Compressing image data from a medical imaging device

Abstract

A computer-implemented method for compressing image data from a medical imaging device, wherein the image data comprises pixels with pixel values ​​at defined positions, the method comprising the following steps: - setting at least one threshold for the pixel values ​​to assign the pixels to user data (3) and to background data on opposite sides of the threshold; - reducing the image data to data points belonging to the user data (3) and storing the pixel values ​​of the user data (3) without the pixel values ​​of the background data, as well as storing location information that allows the location of the data points of the user data (3) to be reconstructed together with the respective pixel values, on at least one computer-readable storage medium, wherein for reducing the image data a starting point and a linear path is defined that traverses the pixels of the image data.wherein, when reducing the image data, image point values ​​of directly adjacent image points of the user data (3) along the linear path are stored without location information and in a defined order, and wherein, in the case of gaps between image points of the user data (3) along the linear path, at least one location piece of information from which the gap can be reconstructed, in particular at least one coordinate of a coordinate position of the next image point on the linear path after the gap, which is changed compared to the last image point, is stored.

Description

[0001] Regardless of the grammatical gender of a particular term, persons with male, female or other gender identities are included.

[0002] The invention relates to a computer-implemented method for compressing image data of a medical imaging device, a method for recording and storing image data of a medical imaging device, a computer program product and a medical imaging device.

[0003] In a positron emission tomography (PET) medical imaging device, gamma quanta are slowed down by matter located between the PET signal source and the gamma detectors. The signal source can be defined as the point in space where electron-positron annihilation occurs, resulting in the emission of two gamma quanta in opposite directions. According to current technology, the gamma detectors can include LSO crystals, which act as scintillators by converting the gamma quanta into photons, and CMOS photodetectors, which convert these photons into electrons. This slowing of the gamma quanta weakens the PET signal, negatively impacting its signal quality.This signal attenuation should therefore typically be avoided as much as possible (for example, by ensuring that there are no obstacles in the path) or the PET signal is subsequently corrected. For this attenuation correction (AC), so-called µ-maps are used, which generally comprise pixels with pixel values ​​at defined positions. These µ-maps typically exist, for example, for the patient table and the patient (in the case of PET, PET-CT, and MR-PET; CT stands for computed tomography, MR for magnetic resonance imaging) and for local MR coils (usually only for MR-PET). These µ-maps are typically acquired either with a CT scanner (e.g., table, coils) or directly with MR (e.g., patient). After acquisition, this data is typically stored as voxel data resolved in three spatial dimensions.

[0004] The resulting data volumes are typically quite large, for example in the range of 1 to 8 gigabytes, and therefore require a correspondingly large amount of storage space. This can be particularly problematic when data needs to be archived and backed up, for example on online servers. Thus, large data volumes incur additional costs that typically correlate directly with the size of the data.

[0005] It would therefore be desirable to find a way to reduce the size of such data as much as possible. Currently, micro-map data is typically not compressed using state-of-the-art methods. Furthermore, common lossless data compression techniques often only result in a slight reduction in the data size of such micro-maps.

[0006] It is therefore an object of the present invention to find a way to reduce the size of micro-map data or to find an improved method for compressing micro-map data, preferably without any loss of information in the data. Furthermore, it would be desirable to find a general method for compressing or improving the compression of medical image data, or at least certain types of medical image data.

[0007] This problem is solved by a method according to claim 1, a method according to claim 8, a computer program product according to claim 9, and a medical imaging device according to claim 10. Further features and advantages will become apparent from the dependent claims, the description, and the accompanying figures.

[0008] According to a first aspect of the invention, a computer-implemented method for compressing image data from a medical imaging device is provided, wherein the image data comprises pixels with pixel values ​​at defined positions. The method comprises the following steps: - Setting at least one threshold for the pixel values ​​to assign pixels to user data and background data on different sides of the threshold; - Creating a data header containing comprehensive information about the applied compression, so that image data can be reconstructed from the compressed data based on the information in the data header; -Reducing the image data to data points belonging to the user data and storing the pixel values ​​of the user data without the pixel values ​​of the background data, as well as storing location information that allows the location of the data points of the user data to be reconstructed together with the respective pixel values, on at least one computer-readable storage medium.

[0009] Image data from PET attenuation maps (i.e., micro-maps) are typically multidimensional volume files sparsely populated with user data. This is primarily due to the fact that only a (relatively small) portion of the observed space is filled with signal-attenuating obstacles, such as a patient table (especially lightweight models with a large air content in the frame). The method according to the invention is particularly well-suited for such micro-maps or other medical image data sparsely populated with user data. Advantageously, the method according to the invention allows for data compression, enabling the image data (for example, the micro-maps from an MR-PET scanner) to be stored and read from a data carrier very easily, quickly, and in a space-saving manner. In particular, this reduces the costs of archiving the image data.Reading image data from a storage medium into the main memory of a user PC ("personal computer") can also be significantly faster due to the reduced file size. "Low data content" in this context means that a significant portion of the data points are not data (i.e., they are background data). Specifically, "sparse data content" can mean that a maximum of 85%, preferably a maximum of 50%, particularly preferably a maximum of 25%, and most preferably a maximum of 10% of the pixels are background data. The greater the proportion of data content to the total image data, the more effective the compression using the method according to the invention can be. The proportion of data content up to which effective compression can still be achieved can depend, in particular, on the specific implementation or the respective embodiment of the method according to the invention.With a particularly preferred embodiment, as described herein, effective compression of the image data can still be achieved even with a share of over 80% of the usable data.

