(19)
(11)EP 3 320 843 B1

(12)EUROPEAN PATENT SPECIFICATION

(45)Mention of the grant of the patent:
17.06.2020 Bulletin 2020/25

(21)Application number: 17177310.4

(22)Date of filing:  22.06.2017
(51)International Patent Classification (IPC): 
G06T 5/00(2006.01)
A61B 6/00(2006.01)
G06T 7/33(2017.01)
G06T 7/00(2017.01)
G06T 5/50(2006.01)

(54)

METHOD AND APPARATUS FOR GENERATING A SET OF PROCESSED IMAGES

VERFAHREN UND VORRICHTUNG ZUR ERZEUGUNG EINES SATZES AUS VERARBEITETEN BILDERN

PROCÉDÉ ET APPAREIL DE PRODUCTION D'UN ENSEMBLE D'IMAGES TRAITÉES


(84)Designated Contracting States:
AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

(43)Date of publication of application:
16.05.2018 Bulletin 2018/20

(73)Proprietor: Siemens Healthcare GmbH
91052 Erlangen (DE)

(72)Inventor:
  • Manhart, Michael
    90765 Fürth (DE)


(56)References cited: : 
US-A1- 2005 127 910
US-A1- 2015 208 996
US-A1- 2008 152 088
US-A1- 2016 247 325
  
      
    Note: Within nine months from the publication of the mention of the grant of the European patent, any person may give notice to the European Patent Office of opposition to the European patent granted. Notice of opposition shall be filed in a written reasoned statement. It shall not be deemed to have been filed until the opposition fee has been paid. (Art. 99(1) European Patent Convention).


    Description


    [0001] The invention relates to a method and an apparatus for generating a set of processed images of an object from overlapping raw images. In particular, the present invention can be used for a peripheral digital subtraction angiography.

    [0002] US 2008 / 0 152 088 A1 discloses a method for long length imaging with a digital radiography apparatus. Therein, a set of imaging positions for an exposure series is determined. Responsive to an operator command, an image is automatically captured at each imaging position. A long length image is then generated by combining two or more partial images.

    [0003] US 2005 / 0 127 910 A1 discloses a method of data-processing to form a compound object data set from a plurality of basis datasets. The basis datasets assign datavalues to spatial positions in at least three-dimensional space, and are associated with mutually overlapping regions. From the datavalues in the basis datasets, by way of interpolation compound datavalues for the compound object data set are calculated in the overlapping regions.

    [0004] The DE 10 2008 003 945 B3 discloses a method for creating images relating to a system of blood vessels of a patient using digital subtraction angiography. Therein, multiple images are taken in different positions following a dispersal of a contrast agent through the system of blood vessels. From a first image an image part to that the contrast agent has not yet spread is used as a mask for a spatially corresponding image part of a second image.

    [0005] US 8 792 616 B2 discloses an x-ray image diagnosis apparatus and an x-ray image processing method. Therein, an imaging unit can perform a step movement process such that images of a subject are taken at a plurality of stages. A plurality of regions of interest for a plurality of locations in both legs of the subject is set when an image is taken after a contrast medium is injected into a lower limb of the subject. A change of an image level in the regions of interest is then measured to detect a flow of the contrast medium. The imaging unit and/or a top panel is then moved to a next imaging stage on the basis of a change in the image level of a leg in which the contrast medium flows more slowly. Image data taken at the plurality of stages is then processed.

    [0006] It is an objective of the present invention to provide a means for obtaining improved processed images contrasted with a contrast agent, with improved flexibility and usability.

    [0007] This objective is achieved by a method having the features of patent claim 1 and an apparatus having the features of patent claim 10. Advantageous embodiments with expedient developments of the present invention are indicated in the dependent patent claims as well as in the following description and the drawings.

    [0008] A method according to the present invention is concerned with generating a set of processed images of an object from overlapping raw images, wherein the raw images are obtained by means of a single contrast run. A contrast run refers to the process of injecting a contrast agent into the object and taking a series of raw images of the object using an imaging device to picture different stages of dispersal of the contrast agent in the object. The imaging device may for example be an x-ray apparatus. The images can therefore for example be x-ray images. After the raw images are obtained and provided for processing, they are aligned so that pixels representing the same part of the pictured object in different raw images are aligned forming a respective pixel stack for each pictured part of the object. This means that each pixel is correlated to all other pixels corresponding to, that is, picturing or representing the same part of the object. A part of the object in this sense is a detail or an element of the object that is captured in only a single pixel in each image picturing this specific part.

