System and method for analysis of medical image data based on interaction of quality metrics

Abstract

The present disclosure relates to a system for analyzing medical image data representing a two-dimensional or three-dimensional medical image. The system reads and / or determines a plurality of image quality metrics for the medical image and determines a combined quality metric based on the image quality metrics. The system is further configured such that the determination of the combined quality metric takes into account interactions between the image quality metrics for their combined effect on the combined quality metric.

Description

[Technical Field]

[0001] The present invention relates to a system, method, program element, and computer-readable medium for the analysis of medical image data. In particular, the analysis is performed to consider interactions between multiple image quality metrics for the combined effect of the image quality metrics on a combined quality metric. [Background technology]

[0002] In radiology, the functions of generating radiological images and assessing their diagnostic quality are divided between radiologists, called radiographers, who operate the image acquisition units. If an image shows severe degradation of image quality, caused for example by errors in patient positioning relative to the imaging unit, the radiologist, when performing a rejection analysis, will classify the image as unusable for diagnosis and reject the image. In such cases, a repeat image must be acquired from the patient. Summary of the Invention [Problem to be solved by the invention]

[0003] However, repeated imaging incurs additional costs, reduces radiology department productivity, increases patient radiation dose, and does not fit into clinical workflows that must manage ever-increasing volumes of diagnostic data.

[0004] A similar problem is that some radiologists tend to reject images without consulting a radiologist, anticipating the possibility of rejection by the radiologist, resulting in unnecessary repeat imaging because the images were rejected by someone with only basic medical training who is not qualified to judge whether the images are suitable for diagnosis.

[0005] Because the current process for generating medical images and assessing their diagnostic quality is flawed, it would be desirable to have a more efficient system for analyzing medical images.

[0006] US Patent Application Publication No. 2010 / 0305441A1 relates to an automated ultrasound image optimization system.

[0007] EP 3644273 A1 relates to a system for determining image quality parameters of medical images. [Means for solving the problem]

[0008] The present disclosure relates to a system for analyzing medical image data representing two-dimensional or three-dimensional medical images acquired from a human subject. The system includes a data processing system that reads and / or determines a plurality of image quality metrics for the medical image. The data processing system further determines a combined quality metric based on the image quality metrics. The data processing system is configured such that the determination of the combined quality metric takes into account interactions between the image quality metrics with respect to their combined effect on the combined quality metric.

[0009] The image data is acquired using a medical imaging unit, which includes an X-ray imaging unit, a magnetic resonance imaging unit, and / or an ultrasound imaging unit, where the X-ray imaging unit is a unit for planar radiography and / or X-ray computed tomography.

[0010] Medical images acquired of a human subject include images of the lungs, chest, ribs, bones, ankles, or knees.

[0011] A three-dimensional medical image represents a tomographic image, which may be generated from a two-dimensional image using a reconstruction algorithm. Two-dimensional medical images are generated using projection imaging, specifically projection radiography.

[0012] The data processing system retrieves at least some of the quality metrics from one or more files stored on a storage device of the data processing system. Additionally or alternatively, the data processing system may determine at least some of the quality metrics using an analysis algorithm that uses at least some of the medical images as input. The analysis algorithm may include, but is not limited to, one or a combination of a segmentation algorithm, a registration algorithm for aligning the image to an atlas, a classifier, and a machine learning algorithm. The machine learning algorithm may be configured for supervised and / or unsupervised learning. In particular, the machine learning algorithm is an artificial neural network (ANN).

[0013] Additionally or alternatively, the data processing system may receive at least some of the quality metrics via a user interface of the data storage system and / or from a further data processing system, wherein the data processing system determines the image quality metrics automatically or interactively (i.e., requiring user intervention).

[0014] One or more or each of the quality metrics relates to parameters of an imaging process for acquiring medical images using a medical imaging unit. One or more or each of the quality metrics may indicate or include parameters of a position and / or orientation of a body part of a human subject relative to components of the medical imaging unit. By way of example, the components include a detector (such as an X-ray detector) of the imaging unit and / or a radiation source of the imaging unit. An example of an orientation parameter is the rotation angle of an ankle joint about a longitudinal axis. An example of a position parameter is the position of a lung field relative to the field of view of an X-ray radiosensitive detector. The field of view is adjusted by adjusting the position of a collimator of the X-ray imaging unit and / or by adjusting the relative positions of the subject and the radiosensitive detector. Additionally or alternatively, the extent of the lung field relative to the field of view can also be adjusted by adjusting an opening formed by the collimator.

[0015] Additionally or alternatively, one or more or each of the quality metrics may indicate or include a parameter of the orientation of parts of a human subject relative to one another. For one or more or all of the body parts, a corresponding body part is imaged in the medical image. One example of the orientation of two body parts relative to one another is the flexion angle of the ankle (i.e., the angle of dorsiflexion or plantar flexion).

[0016] The combination quality metric indicates whether or to what extent a medical image is sufficient for a medical diagnosis. The data processing system may determine the combination quality metric using an algorithm stored in the data processing system. The determination of the combination quality metric may be performed automatically or interactively (i.e., requiring human intervention). For example, the data processing system may receive user input used to determine the combination quality metric via a user interface of the data processing system. One example of such user input is a desired diagnosis based on the medical image. Based on this user input, the data processing system selects one or more image quality metrics and / or an algorithm for determining the combination quality metric. This allows for the determination of a combination quality metric that reliably enables the determination of whether the acquired images are suitable for making the desired diagnosis.

[0017] At least some of the combined quality metrics may have a plurality of predefined quality level values, where one or more of the quality levels may indicate that the medical image meets the predefined criteria, and one or more second levels may indicate that the medical image does not meet the predefined criteria, where the predefined criteria indicate whether or to what extent the medical image is suitable for a predefined diagnosis.

[0018] The combined quality metric may be a tabular or analytical function that depends on the image quality metrics. The combined quality metric may include or be a scalar value, vector, and / or state variable. Combinations of values ​​of the image quality metrics that represent the same value of the combined quality metric may represent contour lines in a height map that indicate at least a portion of the combined quality metric.

[0019] According to a further embodiment, the interaction between the image quality metrics is such that a change in the combined quality metric caused by a change in a first quality metric among the quality metrics can be compensated for by a change in one or more quality metrics among the remaining quality metrics.

[0020] According to further embodiments, each of the image quality metrics is associated with a corresponding predefined optimum value or a predefined optimum range. By way of example, one of the optimum values ​​is an optimum orientation of the body part, in particular the imaged body part, relative to the components of the imaging unit (such as the detector and / or the source of the medical imaging unit). Additionally or alternatively, the optimum value may be an optimum orientation of the body parts relative to each other. By way of example, the optimum angle is an optimum angle of the ankle joint (i.e., dorsiflexion angle or plantar flexion angle).

[0021] The interaction between the image quality metrics may be such that an increase in deviation of a first one of the metrics from an optimal range or value can be compensated for by a decrease in deviation of one or more of the remaining image quality metrics from their optimal ranges or values.

[0022] According to further embodiments, the or each quality metric is indicative of one or more parameters of the position and / or one or more parameters of the orientation of the imaged body part relative to a component of the imaging unit or relative to a further body part. The component may be a detector and / or a source of the medical imaging unit. The or each body part is at least partially imaged in the medical image.