[0010] The method according to the invention is a computer-implemented method. The method can, for example, be performed directly on a control unit of the medical imaging device. However, the method can also be performed externally by the medical imaging device. The image data can generally be any image data comprising user data and background data. Background data can, in particular, be pixels representing a substantially empty space, a space with a substantially zero signal, or a pixel value below a minimum threshold. Preferably, at least 15%, more preferably at least 50%, more preferably at least 75%, and most preferably at least 90% of the pixels in the image data are not user data. The image data can be image data of an attenuation map. In particular, the image data can be image data of a PET attenuation map (µ-map).The medical imaging device may be, in particular, a medical imaging device based on positron emission tomography (PET). For example, the medical imaging device may be or include a PET scanner, an MR-PET (magnetic resonance PET), and / or a PET-CT (PET-computed tomography) scanner. An MR-PET is a hybrid device combining PET and MRI (magnetic resonance imaging). A PET-CT is a combined PET-computed tomography device. Optionally, the image data may also be other image data from a medical imaging device that includes both payload and background data.

[0011] The image data can preferably be multidimensional image data, in particular volumetric image data. The image points can be, in particular, pixels or voxels. The image point values ​​can be, in particular, pixel values ​​or voxel values. For example, the positions of the image points can be defined by coordinates. For example, the coordinates of the image points can be specified by integer indices that uniquely indicate their position in the image, for example, in an image volume.

[0012] According to the invention, at least one threshold value is set for the pixel values. The threshold value can apply to the entire volume of the image data, i.e., it can be a fixed threshold value. Alternatively, the threshold value can apply locally to a sub-area of ​​the volume and / or be location-dependent. For example, several threshold values ​​can be provided, each assigned to a sub-area. A threshold value that applies locally to a sub-area or differs for different sub-areas can be called an adaptive threshold value. Adaptive threshold values ​​are useful, for example, when the boundary between user data and background data depends on the position of the pixels, in particular voxels, within the volume of the image data. This can be the case, for example, with brightness fluctuations in a video stream. This allows the pixels to be assigned to user data and background data.In particular, background data can be pixels whose pixel values ​​represent a background to the image data, e.g., air. The threshold can be a predefined threshold. For example, the threshold can be determined based on the type of image data. For instance, the threshold can be determined in a calibration procedure. A correspondingly defined threshold can then be used repeatedly for the same type of image data from the same medical imaging device, or always set to the same value. Specifically, the threshold can be set by retrieving the corresponding known threshold. Optionally, the threshold can also be set or defined based on prior knowledge. For example, it is typically known which values ​​occur in PET attenuation maps and what these values ​​represent.Accordingly, for PET attenuation maps, the threshold can be set so that pixel values ​​corresponding to an empty space or air are assigned as background data.

[0013] Furthermore, a data header is created that contains information about the compression method used. This data header allows the image data compressed using this method to be reconstructed. The data header includes information about the compression method used. For example, it can contain information about which specific compression method was employed. Information for decoding the compressed image data can also be stored in the data header. For example, the dimensions of the image data can be included. Additionally, optionally assigned spatial axes and / or a position of the image data in world coordinates can be included. A position of the image data in world coordinates can indicate where the image data was acquired, for example, in relation to the position of the medical imaging device.Optionally, information about the object represented by the image data can be included in the data header. For example, the information "patient bed" or "body coil #2" can be stored in the data header.

[0014] The image data is reduced to data points corresponding to the user data, and the pixel values ​​of the user data are stored without the pixel values ​​of the background data. With sparse image data, this reduces the required storage compared to the state of the art by no longer explicitly storing the pixels of the background data. Additionally, location information is stored, allowing the location of the pixels in the user data to be reconstructed along with their respective pixel values. It can be provided that at least one piece of location information and / or at least one pixel value from at least one pixel, together with a control command describing the compression, is stored on at least one common byte. The pixel values ​​of the user data and / or the location information are stored, in particular, on a computer-readable storage medium or data carrier.The storage medium can be local or remote. For example, it can be a hard drive, SSD, flash memory, online server, or other suitable storage medium. The computer-readable storage medium can be non-volatile. The pixel values ​​of the user data and / or location information can be permanently stored on the computer-readable storage medium.

[0015] According to one embodiment, when reducing image data, the pixel values ​​of the user data are stored together with a coordinate position. Compression according to this embodiment can also be referred to as point cloud encoding compression. In particular, the pixel values ​​of the user data can be stored as a list of pixels, where each pixel is defined as a tuple comprising the pixel value and the position. This allows for a significant reduction in storage requirements when dealing with sparse image data. The coordinate position can, for example, be defined based on a volume grid.