    [0009] A pixel stack therefore comprises all pixels of all raw images corresponding to a specific part of the object. This same part of the object can be pictured in multiple raw images because the raw images or their contents at least partly overlap each other. Every raw image may have the same number of pixels, whereas the pixel stacks may have different numbers of pixels, that is, the pixel stacks may have different heights.

    [0010] Once the raw images are aligned, at least one mask image is generated by means of a maximum intensity projection (MIP) through the pixel stacks of the aligned raw images. This means that from for each pixel stack the respective one pixel with the highest intensity value is determined. These maximum intensity values or pixels from all pixel stacks are then put together to create the at least one mask image. If only a single mask image is created or generated using the maximum intensity pixels from all pixel stacks, then this single mask image covers an entire area pictured or captured during the single contrast run. In this case, the single mask image may be larger than a single raw image. It is, however, also possible to generate multiple mask images, wherein each mask image may for example be the same size as the raw images and may therefore be smaller than the total area or region imaged in the single contrast run. In case multiple mask images are generated, a respective subset of the pixel stacks may be used for each of the multiple mask images.

    [0011] Advantageously, the number of mask images may be equal to the number of raw images, and each mask image can correspond to one raw image, meaning that a respective mask image can picture or cover the same area of the object as the corresponding raw image.

    [0012] In a next step of the method according to the present invention the set of processed images is generated from the raw images using the at least one mask image. This step may comprise subtracting the at least one mask image from the raw images. If only one mask image is generated, this same mask image may be subtracted from each raw image. For this purpose the mask image and the respective raw image can be aligned so that each pixel of the raw image or the corresponding intensity value is subtracted from the corresponding pixel of the respective raw image or the corresponding intensity value. Corresponding pixels represent or picture the same part of the object. Since the mask image is assembled with pixel-level precision after the raw images have been aligned, all data needed for subtracting the at least one mask image from the raw images with pixel-level precision or accuracy is available. The same procedure can be used if more than one mask image is generated. If the number of mask images is equal to the number of raw images and each mask image pictures the same area as the respective corresponding raw image, subtracting the mask images from the corresponding raw images can be done with particularly little effort since the same coordinate system may be used for the mask images and the raw images, meaning that no alignment or registration is necessary.

    [0013] In terms of the present invention a raw image can be image data captured and output by a detector of the imaging device. The detector may be or comprise an x-ray detector. The detector may comprise a multiplicity of pixels which can correspond to the pixels of the raw images. A raw image may, however, have been reconstructed or generated from raw image data output by the detector or the imaging device. Therefore, one or more image processing steps may be executed using a conventional image processing algorithm prior to the maximum intensity projection. Similarly, one or more image processing steps may be applied to the raw images in addition to using the at least one mask image to generate the set of processed images. Using the at least one mask image to generate the set of processed images can in particular comprise subtracting the at least one mask image from the raw images.

    [0014] The raw images show all parts of the object made visible by the respective imaging technique used to capture or generate the raw images. This may for example include bones as well as different kinds of soft tissue, organ tissue, and/or connective tissue in addition to the contrast agent in its respective stage of dispersal. Since only a respective maximum intensity value for each part of the object or the region captured during the single contrast run is used for the at least one mask image and the contrast agent lowers the detected intensity, the mask image does not show the contrast agent. This is at least the case for every part of the object or the imaged area that has been imaged at least once during the single contrast run or contrast acquisition without the contrast agent flowing through this particular part. By using the mask image, that is, by combining the at least one mask image with the raw images, the contrast agent or the respective corresponding parts of the object flooded with the contrast agent can be emphasised and made visible more clearly in the processed images. At the same time, parts of the object not flooded with the contrast agent can be removed and are therefore not shown in the processed images. The proposed method can, in other words, be advantageously employed for techniques such as peripheral digital subtraction angiography.