[0023] Further examples of quality metrics include, but are not limited to, inhalation status, collimation parameters, field of view position parameters, and distances between body parts and components of the medical imaging unit.

[0024] For example, the inhalation status is represented using the number of ribs that are shown above the diaphragm in the image. The analysis algorithm detects the ribs that are shown above the diaphragm in the image and determines the number of these ribs. The detection of the ribs can be performed using a segmentation algorithm.

[0025] Determining the collimation parameters may include determining the distance between the boundary of the image field of view and the optimal collimation boundary that covers the target anatomical structure (such as the lung field) without under- or over-collimation. The collimation parameters may make it possible to detect under-collimation and over-collimation in medical images. An over-collimated image shows only a portion of the target anatomical structure (such as the lung field), thereby making only some of the information necessary to make a diagnosis available. On the other hand, an under-collimated image shows body parts that do not need to be imaged to make a diagnosis, thereby exposing these body parts to radiation that could be avoided with proper collimation.

[0026] The field of view position parameter may be a parameter of the position of the field of view of the image relative to an optimal field of view that covers the target anatomy without under- or over-collimation.

[0027] Other quality metrics include, but are not limited to, (a) a metric indicating whether a given anatomical structure or a minimum amount of anatomical structures is visible in the image (e.g., whether one or both scapulae are visible in the imaged lung field), and (b) a metric indicating the percentage of overlap between matching anatomical structures (e.g., femoral condyles). Matching anatomical structures are defined as structures that are symmetrical or nearly identical. In particular, in two-dimensional projection images, it can be advantageous to have matching anatomical structures overlap or substantially overlap with respect to their contours, which increases the reliability of the diagnosis.

[0028] According to a further embodiment, the system determines, for one or more of the images, parameters or state variables representing changes in imaging conditions for using the imaging unit based on the determined combined quality metric. By way of example, the state variables represent operational states of the medical imaging unit. Further, examples of parameters representing changes in imaging conditions include, but are not limited to, (a) operational parameters of the medical imaging unit, (b) parameters of position and / or orientation of a body part relative to components of the medical imaging unit, and (c) parameters of position and / or orientation of two or more body parts relative to each other.

[0029] By way of example, the change in imaging condition is a change in the position and / or orientation of the body parts relative to the components of the imaging unit and / or a change in the position and / or orientation of the body parts relative to each other. Additionally or alternatively, the change in imaging condition is a change in the operating parameters of the imaging unit, such as the position of the collimator, the tilt of the X-ray tube and / or X-ray sensitive detector, the position of an anti-scatter grid or whether an anti-scatter grid is used, the geometry of the patient support (such as the geometry of the bending unit), etc.

[0030] When the imaging conditions are changed, one or more of the image quality metrics are changed, i.e., the system provides suggestions to the user or medical imaging unit to adapt the imaging conditions to obtain a change in the combined quality metric.

[0031] According to a further embodiment, determining the image quality metric includes segmenting at least a portion of the medical image. Additionally or alternatively, the determining includes registering the medical image to an atlas. The registration of the image data with the atlas may be performed using the segmented image. The image segmentation determines one or more image regions. The segmentation may be performed using a data processing system and / or a further data analysis system from which the data processing system receives data.

[0032] Segmentation of medical images is performed using any or a combination of the following segmentation techniques: thresholding, region growing, watershed transformation, edge detection, shape models, appearance models, and manual segmentation through user interaction with a graphical user interface. Additionally or alternatively, segmentation may be performed using an artificial neural network. Segmentation may be performed automatically or interactively (i.e., requiring user intervention). In interactive segmentation, a computer system receives user input indicating one or more parameters of the location, orientation, and / or outer contour of an image region. The artificial neural network is trained using segmented images, particularly those performed by interactive segmentation.

[0033] An atlas includes a statistically averaged anatomical map of one or more body parts. At least a portion of the atlas shows or represents the two-dimensional or three-dimensional shape of a body part, particularly an anatomical part of the body. By way of example, at least a portion of the atlas shows or represents the three-dimensional exterior surface of one or more anatomical or functional parts of the body. By way of example, the anatomical part of the body is one or more bones and / or one or more parts of a bone.

[0034] The atlas is generated based on anatomical data acquired from multiple human bodies. The anatomical data is acquired using medical imaging units such as X-ray imaging units, magnetic resonance imaging units, and / or ultrasound imaging units. The human bodies used to generate the atlas data may share a common characteristic or range of characteristics, such as gender, age, ethnicity, body size, weight, and / or pathological condition.

[0035] The present disclosure relates to a system for analyzing medical image data representing multiple medical images, each of which may be a two-dimensional or three-dimensional image. The medical images are acquired from one or more human subjects using one or more medical imaging units, each of which acquires a medical image. The system includes a data processing system that reads and / or generates one or more quality metrics for each of the medical images. The data processing system further receives, via a user interface, user input for at least some of the images indicating a user-specified quality assessment of the corresponding image. The data processing system further (a) determines or adapts an algorithm for determining a combined quality metric based on the image quality metrics and based on the user input, and / or (b) determines or adapts an algorithm for classifying the images or selecting portions of the images, the classification or selection being based on the user input and based on the combined quality metric. The combined quality metric relies on interactions between the image quality metrics for their combined effect on the combined quality metric.

[0036] The data processing system retrieves at least some of the quality metrics from one or more files stored on a storage device of the data processing system. Additionally or alternatively, the data processing system may determine at least some of the quality metrics using an analysis algorithm that uses at least some of the medical images as input. The analysis algorithm may include, but is not limited to, one or a combination of a segmentation algorithm, a registration algorithm for aligning the image to an atlas, a classifier, and a machine learning algorithm. The machine learning algorithm may be configured for supervised and / or unsupervised learning. In particular, the machine learning algorithm is an artificial neural network (ANN).

[0037] Additionally or alternatively, the data processing system may receive at least some of the quality metrics via a user interface of the data storage system and / or from a further data processing system, wherein the data processing system determines the image quality metrics automatically or interactively (i.e., requiring user intervention).

[0038] The user input indicating the user-specified quality rating includes one of a plurality of predefined levels, and the user input is received via a user interface, including, but not limited to, any one or combination of a keyboard, a computer mouse, touch control, voice control, and / or gesture control.

[0039] The image classification algorithm is an algorithm for classifying each image into one or more of a plurality of predefined classes. The classification includes binary classification data and / or probabilistic classification data. Probabilistic classification data is defined as data including one or more probability values ​​for one or more of the predefined classes. Binary classification data is defined as data including, for one or more of the predefined classes, a value indicating that the image is a member of the class or a value indicating that the image is not a member of the class.

[0040] The predefined classes include a class of diagnostic images and a class of non-diagnostic images. The predefined classes are represented by a class of diagnostic images and a class of non-diagnostic images.

[0041] The algorithm for selecting may, for example, be an algorithm for selecting portions of diagnostic images. In another embodiment, the algorithm for selecting is an algorithm for selecting non-diagnostic images.