[0016] According to one embodiment, for reducing the image data, a starting point and a linear path are defined that traverses the image data pixels. During the reduction of the image data, pixel values ​​of directly adjacent pixels of the user data along the linear path are stored without location information and in a defined order. In the case of gaps between pixels of the user data along the linear path, at least one piece of location information is stored from which the gap can be reconstructed. The location information from which the gap can be reconstructed can include coordinate information. In particular, the location information from which the gap can be reconstructed can describe at least one coordinate, and especially all coordinates, of a coordinate position of the next pixel along the linear path after the gap that has changed compared to the last pixel.Preferably, only the coordinates of the position that have changed compared to the coordinates of the last voxel along the linear path are stored as part of the location information. Advantageously, this embodiment allows for storing fewer individual coordinates overall, thus reducing the required storage space. Storing fewer individual coordinates is made possible in particular by the defined linear path. The linear path can be defined in the data header. For example, a starting point and the path of the linear route can be defined in the data header. Especially when many adjacent pixels are to be assigned to the user data, a further significant reduction in storage space can be achieved. Typically, adjacent pixels are to be expected in corresponding image data when the image data concerns related objects.It may be possible to traverse a linear path during the compression or reduction of the image data, comparing each pixel value with at least one threshold value to determine whether the respective pixel belongs to the user data. In particular, each pixel value can be compared to a fixed or adaptive threshold value. This advantageously allows for the direct determination of which pixels belong to the user data. It may also be possible to store a control command describing the compression process along with location information and / or a pixel value. For example, a control command could specify that stored information is location information in a particular coordinate direction. Similarly, a control command could specify that stored information is a pixel value.For example, one control command can be provided for each coordinate direction, and another control command can be provided to specify the pixel values. For example, N+1 control commands can be provided for N-dimensional image data. For example, 4 control commands can be provided for three-dimensional volume data. It can be provided that the location information from which the gap can be reconstructed and / or at least one pixel value is stored together with a control command on at least one common byte. For example, according to this embodiment, it may be possible to store a control command on only two bits. Therefore, it may be advantageous not to use a separate byte for the control command, but rather to store the respective location information and / or at least one pixel value together with the control command. It can be provided that a data packet length is defined, which describes a number of bytes per unit of information.An information unit can, in particular, comprise one control command as well as at least one location piece of information and / or at least one pixel value. For example, a data packet length of 16 bits to 32 bits can be specified.

[0017] According to one embodiment, coordinates of the coordinate system are defined for reducing the image data, with the linear path being defined such that it traverses the coordinates as nested loops, with the first set of coordinates as the innermost loop and the last set of coordinates as the outermost loop. The coordinate system can correspond to a defined coordinate system of the medical imaging device. For example, the coordinate system can be a Cartesian coordinate system, particularly with the coordinates X, Y, and Z. However, another suitable coordinate system can also be chosen, such as a two-dimensional or, more generally, N-dimensional coordinate system, or a cylindrical coordinate system. Nested loops are to be understood in the sense of the term used in programming.For example, the linear path can be defined as nested counting loops, where, in particular, the coordinate value of the coordinates in the counting loops is incrementally increased according to the resolution of the respective coordinates. The coordinate of the innermost loop can be considered the preferred direction. The preferred direction can, in particular, correspond to the principal longitudinal extent of an object in the image data. Preferably, the loops are defined in such a way as to optimize compression.

[0018] According to one embodiment, the order of the nested loops is chosen such that the number of gaps along the linear path is minimized, and / or the data size of the compressed image data is minimized, and / or the coordinate of the innermost loop contains the fewest gaps, and / or the coordinate of the innermost loop describes the largest dimension of the image data. Advantageously, the nesting of the loops can thus be chosen so that the compression is most effective or the size of the image data is reduced the most. For example, the nesting can be determined relatively easily by trial and error, by applying compression to a representative example of image data with variable nesting of the loops of coordinates.Typically, a nesting structure defined in this way can be reused in the future, as it can be assumed that similar problems favor corresponding nesting methods. The chosen nesting structure is preferably stored in the data header, particularly to enable later data reconstruction.

[0019] According to one embodiment, the coordinate selected as the first coordinate for the innermost loop is the one along whose direction there is the greatest number of directly adjacent pixels in the user data. This measure can make compression particularly efficient.

[0020] According to one embodiment, the image data, in particular the volume data to be compressed, are image data of an attenuation map, specifically an attenuation map for a positron emission tomography (PET) scanner. Within the scope of this invention, an attenuation map for a PET scanner can also be referred to as a PET attenuation map. Such an attenuation map is also called a µ-map. Due to the often relatively large amount of background data on such attenuation maps, the method according to the invention is particularly well suited for attenuation maps. Background data in attenuation maps are, in particular, pixels where essentially no attenuation is recorded.

[0021] According to one embodiment, the image data are image data of at least one hardware component of the medical imaging device, in particular image data of a magnetic resonance imaging hardware component, for example image data of a patient bed, in particular an MR patient bed, a magnetic resonance imaging coil and / or a magnetic resonance imaging phantom.

