 generates a corresponding adjustment signal and provides the adjustment signal to the medical imaging modality 2 to automatically adjust the imaging trajectory. In response to the adjustment signal, the medical imaging modality adjusts the imaging trajectory for acquiring a second subset of the diagnostic image data 30. Then, starting from step S303, the method is repeated in the loop with the second subset of the diagnostic image data 30.

[0099] As indicated above, if visibility is determined to be sufficient (“Y”) in step S403, the method proceeds to step S404. In step S404, the calculation unit 400 evaluates a first subset of the diagnostic image data 30 to determine whether sufficient diagnostic information can be derived from the accumulation of acquired images 31 in the first subset of the diagnostic image data 30. If this is the case (“Y”), the calculation unit 400 generates a termination signal in step S545 and provides the termination signal to the medical imaging modality 2. This results in the termination of image acquisition in step S406.

[0100] If this is not the case (“N”), i.e., if it is determined in step S404 that the diagnostic information derived from the accumulation of acquired images 31 in the first subset of diagnostic image data 30 is insufficient, the method proceeds to step S405'. In step S405', no termination signal is generated, and the method proceeds to continue receiving and processing a second subset (or subsequent subset) of diagnostic image data 30 in steps S303 to S404. With this arrangement, a feedback loop is implemented, which allows for real-time adjustment of acquisition parameters to optimize diagnostic image data acquisition.

[0101] Figure 3 A flowchart of a method S2000 for generating training image data according to an embodiment is shown. In step S2001, a simulation unit obtains at least one medical image of a patient and generates a three-dimensional geometric model of the patient's vascular system from it. The medical image may have been obtained specifically through a medical imaging modality. The medical imaging modality may correspond to medical imaging modality 2, or it may be a different imaging modality.

[0102] In step S2002, the simulation unit also obtains at least one two-dimensional background image of the patient's vascular system in order to accurately distinguish the background from the vascular system.

[0103] In step S2003, the simulation unit then performs vessel identification and identifies blood vessels in the vascular system. Furthermore, in step S2004, the simulation unit uses three-dimensional medical images and / or two-dimensional background images to generate a hydrodynamic model of blood flow through the patient's vascular system. Figure 3 In certain embodiments, the fluid dynamics model may include or correspond to a lumped parameter model, i.e., a model in which the fluid dynamics of blood vessels are approximated by the topology of discrete entities.

[0104] In step S2005, the model is used to simulate the flow of contrast agent fluid through the patient's vascular system. In step S2006, the simulation unit may optionally also receive deformation translation and rotation data as additional information. In step S2007, the simulation unit can then use the additional information to enhance the training image data. In step S2008, the training image data is output to the classification unit 400.

[0105] Although the training data in the above embodiments has been generated based on simulations using a fluid dynamics model, it should be understood that the training data can also be derived from the historical clinical data of one or more patients.

[0106] Although the adjustment of the adjustable acquisition parameters in the above embodiments involves changes to the imaging trajectory and termination of the acquisition process, it should be understood that other types of adjustments can be made automatically based on the classification of the received diagnostic image data, such as the adjustment of the radiation dose to be delivered to the target area and / or the adjustment of the injection rate into the vessel of interest.

[0107] Those skilled in the art, through studying the accompanying drawings, the disclosure, and the claims, can understand and implement other variations of the disclosed embodiments when practicing the claimed invention.

[0108] In the claims, the word "comprising" does not exclude other elements or steps, and the words "a" or "an" do not exclude a plurality.

[0109] A single unit or device can perform the functions of several items recited in the claims. Although some measures are recited in different dependent claims, this does not mean that combinations of these measures cannot be used advantageously.

[0110] Procedures that might be interpreted as being performed by a single unit, such as generating training datasets, training classification devices, classifying image data, or simulating training image data to generate training image data, can also be performed by multiple units. Furthermore, some processes can be performed by the same unit rather than by separate units.

[0111] Computer programs can be stored / distributed on suitable media, such as optical storage media or solid-state media supplied together with or as part of other hardware, but computer programs can also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems.

[0112] No reference numerals in the claims should be construed as limiting the scope.

[0113] This invention relates to a method for analyzing diagnostic image data, comprising the following steps: receiving diagnostic image data, including multiple acquired images of a vessel of interest, at a trained classification device, the diagnostic image data having been acquired using a predefined acquisition method; classifying the diagnostic image data to extract at least one quantitative feature of the vessel of interest from at least one of the multiple acquired images; outputting the at least one quantitative feature of the vessel of interest associated with the at least one acquired image, while the acquisition of the diagnostic image data is still in progress; and adjusting one or more adjustable image acquisition settings based on the at least one quantitative feature to optimize the acquisition of the diagnostic image data.