[0042] According to a further embodiment, determining an algorithm for determining the combination quality metric and / or determining an algorithm for classifying the image comprises training a machine learning algorithm, in particular an artificial neural network (ANN). The artificial neural network comprises an input layer, one or more hidden layers, and an output layer. The artificial neural network is configured as a convolutional neural network, in particular a deep convolutional neural network.

[0043] According to a further embodiment, the system displays, via a user interface, an output indicative of a combined quality metric for at least some of the images. Additionally or alternatively, the system may classify the medical image based on the combined quality metric and display an output indicative of a classification of the medical image. The output may represent probabilistic and / or binary classification data. The classification is determined and / or adapted by the data processing system based on user input indicative of a user-specified quality assessment.

[0044] According to one embodiment, the data processing system receives, via a user interface, a user-specified selection of one or more of the medical images for inputting a user input indicating a user-specified quality assessment for the images, the images being selected by the user-specified selection. The user-specified selection is received via a graphical user interface, in particular a computer mouse of the graphical user interface. In response to the selection, the data processing system requests the user to input a user input indicating a user-specified quality assessment for the selected one or more images.

[0045] According to one embodiment, the system outputs, for at least some of the images, an output indicative of one or more image quality metrics of the corresponding image.

[0046] According to a further embodiment, the system outputs a graphical representation indicating a coordinate system for one or more of the image quality metrics. The graphical representation is displayed using a graphical user interface. The graphical representation is displayed on a display device of the data processing system. For one or more of the images, the graphical representation further indicates a spatial relationship between the corresponding image and the coordinate system. The graphical representation includes, for each of the images, a graphical representation representing the image, such as one or more icons, where the position and / or orientation of the graphical representation representing the image relative to the coordinate system indicates the image quality metric for the corresponding image.

[0047] For each of the images, for the image quality metric used to form the coordinate system, the spatial relationship indicates the value of the image quality metric for the corresponding image. The spatial relationship can be a one-dimensional, two-dimensional, or three-dimensional spatial relationship.

[0048] According to one embodiment, the data processing system receives, via a user interface, a user-specified selection of one or more of the images whose graphical representations indicate spatial relationships. The user interface is configured as a graphical user interface in which one or more regions are user-selectable. The selectable regions are displayed on a display device of the data processing system, thereby allowing the user to visually recognize the location and extent of the regions. The spatial arrangement of the regions depends on the graphical representation, and in particular the spatial relationships that the graphical representations indicate. The graphical representation may indicate the user-selectable regions.

[0049] The user-selectable regions are selectable by a user using a computer mouse in the user interface. Each of the user-selectable regions may represent a medical image. Each of the user-selectable regions is positioned in a coordinate system of one or more of the image quality metrics. The graphical user interface displays a graphical representation, such as an icon, for each of the user-selectable regions, representing a medical image selectable by user input.

[0050] By way of example, the graphical representation may include a graph, which may show the combined quality metric of the image as discrete values ​​of a height function.

[0051] The coordinate system may be a one-dimensional, two-dimensional, or three-dimensional coordinate system. Additionally or alternatively, the coordinate system may be a Cartesian, spherical, or cylindrical coordinate system. The data processing system receives user input indicating one or more selected image quality metrics. The data processing system displays the coordinate system including the selected image quality metrics.

[0052] According to further embodiments, determining or adapting an algorithm for classifying images and / or determining a combined quality metric is performed using one or a combination of a maximum likelihood model and a machine learning algorithm. The machine learning algorithm may be configured for supervised or unsupervised learning. The machine learning algorithm may be implemented using a support vector machine or an artificial neural network.

[0053] According to one embodiment, the system includes one or more medical imaging units, each of which acquires a medical image.

[0054] The present disclosure further relates to a computer-implemented method for analyzing medical image data representing a two-dimensional or three-dimensional medical image, the image being an image of at least a portion of a human subject. The method is performed using a data processing system. The method includes using the data processing system to read and / or determine a plurality of image quality metrics for the medical image. The method further includes using the data processing system to automatically or interactively determine a combined quality metric based on the image quality metrics. In determining the combined quality metric, interactions between the image quality metrics are taken into account for the combined effect of the image quality metrics on the combined quality metric.

[0055] The present disclosure relates to a method for acquiring and analyzing medical images, the method including acquiring medical images using a medical imaging unit, the method further including the computer-implemented method described in the previous paragraph for analyzing the acquired medical images.

[0056] The present disclosure further relates to a computer-implemented method for analyzing medical image data representing a plurality of medical images, each of which may be a two-dimensional or three-dimensional image, of portions of one or more human subjects. The method is performed using a data processing system. The method includes using the data processing system to read and / or generate one or more quality metrics for each of the medical images. The method further includes receiving, via a user interface of the data processing system, user input for at least some of the images indicating a user-specified quality assessment of the corresponding image. The method further includes at least one of the following steps: (a) determining or adapting, using the data processing system, an algorithm for determining a combined quality metric based on the image quality metrics and based on the user input; and / or (b) determining or adapting, using the data processing system, an algorithm for classifying the images or selecting portions of the images, wherein the classifying or selecting is based on the user input and the combined quality metric. The combined quality metric relies on interactions between the image quality metrics for their combined effect on the combined quality metric.

[0057] The present disclosure further relates to a method for acquiring and analyzing medical images, the method including acquiring a plurality of medical images using one or more medical imaging units, each of the medical imaging units acquiring a medical image, the method further including the computer-implemented method described in the previous paragraph for analyzing the plurality of medical images.

[0058] The present disclosure relates to a program element for analyzing medical image data representing two-dimensional or three-dimensional medical images, where the images are images of at least a portion of a human subject, the method being carried out using a data processing system. The program element, when executed by a processor of the data processing system, causes the data processing system to read and / or determine a plurality of image quality metrics for the medical image. The program element, when executed by the processor, further causes the data processing system to automatically or interactively determine a combined quality metric based on the image quality metrics. In determining the combined quality metric, the interaction between the image quality metrics is taken into account for the combined effect of the image quality metrics on the combined quality metric.

[0059] The present disclosure relates to a program element for analyzing medical image data representing multiple medical images, each of which may be a two-dimensional or three-dimensional image. The medical images are images of portions of one or more human subjects. When executed by a processor of a data processing system, the program element causes the program element to read and / or generate, using the data processing system, one or more image quality metrics for each of the medical images. The program element is further configured to receive, via a user interface of the data processing system, user input for at least some of the images indicating a user-specified quality assessment of the corresponding image. When executed by the processor, the program element causes the program element to perform at least one of: (a) determining or adapting, using the data processing system, an algorithm for determining a combined quality metric based on the image quality metrics and based on the user input; and / or (b) determining or adapting, using the data processing system, an algorithm for classifying the image or selecting a portion of the image, wherein the classification or selection is based on the user input and the combined quality metric. The combined quality metric relies on interactions between the image quality metrics for their combined effect on the combined quality metric.