[0022] According to one embodiment, the image data is image data of a patient bed. In particular, the image data can be a PET attenuation map of a patient bed. For example, most pixels of a PET attenuation map of a patient bed are typically black in cross-section, or represent air. These pixels are then primarily background data. Accordingly, there are usually only relatively few truly relevant data points that are relevant, for example, for attenuation correction. Due to the typically relatively large number of background data points in the image data of a patient bed, the method according to the invention can therefore be particularly effective when applied to image data of a patient bed.

[0023] According to one embodiment, the image data consists of images of a patient bed, with the first coordinate for the innermost loop being chosen as a coordinate parallel to a longitudinal direction of the patient bed. This embodiment takes into account that a patient bed typically contains the most contiguous data in the image data along its longitudinal direction. Therefore, compression can be particularly effective.

[0024] According to one embodiment, during the step of setting the at least one threshold, an input option is provided for a user to adjust the threshold, whereby user input regarding the at least one threshold is received and the at least one threshold is adjusted based on the user input. The input option can allow the process to be customized according to user preferences. The input option can enable individual optimization of the compression tailored to specific image data.

[0025] Another aspect of the invention is a method for recording and storing image data from a medical imaging device, in particular image data of an attenuation map for a positron emission tomography device, comprising the following steps: - Acquiring image data with an imaging device; - Compress and store the image data as described herein.

[0026] All the advantages and features of the method for compressing image data from a medical imaging device can be applied analogously to the method for acquiring and storing image data, and vice versa. The medical imaging device may be, for example, a PET scanner, an MR-PET (magnetic resonance PET), and / or a PET-CT (PET-computed tomography) scanner. For example, the image data may be acquired using a computed tomography (CT) scanner, a PET scanner, or a magnetic resonance imaging (MRI) scanner.

[0027] According to one embodiment, the image data are first acquired as image data by the imaging device and, after acquisition by the imaging device, converted into an attenuation map for a positron emission tomography (PET) device, wherein the pixel values ​​of the attenuation map are based, in particular, on a linear attenuation coefficient. The image data can be a PET attenuation map acquired by CT, PET, or MRI, or be based on such an acquisition.

[0028] Another aspect of the invention is a computer program product or a computer-readable storage medium comprising instructions which, when executed by a computer, cause the computer to perform the steps of the method as described herein. All the advantages and features of the method for compressing image data from a medical imaging device and the method for acquiring and storing image data can be transferred analogously to the computer program product and vice versa. The computer program product can, for example, be stored on the computer-readable storage medium, in particular a non-volatile storage medium. The storage medium can be, for example, a hard drive, an SSD, flash memory, an online server, etc.

[0029] Another aspect of the invention is a medical imaging system comprising an imaging device, wherein the imaging system is configured to perform a method as described herein. All the advantages and features of the method for compressing image data of a medical imaging device, the method for acquiring and storing image data, and the computer program product can be transferred analogously to the medical imaging system and vice versa.

[0030] Another aspect of the invention is image data from a medical imaging device, wherein the image data is compressed using a method as described herein. All the advantages and features of the method for compressing image data from a medical imaging device, the method for acquiring and storing image data, the computer program product, and the imaging system can be applied analogously to the image data of a medical imaging device, and vice versa.

[0031] Another aspect of the invention is a method for reading image data from compressed data that has been compressed according to a method described herein. The method comprises the following steps: - Reading the data header to obtain information about the compression applied; and - Reconstructing the image data from the compressed data based on information about the applied compression, in particular converting the image data back into the uncompressed format.

[0032] All the advantages and features of the method for compressing image data from a medical imaging device, the method for acquiring and storing image data, the computer program product, the imaging system, and the image data of a medical imaging device can be applied analogously to the method for reading image data from compressed data, and vice versa. Information about the applied compression can, for example, include the number of dimensions of the image data and their size. When reconstructing the image data, the actions used for compression can be performed in reverse. For example, an initial value of a linear path can be determined based on which the image data was compressed, and image points can be read along the linear path, particularly along nested loops of the linear path.In particular, the same control commands used during compression can be used to assign and decode the stored compressed data and write the image values ​​to the correct target position. Optionally, image data formatting can then be performed. For example, if the image data is PET attenuation map data, formatting can be applied to allow attenuation correction. For instance, the resolution and / or scaling of the image data can be adjusted to a required resolution through interpolation and, optionally, multiplanar reconstruction.

[0033] All embodiments described herein can be combined with one another, unless explicitly stated otherwise.

[0034] The following describes embodiments with reference to the attached figures. Fig. Figure 1 shows a layer of a PET attenuation map of a patient bed in a transverse section or cross-section. Fig. Figure 2 shows a flowchart of a computer-implemented method for compressing image data from a medical imaging device according to an embodiment of the invention. Fig. Figure 3 shows a representation of a spatial volume with three-dimensional image data in a coordinate system to illustrate an embodiment of this invention. Fig. Figure 4 shows a diagram in which the size of the image data is plotted as a function of the proportion of useful data in the image data before and after the inventive method according to different embodiments. Fig. Figure 5 shows a medical imaging system comprising an imaging device according to an embodiment of the invention and Fig. Figure 6 shows a flowchart of a method for reading image data from compressed data according to an embodiment of the invention.