    [0015] Since the at least one mask image can be a compound image put together using parts of different raw images, there is no need for a separate mask run to obtain a mask image prior to injecting the contrast agent. This can advantageously allow employing imaging technique with lowered exposure of the object due to the omitted mask run. Another advantage of the present invention is that by omitting the mask run overall less time is required for the imaging process, which again reducing strain on the object, which may for example be a patient. Additionally, the resulting processed images can be improved while at the same time lowering the requirements that need to be met by the imaging device and respective operating personnel for a successful execution of the imaging procedure or process and the proposed method. This is the case, because only the single contrast run and no separate mask run is necessary, which eliminates the need for precise repositioning of the imaging device for two separate runs. Additionally, since it is not required to use fixed and predetermined parts of each raw image to generate the at least one mask image, it becomes possible to use raw images acquired in a continuous fashion, such as for example as a video feed. This can advantageously also eliminate the need for a respective operating personnel to manually advance the imaging device during the contrast run, which can be challenging and may require significant experience and expertise. The acquisition of the raw images as well as the generation of the processed images can therefore be fully automated.

    [0016] The method of aligning the overlapping raw images with pixel-level precision and exploiting or using the resulting pixel stacks also enables the use of noise reducing image processing techniques to improve an image quality of the resulting processed images.

    [0017] In an advantageous development of the present invention at least one noise-reduced mask image is generated from the aligned raw images. This is done by discarding all pixels from the pixel stacks having intensity values lower than a predetermined threshold value. A weighted average of the intensity values of the remaining pixels of each pixel stack is then calculated to generate the at least one noise-reduced mask image. The at least one noise-reduced mask image is then subtracted from each corresponding raw image to generate the set of processed images.

    [0018] The methods and possibilities as well as the corresponding advantages as described with reference to the at least one mask image can also be applied to the at least one noise-reduced mask image. This refers to generating the at least one mask image as well as to subtracting it from the raw images.

    [0019] The predetermined threshold value can be chosen or provided in dependence of the used contrast agent and/or an expected intensity for pixels imaging parts of the object that are flooded with the contrast agent. The predetermined threshold value may therefore also depend on an imaging technique used to acquire the raw images and/or on a relative location or position of a region to be flooded with the contrast agent within the object. If, for example, this flooded region is located deep within the object and surrounded by dense material such as bone, the expected intensity and therefore the predetermined threshold value can be relatively lower. In case of a thinner object not containing any dense material the expected intensity and therefore the predetermined threshold value may be relatively higher.

    [0020] Discarding all pixels with intensity values below the predetermined threshold value can advantageously ensure that parts of the raw images picturing or corresponding to parts of the object flooded with the contrast agent are not used to generate the noise-reduced mask image or images. The predetermined threshold value may, for example, be given as an absolute intensity value or as a relative intensity value with respect to a characteristic of the imaging device, such as an intensity output by a radiation source of the imaging device and/or a sensitivity of the detector of the imaging device.

    [0021] Since the respective weighted average intensity is used for the pixels of the noise-reduced mask image, noise, detection errors and/or measuring errors, et cetera are averaged out or minimised resulting in an improved image quality of the at least one noise-reduced mask image - and ultimately in an improved image quality of the processed images. In case the at least one noise-reduced mask image is generated, only this noise-reduced mask image and not the at least one mask image may be subtracted from the raw images to generate the set of processed images.

    [0022] In a further advantageous development of the present invention the at least one mask image is used as a reference for generating the at least one noise-reduced mask image. Since the at least one mask image features a maximum measured intensity for each of its pixels, the predetermined threshold value can be given as a relative value in relation to the corresponding intensity of the at least one mask image. This approach is especially advantageous since external factors such as the intensity output by a radiation source, or a characteristic of the detector, as well as a geometry of the imaged object do not have to be explicitly taken into account or considered in determining the threshold value.

    [0023] For calculating the weighted average of all intensity values of the remaining pixels a monotonous weight function can be provided and applied to the intensity values of the remaining pixels for each pixel stack. The weight function can be constant, meaning that all intensity values in each respective pixel stack are weighted equally. This can result in optimal noise reduction. Alternatively, a non-constant weight function can be used. For example, a step function can be used as the weight function to combine this step with the above-mentioned thresholding. The step function can apply equal non-zero weights to all intensity values above the predetermined threshold value and a weight of zero to all intensity values equal to or lower than the threshold value.