[0060] The present disclosure relates to a computer program product having stored thereon the computer program elements of any one of the above embodiments. [Brief explanation of the drawings]

[0061] [Figure 1] FIG. 1 is a schematic diagram of a system for analysis of medical image data in accordance with an exemplary embodiment. [Figure 2] FIG. 2 illustrates schematically two image quality metrics of an ankle radiograph determined by a system according to an exemplary embodiment. [Figure 3]FIG. 3 illustrates a schematic representation of an image display window of a graphical user interface of a system according to an exemplary embodiment, displaying a normalized X-ray radiograph. [Figure 4] FIG. 4 shows a schematic diagram of the dependency of the combined quality metric on two image quality metrics, which is calculated by the computer system of the system according to the exemplary embodiment. [Figure 5] FIG. 5 is a schematic illustration of a graphical representation of a user interface of a central computer system of a system according to an exemplary embodiment, where for each image, the graphical representation shows the value of the combined quality metric and the image quality metric of the corresponding image. [Figure 6] FIG. 6 illustrates schematically the operation of the graphical user interface of the central computer system. [Figure 7] FIG. 7 is a schematic diagram of an artificial neural network (ANN) used by the computer system of the system according to an exemplary embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0062] 1 is a schematic diagram of a system 1 for analyzing medical image data according to an exemplary embodiment. System 1 includes multiple medical imaging units 2 a, 2 b, and 2 c, each of which acquires X-ray images using planar projection radiography. Each of medical imaging units 2 a, 2 b, and 2 c includes a radiation source 3 a, 3 b, and 3 c and a radiation-sensitive detector 4 a, 4 b, and 4 c that detects X-ray imaging radiation emitted by radiation sources 3 a, 3 b, and 3 c.

[0063] Each of the imaging units 2a, 2b, 2c acquires projection X-ray images from a subject 8a, 8b, 8c positioned between a radiation source 3a, 3b, 3c and a radiation-sensitive detector 4a, 4b, 4c. System 1 includes a computer system 5a, 5b, 5c for each of the medical imaging units 2a, 2b, 2c. The computer systems are operated by a radiologist who positions the subject 8a, 8b, 8c between the X-ray source 3a, 3b, 3c and the radiation-sensitive detector 4a, 4b, 4c so that a body part of the subject 8a, 8b, 8c is imaged using X-rays emitted from the radiation source 3a, 3b, 3c.

[0064] 1, the imaged body parts of the subjects 8a, 8b, 8c are the chests of the subjects 8a, 8b, 8c, and therefore the image data generated using the medical imaging units 2a, 2b, 2c represent chest radiographs, although it is conceivable that the acquired image data may represent other body parts, such as the ankles, knees, shoulders, elbows, wrists, hips, or spine.

[0065] The system 1 further includes a central computer system 7, operated for example by a radiologist. The central computer system 7 may be a dedicated server that receives all images acquired using the medical imaging units 2a, 2b, and 2c for analysis by the radiologist. The computer systems are in signal communication with each other via a computer network, including a LAN (Local Area Network) 6 and / or the Internet.

[0066] Each medical imaging unit 2 a, 2 b, and 2 c determines, for each acquired image, a number of image quality metrics based on the image data of the respective image. The determination of the imaging quality metrics is performed automatically or semi-automatically (i.e., requiring user intervention). One or more of all the image quality metrics may be indicative of or a parameter of the position and / or orientation of the imaged body part of the imaged subject 8 a, 8 b, 8 c relative to components of the medical imaging unit 2 a, 2 b, 2 c, including at least a part of the X-ray radiation source 3 a, 3 b, 3 c and / or the radiation-sensitive detector 4 a, 4 b, 4 c.

[0067] FIG. 2 is a schematic diagram of an ankle X-ray radiograph acquired by one of imaging units 2a, 2b, and 2c (shown in FIG. 1). In some ankle X-ray examinations, it is advantageous for the angle of rotation about the longitudinal axis of the tibia (denoted as α in FIG. 2) and the angle of ankle flexion (i.e., the angle of dorsiflexion and plantarflexion, denoted as β in FIG. 2) to be within predetermined ranges, thereby allowing the radiograph to show diagnostic features, such as the joint space, that allow a radiologist to make a medical diagnosis based on the radiograph. Furthermore, radiographs in which these parameters do not fall within the predetermined ranges may exhibit artifacts, making it difficult or even impossible to make a reliable diagnosis. These parameters therefore represent image quality metrics and can be used to determine whether a radiograph has diagnostic image quality.

[0068] It should be noted that these image quality metrics are merely examples, and the present invention is not limited to these image quality metrics. In particular, the image quality metrics, as well as the predetermined ranges and / or optimal values ​​of the image quality metrics, depend on the body-part imaged and / or the particular diagnosis to be made based on the radiograph.

[0069] Determination of one or more or all of the image quality metrics is performed using image processing applied to the image. Image processing may include image segmentation. Segmentation of medical images is performed using any or a combination of the following segmentation techniques: thresholding, region growing, watershed transformation, edge detection, use of shape models, use of appearance models, and manual segmentation through user interaction with a graphical user interface. Additionally or alternatively, segmentation may be performed using an artificial neural network. The artificial neural network is trained using manual segmentation. Segmentation is performed automatically or interactively (i.e., requiring user intervention). In interactive segmentation, a computer system receives user input indicating one or more parameters of the location, orientation, and / or outer contour of an image region.

[0070] The segmentation results in one or more segmented image regions and / or one or more contours. The image regions and / or contours are two-dimensional. In a three-dimensional image, the image regions or contours are three-dimensional. At least a portion of the image regions or contours represent anatomical or functional parts of the body or its surface. An anatomical part of the body is the bony and / or tissue structure of the body. A functional part of the body is a part of the body that performs an anatomical function.

[0071] Additionally or alternatively, determining the image quality metric includes registering at least a portion of the image with the atlas. The data processing system extracts features from the image, which are used to register the image with the atlas. The features are extracted using image processing. The atlas registration is performed based on a contact gating of the image. Additionally or alternatively, the atlas registration is performed based on landmarks detected in the image using the computer system.

[0072] Image segmentation extracts one or more parameters of position, one or more parameters of orientation, and / or one or more parameters of the extent and / or shape of a body part (such as an anatomical or functional part of the body). Atlas registration uses one or more of the extracted parameters to register the image to the atlas.

[0073] The inventors have shown that atlas registration allows for more reliable determination of image quality metrics, such that a combined quality metric, determined based on the determined image quality metrics and described in detail below, can more accurately indicate whether an image is of diagnostic image quality. However, it has also been shown that sufficient accuracy can be achieved without using segmentation and atlas registration techniques.

[0074] Each of computer systems 5a, 5b, and 5c (shown in FIG. 1) displays images to a user via a graphical user interface, and simultaneously displays a graphical representation that at least partially illustrates one or more of the features used to register the images to the atlas.

[0075] 3 is a schematic diagram of an image display window 27 of a graphical user interface showing an X-ray image 25 of a knee joint and a graphical representation 26 showing the outer contours of the bones (i.e., the tibia, fibula, patella, and femur). In the exemplary embodiment shown, the computer system uses at least a portion of these outer contours to register the image to the atlas. A graphical representation (e.g., as shown in FIG. 3) showing the parameters used to register the image to the atlas allows the user to verify whether the computer system correctly determined the image quality parameters.