[0035] Fig. Figure 1 shows a layer of a PET attenuation map of a patient bed in a transverse section. Such an attenuation map can be created, for example, by a computed tomography (CT) scan. The Hounsfield data from the CT scan can be converted into PET attenuation values, specifically LAC values, using established conversion methods. LAC stands for Linear Attenuation Coefficient. LAC values ​​are typically given in units of 1 / cm. As can be seen in this cross-section, most pixels, in this case voxels, are black, meaning they have a value of zero or approximately zero. The composite plastic of the patient bed (dark white) and significantly denser copper components, such as copper wires, copper screws, etc. (light white), are shown in light white. The white voxels (light and dark white) correspond to usable data 3.The black voxels correspond to background data. Essentially, the useful data 3 are relevant for attenuation correction, meaning that, compared to the total volume, only a relatively small number of voxels actually contain relevant data. The same applies to attenuation maps of MR coils (e.g., head coils). MR coils are also voluminous, but only sparsely populated with significant voxel data, i.e., within the scope of the invention, particularly useful data 3. Image data of this type can be referred to as sparsely populated image data within the scope of the invention. The inventive method for compressing the image data can be particularly advantageous for sparsely populated image data. In general, the inventive method can also be applied to sparsely populated image data, especially sparsely populated multidimensional volume data.In general, for such data, voxels can be specified by integer indices that uniquely define the position of the voxels within the volume. For example, a position can be specified by the coordinates (x1, x2, x3... x). n ) with x i as an index along a volume axis X i and n of the maximum dimension of the volume. For a three-dimensional volume, n=3. Furthermore, each pixel, or in this example voxel, contains a pixel value. The pixel value can also be referred to simply as a "value". The pixel value of a voxel can also be called the voxel value. In general, the pixel value does not necessarily have to be an integer scalar. For example, the pixel value can contain arbitrarily complex values ​​(e.g., vectors), especially with a required or desired precision / bit depth. In the specific example of Fig. 1. The pixel value indicates the strength of the attenuation, which in Fig. 1 is represented by the degree of brightness, so that greater brightness means greater attenuation. The pixel value can be expressed as a value (x1, x2, x3... x) depending on the position. n ) are specified. Attenuation maps typically contain integer, positive intensities with 16-bit resolution as pixel values ​​or voxel values. In the case of an attenuation map of the patient bed as in Fig. As can be seen, for example, a 1 mm scan can result in a volume with a total of 2200x400x600 voxels. Each voxel can, for instance, have 16-bit pixel values ​​(corresponding to the accuracy of the attenuation intensity value). In this example, the uncompressed data size is approximately 1 GB (1056 MB). Typically, such image data is stored linearly in memory, sorted by rows, columns, and layers of the volume image data. The position of a voxel in memory can then be specified as a memory index for linear memory access, e.g., according to: memoryIndex=(((z∗VolumeSizeY)+y)∗VolumeSizeX+x)

[0036] This allows the pixel value or voxel value to be determined, e.g. in the form Value=muMap[memoryIndex], where the position of the voxel is given in integer volume coordinates (x, y, z). If one assumes that only a small portion, for example one percent, of the volume data contains useful information, i.e., is payload data 3, then a significant reduction in the required storage capacity can be achieved with a method according to the invention.

[0037] Fig. Figure 2 shows a flowchart of a computer-implemented method for compressing image data from a medical imaging device according to an embodiment of the invention. As with reference to Fig. As described in example 1, the image data contains pixels with pixel values ​​at defined positions. In a first step 101, at least one threshold value for the pixel values ​​is set. The threshold value serves to assign the pixels to user data 3 and to background data on different sides of the threshold. The threshold value can be set, for example, by retrieving it from a database, from storage, or from another source where it is stored. Alternatively, the threshold value can also be set, for example, by user input.

[0038] In a further step, 102, a data header is created. The data header contains information about the applied compression, so that a reconstruction of the compressed data is possible based on the information in the data header.

[0039] In a further step 103, the image data is reduced to data points belonging to the user data 3, and the pixel values ​​of the user data 3 are stored without the pixel values ​​of the background data. Furthermore, location information is stored that allows the location of the data points of the user data 3 to be reconstructed together with the respective pixel values. During the reduction of the image data, the pixel values ​​of the user data 3 are each stored together with a coordinate position.

[0040] According to one embodiment, when reducing the image data, the pixel values ​​of the user data 3 are each stored together with a coordinate position. Within the scope of the invention, this can also be referred to as point cloud encoding. In particular, the image data of the user data 3 can be stored as a list of voxels, where each voxel is defined as a (position, value) tuple. In the example, which refers to Fig. As explained in section 1, this can mean that 8 bytes of storage space are required per voxel. These 8 bytes consist of the three coordinates (x, y, z) with 2 × 8 bits and the pixel value or voxel value with a resolution of 2 bytes (thus 3*2+1*2 = 8 bytes). Under the exemplary assumption made above of 1% user data and 99% background data (especially air), this results in a reduction of the data volume to 42.3 MB, and thus approximately 4% of the original image data of about 1 GB according to this example.