Claims

1. A method for analyzing diagnostic image data, comprising: Diagnostic image data, including multiple acquired images of the vessel of interest, is received at a trained classification device. This diagnostic image data has been acquired using a predefined acquisition method. The diagnostic image data is classified to extract at least one quantitative feature of the vessel of interest from at least one of the multiple acquired images. While the acquisition of the diagnostic image data is still in progress, output at least one quantitative feature of the vessel of interest associated with the at least one acquired image, and One or more adjustable image acquisition settings are adjusted based on the at least one quantitative feature to optimize the acquisition of the diagnostic image data.

2. The method according to claim 1, wherein, Adjusting the one or more adjustable image acquisition settings includes: If it is determined that the acquired portion of the diagnostic image data meets at least one predefined reliability criterion, then the acquisition of the diagnostic image data is terminated prematurely.

3. The method according to claim 1, wherein, Adjusting the one or more adjustable image acquisition settings includes: Adjust the image acquisition trajectory to improve the visibility of the vessel of interest in the diagnostic image data.

4. The method according to claim 1, wherein, Adjusting the one or more adjustable image acquisition settings includes: Adjust the injection rate of contrast agent into the vessel of interest during image acquisition.

5. The method according to claim 1, further comprising: Training image data of the vessel of interest is obtained according to the predefined acquisition method, and at least one quantitative feature is extracted from the training image data. Generate at least one training dataset for the classification device, the training dataset comprising the training image data associated with the at least one quantitative feature, and The classification device is trained using the at least one training dataset.

6. The method according to claim 5, wherein, The training image data includes simulated training image data generated by simulating image acquisition according to the predefined acquisition method, wherein the simulation includes: Obtain at least one three-dimensional geometric model of the blood vessel of interest; Obtain at least one two-dimensional background image of the blood vessel of interest; and The contrast agent fluid dynamics through the patient's vascular system are simulated based on at least one contrast agent fluid parameter.

7. The method according to claim 6, wherein, The simulation also includes: Obtain deformation translation and rotation data. The simulated training image data is enhanced based on the translation and rotation data.

8. The method according to claim 5, wherein, Generating the at least one training dataset also includes: Receive additional patient data, and The at least one training dataset is adjusted based on the additional patient data.

9. The method according to claim 1, wherein, The at least one quantitative feature includes one or more of the following: vascular label of the vessel in the patient's vascular system, and / or vessel length of the vessel in the patient's vascular system, and / or severity of lesions in the vessel in the patient's vascular system, and / or vessel diameter of the vessel in the patient's vascular system, and / or visibility score of lesions and / or vessels in the patient's vascular system, and / or integrity score for at least one of the acquired images, and / or myocardial contrast value.

10. The method according to claim 1, wherein, Outputting at least one quantitative feature for further evaluation includes: Display the at least one quantitative feature to the user, and / or The at least one quantitative feature is output in a predefined format for automatic reporting to the reporting entity.

11. An apparatus for analyzing diagnostic image data, comprising: A trained classification device, wherein the trained classification device is configured to: Receive diagnostic image data including multiple acquired images of the vessel of interest, said diagnostic image data having been acquired using a predefined acquisition method. The diagnostic image data is classified to extract at least one quantitative feature of the vessel of interest from at least one of the multiple acquired images, and While the acquisition of the diagnostic image data is still in progress, output at least one quantitative feature of the vessel of interest associated with the at least one acquired image, and A computing unit is configured to adjust one or more adjustable image acquisition settings based on the at least one quantitative feature to optimize the acquisition of the diagnostic image data.

12. The apparatus of claim 11, further comprising: An input unit is configured to obtain training image data of the blood vessel of interest according to the predefined acquisition method; A training dataset generation unit is configured to: extract at least one quantitative feature of the blood vessel of interest from the training image data; generate at least one training dataset for the classification device, the training dataset including the training image data associated with the at least one quantitative feature; and provide the at least one training dataset to the classification device for training.

13. The apparatus of claim 11, further comprising: The display unit is configured to generate a graphical representation of at least one of the multiple acquired images and / or the at least one quantitative feature, and The user interface is configured to receive user input in response to the graphical representation.

14. A computer program product for controlling an apparatus according to any one of claims 11 to 13, wherein the computer program product, when executed by a processing unit, is adapted to perform the method according to claims 1 to 10.

15. A computer-readable medium having thereon a computer program product according to claim 14.

Images