[0076] Each of the computer systems 5a, 5b, and 5c determines at least one combined quality metric for each medical image based on the image quality metrics of the corresponding image. By way of example, the combined quality metric may include parameters, vectors, and / or state variables calculated based on the calculated values ​​of the image quality metrics. The combined quality metric may be a function of the image quality metrics. The function may include analytical and / or tabular functions stored on the computer systems 5a, 5b, and 5c.

[0077] The inventors have discovered that image quality metrics interact in their effect on image quality, which is a measure of the suitability of a medical image for medical diagnosis.

[0078] By way of example, in radiographs of the ankle (such as the radiograph shown in FIG. 2), the inventors have discovered that larger deviations in the rotation angle α about the longitudinal axis LA of the tibia are compensated for by smaller deviations in the flexion angle β of the ankle joint. That is, images in which the deviation of the flexion angle β from the optimal flexion angle is relatively small will have a relatively large deviation in the rotation angle α about the longitudinal axis of the tibia from the optimal rotation angle.

[0079] Additionally or alternatively, in images in which the rotation angle α of the tibia about the longitudinal axis deviates relatively little from the optimal rotation angle, the flexion angle β of the ankle joint will have a relatively large deviation from the optimal flexion angle.

[0080] The inventors further show that a combined quality metric that takes into account this interaction between image quality metrics can more accurately determine whether an image is of diagnostic quality.

[0081] FIG. 4 illustrates a schematic representation of the properties of such a combined quality metric that considers the interaction between image quality metrics. The diagram in FIG. 4 shows equal-height contour lines 10 representing the combination of a first image quality metric (represented by the X-axis 8) and a second image quality metric (represented by the Y-axis 8). The combined quality metrics have a constant value. In the exemplary embodiment of FIG. 4, the first image quality metric (represented by the X-axis 8) is the rotation angle of the tibia about its longitudinal axis (denoted as α in FIG. 2), and the second image quality metric (denoted by the Y-axis 9) is the flexion angle of the ankle joint (denoted as β in FIG. 2). However, it should be noted that the present invention is not limited to this combination of image quality metrics. Each of the image quality metrics has an optimum value, which is shown in FIG. 4 by lines 13 and 14.

[0082] In order for computer systems 5a, 5b, and 5c (shown in FIG. 1 ) to determine the flexion angle β and the rotation angle α about the longitudinal axis of the tibia for each acquired image, each image represents a point (such as points 15, 16, 17, and 18) in the diagram of FIG. 4 . The value of the combined quality metric represented by contour line 10 is selected to represent a boundary separating a first region 19 from a second region 20. First region 19 represents images for which the first and second image quality metrics make the image suitable for diagnosis. Second region 20 represents images for which the first and second image quality metrics make the image unsuitable for diagnosis.

[0083] The boundaries represented by the contour lines 10 therefore represent a criterion for selecting images based on a combined quality metric.

[0084] The images represented by points 15, 16, and 17 are of diagnostic image quality because they are within the region bounded by or at contour line 10. On the other hand, image 18 is of non-diagnostic image quality because it is outside contour line 10.

[0085] Because the contour line 10 of the combined quality parameter deviates from the shape of a rectangle 11, the range 12 within which the rotation angle about the longitudinal axis of the tibia (parameter α in FIG. 2) is acceptable varies with the value of the flexion angle of the ankle joint (parameter β in FIG. 2). Specifically, as can be seen by comparing images 15 and 16 shown in FIG. 4, a small deviation of the flexion angle from its optimal value 14 (as in image 16) leads to a large deviation of the rotation angle about the longitudinal axis of the tibia from its optimal value 13. As can be further seen from FIG. 4, this situation differs from the situation in which fixed boundaries are defined for each image quality metric (represented by rectangle 11).

[0086] Each of computer systems 5a, 5b, and 5c (shown in FIG. 1) outputs a combination quality metric, i.e., an output determined based on the combination quality metric, to a radiologist via a user interface. This allows the radiologist to confirm whether the images are suitable for diagnosis and to acquire repeat images if necessary. The determination of the combination quality metric is performed automatically based on the calculated image quality and metric, so the radiologist does not need to be present when the radiographs are acquired.

[0087] The output may also indicate modifications of imaging conditions so that the resulting image is suitable (or more suitable) for diagnosis.

[0088] As an example, image 18 shown in Figure 4 has a relatively large deviation from optimal rotation angle 14 and a relatively small deviation from optimal flexion angle 13. However, because the combined quality metric takes into account the interaction between the image quality metrics, only a small correction of both angles by approximately equal amounts (indicated by arrow 25) is required to obtain an image 24 suitable for diagnosis.

[0089] The determined image quality metrics and combined quality metrics are transmitted from computer systems 5a, 5b, and 5c (shown in FIG. 1) to a central computer system 7 for further analysis performed by a radiologist.

[0090] The computer system 7 includes a user interface that allows a radiologist to review the determined combination quality metric for multiple images acquired using the imaging units 2 a, 2 b, and 2 c. Based on the review, the central computer system 7 and / or the computer systems 5 a, 5 b, and 5 c can adjust an algorithm for determining the combination quality metric or adjust an algorithm for determining whether an image is a diagnostic image based on the combination quality metric.

[0091] Thus, the central computer system 7 determines data indicative of criteria for selecting images (e.g., diagnostic images) based on a combined quality metric based on a user-specified quality rating for one or more of the images (provided by a radiologist). Further, the central computer system 7 determines data used to determine the combined quality metric based on the image quality metrics based on the user-specified quality rating. This allows for tailoring the determination of the combined quality metric based on user input provided by the radiologist.

[0092] FIG. 5 is a schematic diagram of a graphical representation presented to a radiologist using the central computer system's graphical user interface. Similar to that shown in FIG. 4, each image is shown as an icon within a coordinate system with two axes, each representing one of the image quality metrics. Thus, for each image, the graphical representation shows the spatial relationship between the corresponding image and the coordinate system. As further shown in FIG. 5, each icon represents the combined quality metric for the corresponding image. Each icon has a shape that indicates the level of the combined quality metric, with a circle representing a value of the combined quality metric that indicates high image quality, a square representing a value of the combined quality metric that indicates an intermediate level of image quality, and a hexagon representing a value of the combined quality metric that indicates low image quality. This results in an output that indicates the combined quality metric.

[0093] For each image, the combined quality metric of the corresponding image can conceivably be indicated by means other than the icons described above, such as being displayed in a different color or including a number.

[0094] It is also contemplated that the coordinate system may be a one-dimensional coordinate system of one image quality metric, or that the coordinate system may be a three-dimensional spatial coordinate system. It is also contemplated that one or more dimensions may be indicated using color coding, icons, and / or numbers, thereby providing a graphical representation of four or more image quality metrics. The graphical user interface receives user input for selecting one or more image quality metrics to be used to generate the graphical representation. Additionally or alternatively, the central computer system may automatically or interactively (i.e., require user intervention) determine one or more image quality metrics based on predefined criteria. The predefined criteria may vary depending on the intended diagnosis.

[0095] The graphical user interface of the central computer system further allows the radiologist to select one or more of the images. For example, the graphical user interface may be configured such that each icon represents a user-selectable area of ​​the graphical user interface. The user-selectable areas are selectable using a computer mouse of the data processing system. In response to the selection, the central computer system displays the selected images to allow the radiologist to verify whether the determined combined quality metric adequately indicates whether the images are suitable for diagnosis.