[0041] According to a further embodiment, a starting point and a linear path are defined for reducing the image data. This path traverses the image data pixels. During the reduction process, pixel values ​​of directly adjacent pixels of the user data 3 along the linear path are stored without location information and in a defined order. In the case of gaps between pixels of the user data 3, at least one piece of location information is stored along the linear path from which the gap can be reconstructed. This location information can, in particular, describe the changed coordinates of the next pixel along the linear path after the gap. Preferably, coordinates of the coordinate system are defined for reducing the image data. These coordinates, or the coordinate system, are stored, in particular, in the data header.The linear path is then defined such that it traverses the coordinates as nested loops, with the first set of coordinates as the innermost loop and the last set of coordinates as the outermost loop. The loops are preferably chosen to maximize compression. In the example from... Fig. 1. The coordinate of the innermost loop can preferably correspond to the longitudinal direction of the patient table, i.e., in particular, the direction extending into the image plane. The longitudinal direction of the patient table is often defined as the z-direction in medical imaging devices. Reducing the image data according to this embodiment is described with reference to Fig. 3. Explained in more detail using examples and excerpts.

[0042] Fig. Figure 3 shows a representation of a spatial volume 1 containing three-dimensional image data in a coordinate system with coordinates x, y, and z. The image data shown here can be considered a simplified example of image data acquired using a suitable imaging modality (e.g., computed tomography or magnetic resonance imaging). According to the invention, a threshold value for the pixel values ​​(in this example, voxel values) is defined, which separates the useful data 3 from the background data. The useful data 3 of the image data are represented by small cuboids. It can be seen that in this example, only a relatively small portion of the image data constitutes useful data 3. Consequently, the image data is sparsely populated. Positions in the spatial volume 1 where there is no useful data 3 are correspondingly occupied by background data, which in this representation corresponds to empty space.It can be seen that a large portion of the payload 3 lies on a ray in the direction of the z-axis. For this reason, the z-coordinate is defined here as the innermost loop. For example, the x-coordinate can be set as the middle loop and the y-coordinate as the outermost loop. The information about which coordinates are used and which coordinate is used as the innermost, middle, and outermost loop is stored in the data header of the compressed image data. Further information, such as the world coordinates of the image data, can also be stored in the data header. To compress the data, all pixels (in this example, voxels) of the spatial volume 1 of the image data are now processed by the nested loops (especially counting loops or "for loops") of the coordinates x, y, z, i.e.,As a voxel is visited along a linear path through the spatial volume, its value is compared to the threshold value, and the voxels are assigned to user data or background data. The goal is to obtain a list of voxel values ​​(generally: pixel values) with an additional coordinate encoding that serves as location information for the voxel values. In this example, the coordinate encoding can be provided for three-dimensional data using 2 bits, which can function as control commands. For example, the following states can be described using the two bits: - Bit 00: “SAVE -The `SAVE_Value` command is called when the voxel value is not zero, meaning there is a voxel that belongs to the payload 3. The corresponding voxel value is then stored after these bits. This can, for example, take the form `SAVE_Value(VoxelValue_i)`. It is also implicitly intended that the linear path continues to the next voxel after these bits. No coordinate is explicitly stored for this. A stored sequence of several of these control commands with their associated voxel values ​​in succession therefore means that several consecutive voxels belong to the payload 3. If the next voxel on the linear path does not belong to the payload 3, the linear path continues until the next voxel that does belong to the payload 3. The following three bits are then used in addition. - Bit 01: “SAVE_X” - SAVE_X is called when the x-coordinate changes or has changed and the voxel value is not zero. The x-coordinate of the respective voxel is stored behind these bits. This can, for example, take the form SAVE_X(PositionValue_i_X). In other words, these bits specify the next x-position of the data 3 along the linear path, provided the corresponding voxel has a different x-coordinate than the last voxel of the data 3 along the linear path. - Bit 10: “SAVE_Y” - SAVE_Y is called when the y-coordinate changes and the voxel value is not zero. The y-coordinate of the respective voxel is stored behind these bits. This can, for example, take the form SAVE_Y(PositionValue_i_Y). In other words, these bits specify the next y-position of the data 3 along the linear path, provided the corresponding voxel has a different y-coordinate than the last voxel of the data 3 along the linear path. - Bit 11: “SAVE - The `SAVE_Z` function is called when the Z coordinate changes and the voxel value is not zero. The z-coordinate of the respective voxel is stored behind these bits. This can, for example, take the form `SAVE_Z(PositionValue_i_Z)`. In other words, these bits specify the next z-position of the data 3 along the linear path, provided the corresponding voxel has a different z-coordinate than the last voxel of the data 3 along the linear path.