    [0024] In another example the weights assigned to the different intensity values can decrease or become smaller with an increasing difference between the respective intensity value and the corresponding maximum or reference value taken from the corresponding at least one mask image. Advantageously, this approach can take into account the possibility of low concentrations of the contrast agent being present in a part of the object corresponding to one of the remaining pixels. Using a monotonically decreasing weight function can therefore lower an impact or influence of intensity values of remaining pixels corresponding to parts of the object where some contrast agent is present. If, for example, the predetermined threshold value is set to low, the weight function can be used to mitigate a resulting effect on the image quality of the processed images.

    [0025] In an advantageous development of the present invention a noise-reducing filter is applied to the raw images before generating the set of processed images. The noise-reducing filter can, in other words, be used to prepare or pre-process the raw images before the at least one mask image or the at least one noise-reduced mask image is subtracted. This means that applying the noise-reduced filter results in pre-processed or modified raw images and the respective mask image is subtracted not from the original raw images but from the pre-processed or modified raw images. This approach can advantageously improve the image quality or clarity of the resulting processed images. If, for example, a thresholding technique is used and the intensities of different raw images are smoothed or averaged, the resulting averaged intensity can more closely match the intensity values of the respective mask image, resulting in clearer, more easily readable processed images with less artefacts or features not corresponding to the contrast agent. It can also be possible to use a bilateral filter or a joint bilateral filter, which can reduce noise while preserving detail. It can also be possible to use guided spatial smoothing to de-noise the images while keeping edges sharp and therefore preventing the raw images from becoming less sharp because of the filtering or denoising.

    [0026] In an advantageous development of the present invention after the at least one mask image - or the at least one noise-reduced mask image - is used, a noise-reducing filter is applied to generate the set of processed images. This noise-reducing filter can be the same and/or use the same techniques as described with respect to the approach of applying a noise-reducing filter to the raw images before generating the set of processed images. The noise-reducing filter can, in other words, be applied after the respective mask image has been subtracted from the raw image as an additional image processing step to generate the processed images. This can advantageously also result in an improved image quality of the processed images. This in turn can lead to easier readability and/or more accurate understanding and diagnosis.

    [0027] In an advantageous development of the present invention a detector used to capture the raw images or corresponding image data is moved continuously during the single contrast run to obtain the raw images. The raw images or the corresponding image data can, in other words, be acquired continuously. At the same time there may be a relative movement or motion between the object and the detector. This means that the detector can be moved while imaged object remains stationary. It is also possible to continuously move the object instead of or in addition to moving the detector during the single contrast run to achieve the same result and advantages as only moving the detector. Employing a continuous relative motion between the detector and the object during the single contrast run can advantageously be implemented or realised with less effort and improved precision or control and reproducibility as compared to alternatingly moving the detector from one discrete imaging position to the next and stopping or halting it in between. Another advantage is that a usability of a corresponding imaging apparatus comprising the detector can be improved, resulting in a lowered chance for operating errors or handling errors.

    [0028] In an advantageous development of the present invention an absorbed dose per raw image is varied over the single contrast run in dependence of a respective number of times a respective part of the object is imaged during the single contrast run. The dose per image absorbed by the object or the dose or amount of radiation used to irradiate the object for each image can, in other words, be set in dependence of a height of the pixel stacks to which the respective image contributes. Even though, the pixel stacks are only generated or created after the contrast run is complete by aligning the raw images, the height of the pixel stacks for each part of the object, that is, the number of times each part of the object is imaged during the single contrast run, can be known in advance, that is, before the single contrast run is started. For this purpose, the total number of raw images to be taken as well as a relative motion between the imaging device and the object as well as a timing or temporal correlation between the relative motion and the respective recording time of each raw image can be predetermined. Using this approach it is, for example, possible to lower the dose of radiation for raw images that are completely overlapped by multiple other raw images and therefore picture parts of the object that are imaged multiple times during the single contrast run.

    [0029] Ordinarily, using a lower dose could result in a reduced image quality. With the present invention, however, the image quality can be maintained despite the lowered dose by using multiple pixels of each pixel stack to reduce noise in the corresponding mask image or images and/or the respective raw image and/or the respective corresponding processed image. Specifically, at least first and last raw images taken during the single contrast run can be taken using a higher radiation dose then is used for taking a spatially and/or temporally more central raw image in a middle part of the single contrast run. Overall, this approach can result in advantageously reduced exposure of the object while maintaining image quality.