[0096] Thus, the data processing system receives a user-specified selection of one or more of the medical images via the graphical user interface. In response to the user-specified selection, the central computer system displays the selected one or more images to the user and requests the user to enter user input. The user input indicates, for each of the selected one or more medical images, a user-specified quality rating of the corresponding image.

[0097] The graphical user interface displays a graphical representation showing the parameters used to register the image to the atlas, similar to that described in connection with Figure 3. By displaying the image in a graphical representation, the radiologist can more easily determine whether the image is of diagnostic image quality.

[0098] Additionally or alternatively, the central computer system may register the medical images to the atlas using one or more rigid transformations. The central computer system may further display at least a portion of the transformed images to a user using a graphical user interface.

[0099] Rigid transformations include rotational transformations, scaling transformations, and / or image translations. Because the atlas has a fixed position, orientation, and scale, applying rigid transformations to multiple different images produces a consistent representation of the anatomy across those images. Therefore, using these rigid transformations makes it easier for radiologists to compare images and recognize abnormalities within them. This makes image quality reviews performed by radiologists more time-efficient and reliable.

[0100] 6 illustrates in more detail the operation of the graphical user interface of central computer system 7 (shown in FIG. 1). In the illustrated exemplary embodiment, the graphical user interface displays a graphical representation 21 indicating a boundary of the combined quality metric. The boundary indicates a range of values ​​of the combined quality metric for which an image is of diagnostic image quality. The boundary is thus a graphical representation for classifying images into two or more classes (i.e., diagnostic and non-diagnostic).

[0101] Thus, similar to that described in connection with FIG. 4, the graphical representation 21 of FIG. 6 shows a boundary 21 of the combined quality parameters that separates images having diagnostic image quality from images having non-diagnostic image quality.

[0102] As further explained in relation to Figure 4 above, the boundary graphical representation 21 (shown in Figure 6) does not have a rectangular shape because the combined quality metric takes into account interactions between image quality metrics for their effect on the combined quality metric.

[0103] The graphical representation 21 allows the radiologist to select images close to the boundary to see if the boundary represents an adequate separation between diagnostic and non-diagnostic quality images. The graphical user interface allows the radiologist to select images using a mouse cursor, and the central computer system 7 responds by displaying the selected images to the user.

[0104] As further shown schematically in Fig. 6, the graphical user interface allows for the selection of images that are close to the determined quality threshold 21, thereby facilitating the determination based on a large number of images whether the calculated combined quality metric and the determined threshold 21 adequately indicate whether the images are suitable for diagnosis. Specifically, it facilitates the radiologist to select images that should be provided with a user-specified quality rating in order to efficiently adapt the algorithm for determining the combined quality metric and / or the algorithm for determining the bounds 21 of the combined quality parameter.

[0105] The central computer system 7 further selects images based on the boundaries and further based on the determined combined quality metric, particularly so that the radiologist does not have to decide which images require review.

[0106] The central computer system 7 further receives input from a radiologist via the graphical user interface for one or more of the images indicating a user-specified quality rating for the corresponding image. In particular, the user-specified quality rating includes an indication of whether the image is of diagnostic image quality.

[0107] Based on input received via the user interface, the central computer system adjusts the boundaries of the combined quality metric (which is an example of an algorithm for classifying images based on the combined quality metric) and / or the central computer system adjusts the algorithm for determining the combined quality metric. The central computer system may also determine new algorithms for classifying images and / or new algorithms for determining the combined quality metric based on user input.

[0108] Adapting or determining criteria for selecting images based on combined user input and / or adapting or determining algorithms for determining combined quality metrics may include a machine learning process in which the user's input (i.e., indications of whether the images are suitable for diagnosis and / or user-specified quality ratings) is used as training data.

[0109] Additionally or alternatively, determining an algorithm for classifying images based on combined user inputs and / or adapting an algorithm for determining a combined quality metric can be performed using dimensionality reduction techniques, such as kernel principal component analysis. By way of example, such techniques can be used to simplify boundary determination by reducing the number of images a radiologist needs to review. Specifically, for a given boundary, these techniques can be used to find a low-dimensional phase space representation of the combined quality metric boundary. A low-dimensional phase space representation can facilitate graphical illustration of the boundary via a graphical user interface and can also make adapting the combined quality metric boundary more efficient. After adapting the boundary, a dimensionality reduction technique can be applied again to further simplify the phase space.

[0110] Determining or adapting an algorithm for classifying images and / or determining a combined quality metric can be performed using one or a combination of maximum likelihood models and machine learning algorithms. The machine learning algorithms can be configured for supervised and / or unsupervised learning. The machine learning algorithms can be implemented using support vector machines or artificial neural networks.

[0111] Specifically, the maximum likelihood model is implemented using the assumption that the image quality metric spans a Euclidean space, that the distribution of one of several predefined classes (e.g., the class of diagnostic images) in this space is relatively compact, and that a function f(p), such as a Gaussian-normal distribution, is known to approximate it.

[0112] According to an embodiment, the central computer system generates a maximum likelihood model for a plurality of points p in Euclidean space that indicates the probability that the corresponding image is a member of one of a plurality of predefined classes. By way of example, the predefined classes include a class of diagnostic images and a class of non-diagnostic images. The central computer system further uses a threshold Θ for a combination of image quality metrics to determine the class to which the corresponding image belongs (e.g., if f(p)>Θ, then it is diagnostic; if f(p)<=Θ, then it is non-diagnostic). The corresponding boundary represents the Mahalanobis distance.

[0113] Classification algorithms that use machine learning are implemented using support vector machines and / or artificial neural networks. A description of support vector machines that can be used in the embodiments described herein can be found in "The nature of statistical learning," written by Vladimir Vapnik in 1995 and published by Springer Science+Business Media, New York.

[0114] 7 is a schematic diagram of an artificial neural network (ANN) 119. The ANN is used by computer systems 5a, 5b, and 5c (shown in FIG. 1) to perform segmentation of medical images for determining a combination quality metric. The same configuration of the ANN is also used by central computer system 7 to determine an algorithm for determining the combination quality metric and / or an algorithm for classifying images.

[0115] As can be seen in FIG. 7, the ANN 19 includes a plurality of neural processing units 120a, 120b, ..., 124b. The neural processing units 120a, 120b, ..., 124b are connected to form a network via a plurality of connections 118, each having a connection weight. Each connection 118 connects a neural processing unit in a first layer of the ANN 119 to a neural processing unit in a second layer of the ANN 119, which immediately follows or precedes the first layer. As a result, the artificial neural network has a layered structure including an input layer 121, at least one intermediate layer 123 (also called a hidden layer), and an output layer 125. In FIG. 4a, only one of the intermediate layers 123 is shown schematically. However, it is contemplated that the ANN 119 may include three or more, four or more, six or more, or even eleven or more intermediate layers. Specifically, the ANN may be configured as a deep artificial neural network. The number of layers may be less than 200, less than 100, or less than 50.

[0116] A description of an ANN that can be used in the embodiments described in this disclosure is provided in the article "Foveal fully conventional nets for multi-organ segmentation" written by Tom Brosch and Axel Saalback and published in SPIE 10574, Medical Imaging 2018: Image Processing, 105740U, the contents of which are incorporated herein by reference for all purposes.