[0043] These bits allow the compressed data to be written into a linear, binary stream. For example, the linear path can begin at the origin of the defined coordinate system, position (0,0,0) in the upper left corner. If a voxel of the payload data is present at this position, the control command is executed, and the corresponding "SAVE" bits are set. - Value applied. In the example of Fig. If no voxel of payload 3 is present at the origin of the coordinate system, the process continues along the linear path. This occurs first along the z-coordinates (innermost loop) and then along the outer loops until a voxel of payload 3 is reached. Its position is then recorded by the bits SAVE_X, SAVE_Y, and SAVE_Z, and its voxel value is recorded by the bits "SAVE_Value". This skips the gap of background data, and only the position and voxel value of the next voxel containing payload 3 are saved. These steps are repeated until all voxel values ​​of payload 3 have been copied to the target file.

[0044] In the following example, based on Fig. In section 3, it is assumed that all voxels of the data 3 have the same voxel value of 128. Furthermore, it is assumed that when writing to the target file, we have just reached the beginning of the first segment 32 of the data 3 ray in the z-direction. In other words, we have just arrived at the first voxel 35 of the data 3 ray. The saving of segments 32 and 34 of the data 3 ray is now described as an example. The first voxel 35 of the ray is located at the coordinate position (x=20, y=30, z=0). Since all coordinates have changed after the previous voxel of the data 3, the position of voxel 35 is first saved using the control commands SAVE_X(20), SAVE_Y(30), SAVE_Z(0). Additionally, the voxel value of voxel 35 is saved using SAVE_Value(128). Since the next voxel of the payload 3 is directly adjacent on the linear path, its voxel value is also saved next using SAVE_Value(128).The following voxel values ​​are processed in this way until the last voxel, 36, of this contiguous section, 32, is reached. The gap, 23, containing background data where the voxels have a voxel value of zero, is then skipped, and the z-coordinate of the first voxel, 37, of the second section, 34, containing payload 3 of this z-ray, is saved using SAVE_Z(11). Thus, the first voxel, 37, is located at z-position 11. The other coordinates (x, y) are not explicitly saved here because they have not changed. Starting with the first voxel, 37, of this section, 34, and continuing to the last voxel, 38, the voxel values ​​of 128 of these voxels are saved using the control command SAVE_Value(128). This can then be stored in the target file according to the following scheme: SAVE_X(20), SAVE_Y(30), SAVE_Z(0), SAVE_value(128), SAVE_value(128), SAVE_value(128), SAVE_value(128), SAVE_value(128), SAVE_value(128), SAVE_value(128), SAVE_value(128), SAVE_Z(11), SAVE_value(128), SAVE_value(128), SAVE_value(128), SAVE_value(128), SAVE_value(128), SAVE_value(128), SAVE_value(128), SAVE_value(128)

[0045] Expressed in bits (using the bits "b01", "b10", "b11" and "b00"), this section could look like this, for example: b01 20 b10 30 b11 0 b00 128 b00 128 b00 128 b00 128 b00 128 b00 128 b00 128 b00 128 b11 11 b00 128 b00 128 b00 128 b00 128 b00 128 b00 128 b00 128 b00 128

[0046] Preferably, the coordinates and pixel values ​​are stored together with the respective control command on at least one shared byte of a data carrier. According to the example given here, individual control commands require only two bits of storage space on the data carrier. Therefore, it can be advantageous not to use a separate byte for the control command, but rather to store the respective location information or pixel value together with the control command. This can further improve the compression rate. For example, a data packet length (number of bytes) can be defined, which applies specifically to all information units in a target data stream. Starting, for example, with a data packet length of 2 bytes (corresponding to 16 bits), 14 bits can be allocated for the user data and / or for the X,YZ coordinates.The remaining two bits are reserved for the respective control command, which distinguishes whether this data packet contains an X, Y, Z coordinate or an image data value. Other data packet lengths are also conceivable; for example, data packets with a 32-bit length are possible, where 30 bits are reserved for the coordinates or image values ​​of the user data and 2 bits for the respective control command.

[0047] To read the compressed file, steps corresponding to the compression process can be applied, with the source and target roles reversed. The data can be read from the compressed file and converted into the uncompressed format by following the same linear path. First, the data header is read, and the number of dimensions and their sizes are recorded. Then, the pointer position is set to the starting value (0,0,0). Next, in the N nested loops (in this example, n = 3, and z is the innermost loop), all pixels (in this example, voxels) are read. The SAVE_X / Y / Z commands stored before each pixel or voxel value are used to shift the position within the integer volume 1. Afterward, the next pixel or voxel value is written to the current target position. For example, in the case of a PET attenuation map, it might be necessary to...µ-Map then performs a formatting of the three-dimensional image data to enable attenuation correction according to a standard format. For example, the resolution and / or scaling of the image data can be adjusted to the required resolution by interpolation and, if necessary, multiplanar reconstruction (MPR).