    [0030] In an advantageous development of the present invention a value of at least one acquisition parameter used for obtaining the raw images by means of the single contrast run is changed during the single contrast run. The respective raw images are then grouped corresponding to the different values of the acquisition parameter used during the single contrast run. Each group of the raw images is then processed separately. Since setting the at least one acquisition parameter to a different value can lead to corresponding differences in the raw images, grouping the raw images and treating and processing the groups separately in dependence of the respectively used value of the at least one acquisition parameter can advantageously result in an improved, more consistent image quality of the processed images. Since this approach allows for at least maintaining the image quality, the presently described method advantageously becomes more flexible, because the at least one acquisition parameter can be adjusted according to specific needs, a specific use case or application, and/or specific characteristics of the imaged object.

    [0031] The at least one acquisition parameter can for example be an aperture of an optical element in a path of the radiation used to acquire or generate the raw images. This can, for example, allow for effective and efficient imaging of an irregularly shaped object with varying widths or diameter. Processing the groups of raw images separately can mean that separate pixel stacks are formed and evaluated for each different group, separate MIPs are carried out, and at least one mask image and/or at least one noise-reduced mask image is separately generated for each individual group using only the raw images of that specific group. The mask images are then only used for the corresponding group of raw images. It is also possible to apply different and/or differently tuned or adjusted image processing steps to the raw images of the different groups to generate the processed images. It is for example possible, to adjust a parameter of an image processing algorithm used to generate the processed images in dependence of a value of the at least one acquisition parameter used in acquiring the respective raw image.

    [0032] Advantageously, a boundary between two groups can correspond to a dispersal front of the contrast agent, ensuring that each group of raw images contains pixels corresponding to parts of the object not flooded with the contrast agent so that at least one mask image can be generated for each group.

    [0033] In an advantageous development of the present invention a registration algorithm is used to align the raw images. Using the registration algorithm it can, for example, be possible to analyse the raw images to detect one or more features that are present in multiple raw images, and to then align these multiple raw images using the one or more detected features as a reference. This can be especially advantageous if a position of a detector used to capture the raw images is unknown in relation or with reference to a position of the object, or if the position of the detector is not known with a precision or accuracy enabling an alignment of the raw images with pixel-level accuracy by itself. Even if the detector position is known, the registration algorithm may be used supportively to improve or ensure a precise alignment of the raw images. The registration algorithm can be used very effectively to achieve the pixel-level accuracy in the alignment of the raw images in peripheral digital subtraction angiography (DSA) applications, where only a one-dimensional relative movement, that is, a displacement or offset of different raw images in only one direction needs to be evaluated or determined.

    [0034] Another aspect of the present invention in addition to the described method is an apparatus for generating a set of processed images of an object from overlapping raw images, wherein the raw images are obtained by means of a single contrast run. Therein, the apparatus comprises an alignment unit configured to align the raw images so that pixels representing the same part of the picture object in different raw images are aligned forming a respective pixel stack for each pictured part of the object. The apparatus further comprises a mask generation unit configured to generate at least one mask image by means of a maximum intensity projection through the pixel stacks of the aligned raw images. The apparatus further comprises an image processing unit configured to generate the set of processed images from the raw images using the at least one mask image. The apparatus can, in other words, be configured to carry out at least one embodiment of the method according to the present invention. The apparatus may comprise a processing unit (CPU) and a data store accessible by the processing unit. The alignment unit, the mask generation unit, and/or the image processing unit may be part of the CPU.

    [0035] In an advantageous development of the apparatus according to the present invention the apparatus comprises an image acquisition unit or imaging device and a feeding mechanism. Therein, the feeding mechanism is configured to perform or effectuate a relative motion between the object and the imaging device to execute the single contrast run. The apparatus is further configured to obtain the raw images by means of the imaging device. The imaging device may comprise a radiation source and a corresponding radiation detector. The feeding mechanism may comprise a drive unit as well as electrical and/or mechanical components configured to effectuate the relative motion by moving the imaging device, in particular the detector, the object, and/or an object mount or object support.

    [0036] Another aspect of the present invention is a storage medium containing program code configured to execute at least one embodiment of the method according to the present invention when the program code is executed by a processing unit. This processing unit may be the CPU of the apparatus according to the present invention. The apparatus according to the present invention may comprise the storage medium according to the present invention.