[0117] An ANN is configured as a convolutional neural network. In this specification, the term "convolutional neural network" is defined as an artificial neural network having at least one convolutional layer. A convolutional layer is defined as a layer that applies convolution to the previous layer. A convolutional layer includes multiple neurons, each of which receives input from a predefined section of the previous layer. The predefined section is also called a local receptive field. The weights of the predefined section are the same for each neuron in the convolutional layer. Therefore, a convolutional layer is defined by two concepts: weight sharing and field receptivity. An ANN may include one or more subsampling layers. Each subsampling layer is positioned after (especially immediately after) the corresponding convolutional layer. The subsampling layer downsamples the output of the preceding convolutional layer along the height and width dimensions. The number of convolutional layers followed by a pooling layer is at least one, at least two, or at least three. The number of layers may be less than 100, less than 50, or less than 20.

[0118] Before setting forth the claims, we first set forth the following clauses describing some salient features of certain embodiments of the present disclosure.

[0119] 1. A system for analysis of medical image data representing two-dimensional or three-dimensional medical images acquired from a human subject, comprising a data processing system that reads and / or determines a plurality of image quality metrics for the medical images and automatically or interactively determines a combined quality metric based on the image quality metrics, the data processing system being configured to take into account interactions between the image quality metrics regarding the combined effect of the image quality metrics on the combined quality metric in determining the combined quality metric.

[0120] 2. The system described in paragraph 1, wherein the interaction between image quality metrics is such that a change in the combined quality metric caused by a change in a first quality metric among the quality metrics is compensable by a change in one or more quality metrics among the remaining quality metrics.

[0121] 3. The system of claim 1 or 2, wherein each of the image quality metrics is associated with a corresponding predefined optimal value or predefined optimal range, and the interaction between the image quality metrics is such that an increase in deviation of a first one of the metrics from its optimal range or value can be compensated for by a decrease in deviation of one or more of the remaining image quality metrics from their optimal ranges or values.

[0122] 4. A system described in any one of paragraphs 1 to 3, wherein one or more or each of the quality metrics indicates one or more parameters of the position and / or one or more parameters of the orientation of the imaged body part relative to components of the imaging unit or relative to further body parts.

[0123] 5. A system described in any one of paragraphs 1 to 4, wherein the system determines, for one or more of the images, parameters or state variables representing changes in imaging conditions for using the imaging unit based on the determined combined quality metrics, and the changed imaging conditions change one or more of the image quality metrics.

[0124] 6. The system of any one of paragraphs 1 to 5, wherein determining the image quality metric includes segmenting at least a portion of the image and / or using the segmented image to register the image to the atlas.

[0125] 7. A system for analysis of medical image data representing a plurality of medical images, each of which is a two-dimensional or three-dimensional image, wherein the medical images are acquired from one or more subjects using one or more imaging units, each of which acquires the medical images, the system including a data processing system that reads and / or generates one or more quality metrics for each of the medical images and receives, via a user interface, user input for at least some of the images indicating a user-specified quality rating of the corresponding image, the data processing system further (a) determines or adapts an algorithm for determining a combined quality metric based on the image quality metrics and based on the user input, and / or (b) determines or adapts an algorithm for classifying the images or selecting portions of the images, wherein the classifying or selecting is based on the user input and on the combined quality metric, and the combined quality metric depends on the interaction between the image quality metrics in terms of their combined effect on the combined quality metric.

[0126] 8. The system of claim 7, displaying via a user interface an output indicating a combined quality metric for at least some of the images, and / or an output indicating a classification of the medical images, classifying the images based on the combined quality metric.

[0127] 9. The system of claim 7 or 8, wherein the system outputs, via a user interface, a graphical representation indicating a coordinate system for one or more of the image quality metrics, and for one or more of the images, the graphical representation further indicates a spatial relationship between the corresponding image and the coordinate system.

[0128] 10. The system of any one of clauses 7 to 9, wherein the data processing system receives, via a user interface, a user-specified selection of one or more of the images for inputting a user input indicating a user-specified rating for the images selected by the user-specified selection.

[0129] 11. A computer-implemented method for analysis of medical image data representing a two-dimensional or three-dimensional medical image, the image being an image of at least a portion of a human subject, the method being performed using a data processing system, the method comprising: using the data processing system to read and / or determine a plurality of image quality metrics for the medical image; and using the data processing system to automatically or interactively determine a combined quality metric based on the image quality metrics, wherein in the step of determining the combined quality metric, the interaction between the image quality metrics is taken into account for the combined effect of the image quality metrics on the combined quality metric.

[0130] 12. A computer-implemented method for analysis of medical image data representing a plurality of medical images, each of which is a two-dimensional or three-dimensional image, wherein the medical images are images of portions of one or more human subjects, the method being performed using a data processing system, the method comprising: using the data processing system to read and / or generate one or more quality metrics for each of the medical images; and receiving, via a user interface of the data processing system, user input for at least some of the images indicating a user-specified quality rating of the corresponding image, the method comprising at least one of the following steps: (a) determining or adapting, using the data processing system, an algorithm for determining a combined quality metric based on the image quality metrics and based on the user input; and / or (b) determining or adapting, using the data processing system, an algorithm for classifying the images or selecting portions of the images, wherein the classifying or selecting steps are based on the user input and on the combined quality metric, and the combined quality metric depends on the interaction between the image quality metrics for their combined effect on the combined quality metric.

[0131] 13. A program element for analysis of medical image data representing a two-dimensional or three-dimensional medical image, the image being an image of at least a portion of a human subject, the method being performed using a data processing system, the program element, when executed by a processor of the data processing system, performing the following operations: reading and / or determining, using the data processing system, a plurality of image quality metrics for the medical image; and automatically or interactively determining, using the data processing system, a combined quality metric based on the image quality metrics, wherein in determining the combined quality metric, the program element takes into account interactions between the image quality metrics for the combined effect of the image quality metrics on the combined quality metric.

[0132] 14. A program element for analysis of medical image data representing a plurality of medical images, each of which is a two-dimensional or three-dimensional image, the medical images being images of portions of one or more human subjects, the program element, when executed by a processor of a data processing system, performs the following actions: using the data processing system to read and / or generate one or more quality metrics for each of the medical images; and receiving, via a user interface of the data processing system, user input for at least some of the images indicating a user-specified quality rating of the corresponding image; the program element, when executed by the processor, further performs at least one of: (a) determining or adapting, using the data processing system, an algorithm for determining a combined quality metric based on the image quality metrics and based on the user input; and / or (b) determining or adapting, using the data processing system, an algorithm for classifying the image or selecting portions of the image, wherein the classification or selection is based on the user input and on the combined quality metric, and the combined quality metric depends on the interaction between the image quality metrics for their combined effect on the combined quality metric.

[0133] 15. A computer program product having stored thereon computer program elements according to clauses 13 and / or 14.