[0048] Fig. Figure 4 shows a diagram in which the size of the image data is plotted as a function of the proportion of user data 3 in the image data before and after the method according to the invention, according to various embodiments. It shows a curve 11 of the size of uncompressed image data, a curve 12 of the size with point cloud encoding, and a curve 13 of the size with compression as described above. Fig. 3 Described and shown. The uncompressed data exhibits a constant profile 11, since no distinction is made between user data 3 and background data, and all pixels are treated equally. It is evident that compression according to point cloud coding can be advantageous in this example as soon as the proportion of user data 3 is lower than approximately 27%. However, with a larger proportion of user data 3, this coding requires more storage than in the uncompressed format. Compression based on a linear path, as in relation to Fig. In contrast, the method described in Section 3 can enable a significantly better reduction of the target size. In particular, an advantage can be achieved up to a user data share of approximately 85%. The compression method according to the invention is therefore well suited to reducing the storage size of sparsely populated files (such as µ-maps or PET attenuation maps, which typically have a user data share in the range of 3% to 10%).

[0049] Fig. Figure 5 shows a medical imaging system comprising an imaging device 4 according to an embodiment of the invention. The imaging device 3 can, for example, be an MR-PET or a PET-CT. The imaging system is configured to include a method for compressing image data as described herein, for example, with reference to Fig. 2 and / or Fig. 3. The procedure can, for example, be executed automatically at a control station 5 of the imaging system.

[0050] Fig. Figure 6 shows a flowchart of a method for reading image data from compressed data that has been compressed according to a method according to the invention, according to an embodiment of the invention. In step 210, the data header is read to obtain information about the compression applied. In particular, a number of dimensions of the image data and their size can be obtained. In a further step 220, the image data is reconstructed from the compressed data, in particular by converting the image data back into the uncompressed format. In particular, the actions used for compression can be performed in reverse. For example, starting from a compression such as with reference to Fig. As described in section 3, in step 221 a starting value of the linear path can be determined, e.g., at the coordinate position (0 / 0 / 0). In a further step 222, the pixels, in particular voxels, can be read in the nested loops, for example, three loops. The same control commands (e.g., according to those with reference to) can be used. Fig.The control commands SAVE_Value, SAVE_X, SAVE_Z (described in section 3), which were used during compression, are employed to enable the assignment and decoding of the stored compressed data and to write the image values ​​to the correct target position. Optionally, in a further step, the image data can be formatted. For example, if the image data is PET attenuation map data, formatting can be performed to allow attenuation correction. For instance, the resolution and / or scaling of the image data can be adjusted to a required resolution through interpolation and, optionally, multiplanar reconstruction (MPR).

Claims

[1] Computer-implemented method for compressing image data from a medical imaging device, wherein the image data comprise pixels with pixel values ​​at defined positions, the method comprising the following steps: - Setting at least one threshold for the pixel values ​​to assign the pixels to user data (3) and to background data on different sides of the threshold; - Creating a data header containing comprehensive information about the applied compression, so that image data can be reconstructed from the compressed data based on the information in the data header; - Reducing the image data to data points belonging to the user data (3) and storing the pixel values ​​of the user data (3) without the pixel values ​​of the background data, as well as storing location information that allows the location of the data points of the user data (3) to be reconstructed together with the respective pixel values, on at least one computer-readable storage medium. [2] Method according to claim 1, where a starting point and a linear path are defined for reducing the image data, which traverses the pixels of the image data, wherein when reducing the image data, image point values ​​of directly adjacent image points of the user data (3) along the linear path are stored without location information and in a defined order, and wherein in the case of gaps between image points of the user data (3) along the linear path at least one location information from which the gap can be reconstructed, in particular at least one coordinate of a coordinate position of the next image point on the linear path after the gap, which is changed compared to the last image point, is stored. [3] Method according to claim 2, where coordinates of the coordinate system are defined for reducing the image data, where the linear path is defined such that it traverses the coordinates as nested loops with a first of the coordinates as the innermost loop and a last of the coordinates as the outermost loop. [4] Method according to claim 3, wherein the sequence of the nested loops is selected such that the number of gaps along the linear path is minimized and / or that the data size of the compressed image data is minimized and / or that the coordinate of the innermost loop includes the fewest gaps and / or that the coordinate of the innermost loop describes the largest dimension of the image data. [5] Method according to claim 3 or 4, wherein the coordinate selected as the first coordinate for the innermost loop is the one along whose direction there is the largest number of directly adjacent pixels of the user data (3). [6] Method according to one of the preceding claims, wherein the image data are image data of at least one hardware component of the medical imaging device, in particular image data of a patient bed, a magnetic resonance imaging coil and / or a magnetic resonance imaging phantom. [7] Method according to one of the preceding claims, wherein the image data are image data of an attenuation map, in particular an attenuation map for a positron emission tomography device. [8] Method for recording and storing image data from a medical imaging device, in particular image data of an attenuation map for a positron emission tomography device, comprising the following steps: - Acquiring image data with an imaging device; - Compressing and storing the image data using a method according to one of the preceding claims. [9] Computer program product or computer-readable storage medium comprising instructions which, when executed by a computer, cause the computer to perform the steps of the method according to any of the preceding claims. [10] Medical imaging system comprising an imaging device (4), wherein the imaging system is configured to perform a method according to claim 8.

Images