    [0037] The embodiments and developments of the method according to the present invention as well as the corresponding advantages may be applied to the apparatus according to the present invention and vice versa.

    [0038] Further advantages, features, and details of the present invention derive from the following description of preferred embodiments of the present invention as well as from the drawings pertaining to the present invention. The features and feature combinations previously mentioned in the description as well as the features and feature combinations mentioned in the following description of the figures and/or shown in the figures alone can be employed not only in the respectively indicated combination but also in other combinations or taken alone without leaving the scope of the invention.

    [0039] In the drawings
    FIG 1
    schematically shows a flow chart of steps of a method for generating a set of processed images;
    FIG 2
    schematically illustrates a stack of aligned raw images;
    FIG 3
    schematically illustrates one of the raw images from the stack shown in FIG 2;
    FIG 4
    schematically illustrates a mask image generated from the stack of raw images shown in FIG 2;
    FIG 5
    schematically illustrates four raw images taken during a single contrast run;
    FIG 6
    schematically illustrates a mask images generated from the raw images shown in FIG 5; and
    FIG 7
    schematically illustrates a set of four process images generated from the raw images shown in FIG 5 using the mask image shown in FIG 6.


    [0040] In the FIGs elements that provide the same function are marked with identical reference signs.

    [0041] FIG 1 schematically shows a flowchart 1 of steps of a method for generating a set of processed images. The method starts with an acquisition 2 of image data, specifically of multiple overlapping raw images. The acquisition 2 may comprise accessing a data store containing the raw images. It may, however, also comprise the actual taking or gathering of the image data by means of an imaging device or image acquisition device such as an x-ray apparatus. In particular, a single contrast run of a peripheral angiography procedure may be executed to obtain the image data, which is then used to generate or reconstruct the multiple raw images.

    [0042] The acquisition 2 is followed by an alignment step 3. In the alignment step 3 the raw images are aligned so that pixels of the raw images representing the same part of a pictured object in different raw images are aligned. Pixels of different raw images representing the same part of the object then form a respective pixel stack. For this purpose the multiple raw images may be arranged in an image stack.

    [0043] The alignment step 3 is followed by a maximum intensity projection 4 (MIP). The MIP 4 projects a highest intensity value of each of the pixel stacks on to one plane or level, thereby generating a first or initial mask image. The mask image contains the respective pixel with the maximum measured intensity value from all pixel stacks. Since the multiple raw images are taken at different points in time and may also be taken at different spatial locations they picture different stages of a dispersal of a contrast agent injected into the pictured object. Building up the mask image from only the respective highest intensities for each part of the pictured object or area results in the mask image not showing the contrast agent.

    [0044] The MIP 4 is followed by a process step 5, wherein a noise-reduced mask image is generated. In the process step 5 all pixels from the pixel stacks having intensity values lower than a predetermined threshold value are discarded, since they are interpreted to correspond to object parts flooded with the contrast agent. A weighted average of the remaining pixels is then calculated for each pixel stack. These weighted average intensity values make up the noise-reduced mask image.

    [0045] The process step 5 is followed by a subtraction 6 of the noise-reduced mask image from the raw images. The subtraction 6 removes image data that does not correspond to the contrast agent or parts of the object flooded with the contrast agent. The subtraction 6 therefore emphasises the spreading or dispersing contrast agent and can make the parts of the object that are flooded with the contrast agent more clearly visible. Presently, the subtraction 6 results in a stack of intermediary processing results, that is, intermediate images generated by subtracting the noise-reduced mask from the raw images.

    [0046] In a post-processing step 7 noise in the stack of intermediate images is reduced by applying one or more image processing filters to the stack of intermediate images in a spatial and/or temporal dimension of the stack. Instead of or in addition to the post-processing step 7 such noise reducing filters can be applied to the stack of raw images prior to the subtraction 6.

    [0047] These process steps from the acquisition 2 to the post-processing step 7 generate as a result a set of processed images with the parts of the object flooded with the contrast agent clearly visible at different stages of disposal corresponding to different points in time. The described method does not require a separate mask run to acquire a full set of mask images of the object prior to injection of the contrast agent. The described method can accommodate, that is, can be used with different image acquisition schemes or modes, wherein the imaging device or a corresponding detector can, for example, remain stationary, be moved stepwise between the raw images, or be moved continuously along the object. Exploitation of the fact that parts of the object are imaged multiple times during the single contrast run advantageously allows or a combination of the described method with multiple different processes or methods for noise reduction.