[0134] The above embodiments are illustrative and are not intended to limit the technical approach of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art will understand that the technical approach of the present invention can be modified or equivalently substituted without departing from the scope of protection of the claims of the present invention. In particular, although the present invention has been described based on projection radiography, it is applicable to any imaging technique that produces projection images. In the claims, the word "comprises" does not exclude other elements or steps, and the singular does not exclude a plurality. Any reference signs in the claims should not be construed as limiting the scope.

Claims

1. 1. A system for analysis of medical image data representing two-dimensional or three-dimensional medical images acquired from a human subject, comprising: the system includes a medical imaging unit for acquiring image data, the medical imaging unit including an X-ray imaging unit and / or a magnetic resonance imaging unit; The system further includes a data processing system, the data processing system comprising: reading and / or determining a plurality of image quality metrics for the medical image; automatically or interactively determining a combined quality metric based on the image quality metrics; the data processing system is configured such that the determination of the combined quality metric takes into account interactions between the image quality metrics with respect to their combined effect on the combined quality metric; the interaction between the image quality metrics is such that a change in the combined quality metric caused by a change in a first one of the image quality metrics is compensable by a change in one or more of the remaining image quality metrics; A system wherein the or each one of the image quality metrics is indicative of one or more parameters of a position and / or orientation of an imaged body part relative to components of the imaging unit or relative to a further body part.

2. The system of claim 1 , wherein the component comprises a detector of the imaging unit and / or a radiation source of the imaging unit.

3. each of the image quality metrics is associated with a corresponding predefined optimum value or a predefined optimum range; 3. The system of claim 1, wherein the interaction between the image quality metrics is such that an increase in deviation of a first one of the image quality metrics from the optimal range or value can be compensated for by a decrease in deviation of one or more of the remaining image quality metrics from the optimal range or value.

4. 4. The system of claim 1, wherein the system determines, for one or more of the medical images, parameters or state variables representing changes in imaging conditions for using the imaging unit based on the determined combined quality metrics, wherein the changed imaging conditions change one or more of the image quality metrics.

5. 5. The system of claim 1, wherein the determining of the image quality metric comprises segmenting at least a portion of the medical image and / or registering the medical image to an atlas using the segmented medical image.

6. 1. A system for analysis of medical image data representing a plurality of medical images, each of which may be a two-dimensional or three-dimensional image, comprising: the medical images are acquired from one or more human subjects using one or more medical imaging units, each of the medical imaging units acquiring a medical image; The system includes a data processing system, the data processing system comprising: reading and / or generating one or more image quality metrics for each of said medical images; receiving, via a user interface, user input for at least some of the medical images indicating a user-specified quality assessment of the corresponding image; The data processing system further comprises: (a) determining or adapting an algorithm for determining a combined quality metric based on the image quality metrics and based on the user input; and / or (b) determining or adapting an algorithm for classifying the medical image or selecting a portion of the medical image based on the user input and based on the combined quality metric; the combined quality metric is dependent on interactions between the image quality metrics for their combined effect on the combined quality metric; the interaction between the image quality metrics is such that a change in the combined quality metric caused by a change in a first one of the image quality metrics is compensable by a change in one or more of the remaining image quality metrics; A system wherein the or each one of the image quality metrics is indicative of one or more parameters of a position and / or orientation of an imaged body part relative to components of the imaging unit or relative to a further body part.

7. 7. The system of claim 6, further comprising: displaying, via the user interface, an output indicative of the combined quality metric for at least some of the medical images; and / or an output indicative of a classification of the medical images, the classification of the medical images based on the combined quality metric.

8. outputting, via the user interface, a graphical representation indicating a coordinate system for one or more of the image quality metrics; The system of claim 6 or 7, wherein for one or more of the medical images, the graphical representation further indicates a spatial relationship between the corresponding medical image and the coordinate system.

9. 9. The system of claim 6, wherein the data processing system receives, via the user interface, a user-specified selection of one or more of the medical images to input the user input indicating the user-specified quality assessment of the medical images, the medical images being selected by the user-specified selection.

10. 1. A method for the analysis of medical image data representing a two-dimensional or three-dimensional medical image, said image being an image of at least a portion of a human subject, said method being carried out using a data processing system, The method comprises: acquiring the medical image data using a medical imaging unit, wherein the medical imaging unit comprises an X-ray imaging unit and / or a magnetic resonance imaging unit; using said data processing system to read and / or determine a plurality of image quality metrics for said medical image; using the data processing system to automatically or interactively determine a combined quality metric based on the image quality metrics; Including, determining the combined quality metric takes into account interactions between the image quality metrics for a combined effect of the image quality metrics on the combined quality metric; the interaction between the image quality metrics is such that a change in the combined quality metric caused by a change in a first one of the image quality metrics is compensable by a change in one or more of the remaining image quality metrics; A method wherein the or each one of the image quality metrics is indicative of one or more parameters of a position and / or orientation of an imaged body part relative to components of the imaging unit or relative to a further body part.

11. 1. A method for the analysis of medical image data representing a plurality of medical images, each of which is a two-dimensional or three-dimensional image, the medical images being images of portions of one or more human subjects acquired from a medical imaging unit, comprising: The method is performed using a data processing system; The method comprises: using said data processing system to read and / or generate one or more image quality metrics for each of said medical images; receiving, via a user interface of the data processing system, user input for at least some of the images indicating a user-specified quality rating of the corresponding image; Including, The method comprises: (a) determining or adapting, using said data processing system, an algorithm for determining a combined quality metric based on said image quality metrics and based on said user input; and / or (b) determining or adapting, using said data processing system, an algorithm for classifying said images or selecting portions of said images; and the classifying or selecting is based on the user input and based on the combined quality metric; the combined quality metric is dependent on interactions between the image quality metrics for their combined effect on the combined quality metric; the interaction between the image quality metrics is such that a change in the combined quality metric caused by a change in a first one of the image quality metrics is compensable by a change in one or more of the remaining image quality metrics; A method wherein the or each one of the image quality metrics is indicative of one or more parameters of a position and / or orientation of an imaged body part relative to components of the medical imaging unit or relative to a further body part.

12. 1. A computer program for the analysis of medical image data representing a plurality of medical images, each of which is a two-dimensional or three-dimensional image, the medical images being images of portions of one or more human subjects obtained from a medical imaging unit; The computer program, when executed by a processor in a data processing system, using said data processing system to read and / or generate one or more image quality metrics for each of said medical images; receiving, via a user interface of the data processing system, user input for at least some of the medical images indicating a user-specified quality assessment of the corresponding medical image; Run The computer program further, when executed by the processor, (a) determining or adapting, using said data processing system, an algorithm for determining a combined quality metric based on said image quality metrics and based on said user input; and / or (b) determining or adapting, using said data processing system, an algorithm for classifying said images or selecting portions of said images; and performing at least one of the classifying or selecting is based on the user input and based on the combined quality metric; the combined quality metric is dependent on interactions between the image quality metrics for their combined effect on the combined quality metric; the interaction between the image quality metrics is such that a change in the combined quality metric caused by a change in a first one of the image quality metrics is compensable by a change in one or more of the remaining image quality metrics; A computer program product, wherein the or each one of the image quality metrics is indicative of one or more parameters of a position and / or orientation of an imaged body part relative to components of the medical imaging unit or relative to a further body part.

13. A computer readable medium having stored thereon the computer program of claim 12.

Images