    [0048] FIG 2 schematically illustrates a stack 8 of aligned raw images 9. At a top of the stack 8 is a first raw image 10 and at a bottom of the stack 8 is a last raw image 12. In between the first raw image 10 and the last raw image 12 may be any number of additional raw images 9. Of these, only a third raw image 11 is specifically indicated in FIG 2. In this case, the raw images 9 partly overlap each other giving the stack 8 a spatial structure in at least one spatial direction 13 when the raw images 9 are aligned. Since the different raw images 9 have been taken or captured at different points in time as a time series the stack 8 also has a temporal structure as indicated by arrow 14. This means that the first image 10 shows a different part of a pictured object at an earlier stage than the third raw image 11 as well as the last raw image 12. The raw images 9 are presently aligned so that pixels of different raw images 9 showing the same part of the object form a respective pixel stack at a specific spatial location. As an example, a first pixel stack 15 and a second pixel stack 16 are indicated. The first pixel stack 15 extends from the first raw image 10 through to the third raw image 11 and thus contains three pixels of different raw images 9. All pixels of the first pixel stack 15 show or represent the same part of the object. Due to the raw images 9 only partly overlapping the part of the object corresponding to the first pixel stack 15 has, in this case, only been imaged three times.

    [0049] The second pixel stack 16, on the other hand, extends all the way through the stack 8 from the first raw image 10 to the last raw image 12. The second pixel stack 16 therefore has a larger height then the first pixel stack 15, meaning that it contains more pixels. Nevertheless, all pixels of the second pixel stack 16 show or represent the same part of the object because of the alignment of the raw images 9.

    [0050] FIG 3 schematically illustrates a raw image 17, which may be one of the raw images 9. Presently, the raw image 17 shows an object 18, which may for example be part of a patient. In this case, the object 18 is made up of a part of a left leg 19 containing a first vascular system 20, and a part of a right leg 21 containing a second vascular system 22. The raw image 17 is divided into a contrasted image part 23 and a non-contrasted image part 24 long a line 25. In this case, a contrast agent injected into the object 18 and dispersing through the vascular systems 20, 22 has reached the line 25 only in the second vascular system 22. As soon as the dispersing contrast agent-front reaches the line 25 in one and/or both of the legs 19, 21 an imaging device used to take the raw images 9 including the raw image 17 can be moved into a next position in the spatial direction 13, where a next one of the raw images 9 can then be taken.

    [0051] After all raw images 9 of the stack 8 have been taken or acquired, the MIP 4 through the stack 8 along the pixel stacks 15, 16, that is, along the temporal dimension indicated by arrow 14 is performed to generate a mask image 26, which is schematically illustrated in FIG 4. Since a spatial extent of the stack 8 in the spatial direction 13 is larger than a respective extent of each individual raw image 9 in the spatial direction 13, the mask image 26 may have a larger extent in the spatial direction 13 then each of the individual raw images 9. Specifically, the extent of the mask image 26 in the special direction 13 may be equal to the extent of the stack 8 in the same special direction. The mask image 26 also shows the object 18. It does, however, not show the contrast agent flooding the vascular systems 20, 22. For the mask image 26 to be generated without the contrast agent being visible, it is a prerequisite that during the single contrast run every part of the object 18 pictured in the mask image 26 is imaged at least once without contrast agent being present in the respective part.

    [0052] FIG 5 schematically illustrates a set 27 of raw images 28 of a second object 29. The raw images 28 have been taken at different points in time similar to the raw images 9. In contrast to the raw images 9, however, the raw images 28 completely overlap each other, meaning that they have been taken without spatial displacement, that is, without a relative motion between the second object 29 and an imaging device used to take or capture them.

    [0053] FIG 6 schematically illustrates a second mask image 30 of and for the second object 29.

    [0054] FIG 7 schematically illustrates a set 31 of processed images 32 of the second object 29. The processed images 32 have been generated by subtracting the second mask image 30 from the raw images 28. As compared to the raw images 28, the processed images 32 clearly show different progressive stages of dispersal of a contrast agent 33 through the second object 29.

    [0055] Summing up, the examples described herein illustrate an improved method for processing raw images to generate processed images relying only on image data from a single contrast run. The method comprises a mechanism for robust mask image generation.


    Claims

    1. Method (1) for generating a set (31) of processed images (32) of an object (18, 29) from overlapping raw images (9, 28), wherein the raw images (9, 28) are obtained by means of a single contrast run without a separate mask run,

    - the raw images (9, 28) are aligned so that pixels representing the same part of the pictured object (18, 29) in different raw images (9, 28) are aligned forming a respective pixel stack (15, 16) for each pictured part of the object (18, 29), wherein

    - at least one mask image (26, 30) is generated by means of a maximum intensity projection (4) through the pixel stacks (15, 16) of the aligned raw images (9, 28), so that the at least one mask image (26, 30) does not show a contrast agent (33) used in the contrast run,

    - the set (31) of processed images (32) is generated from the raw images (9, 28) using the at least one mask image (26, 30) by subtracting corresponding pixels or pixel intensity values from the raw images (9, 28).


     
    2. Method (1) in accordance with claim 1, characterised in that

    - at least one noise-reduced mask image (26, 30) is generated (5) from the aligned raw images (9, 28) by

    - discarding all pixels from the pixel stacks (15, 16) having intensity values lower than a predetermined threshold value,

    - calculating a weighted average of the intensity values of the remaining pixels of each pixel stack (15, 16) to generate (5) the at least one noise-reduced mask image (26, 30), and

    - the at least one noise-reduced mask image (26, 30) is subtracted (6) from each corresponding raw image (9, 28) to generate the set (31) of processed images (32).


     
    3. Method (1) in accordance with claim 2, characterised in that the at least one mask image (26, 30) is used as a reference for generating the at least on noise-reduced mask image (26, 30).
     
    4. Method (1) in accordance with any of the preceding claims, characterised in that a noise-reducing filter is applied to the raw images (9, 28) before generating the set (31) of processed images (32).
     
    5. Method (1) in accordance with any of the preceding claims, characterised in that after the at least one mask image (26, 30) is used a noise-reducing filter is applied to generate the set (31) of processed images (32).
     
    6. Method (1) in accordance with any of the preceding claims, characterised in that a detector used to capture the raw images (9, 28) is moved continuously during the single contrast run to obtain the raw images (9, 28).
     
    7. Method (1) in accordance with any of the preceding claims, characterised in that a radiation dose per raw image (9, 28) output by an imaging device and absorbed by the object (18, 29) is varied over the single contrast run in dependence of a respective number of times a respective part of the object (18, 29) is imaged during the single contrast run, so that the dose of radiation per raw images (9, 28) is lower for raw images (9, 28) that are completely overlapped by multiple other raw images (9, 28).
     
    8. Method (1) in accordance with any of the preceding claims, characterised in that

    - a value of at least one acquisition parameter used for obtaining the raw images (9, 28) by means of the single contrast run is changed during the single contrast run, and

    - the respective raw images (9, 28) are grouped corresponding to the different values of the acquisition parameter, and

    - each group of raw images (9, 28) is processed separately.


     
    9. Method (1) in accordance with any of the preceding claims, characterised in that a registration algorithm is used to align the raw images (9, 28).
     
    10. Apparatus for generating a set (31) of processed images (32) of an object (18, 29) from overlapping raw images (9, 28), wherein the raw images (9, 28) are obtained by means of a single contrast run without a separate mask run, and wherein the apparatus comprises

    - an alignment unit configured to align (3) the raw images (9, 28) so that pixels representing the same part of the pictured object (18, 29) in different raw images (9, 28) are aligned forming a respective pixel stack (15, 16) for each pictured part of the object (18, 29),

    - a mask generation unit configured to generate at least one mask image (26, 30) by means of a maximum intensity projection (4) through the pixel stacks (15, 16) of the aligned raw images(9, 28) so that the at least one mask image (26, 30) does not show a contrast agent (33) used in the contrast run, and

    - an image processing unit configured to generate the set (31) of processed images (32) from the raw images (9, 28) using the at least one mask image (26, 30) by subtracting corresponding pixels or pixel intensity values from the raw images (9, 28).


     
    11. Apparatus in accordance with claim 10, characterised in that the apparatus comprises an imaging device and a feeding mechanism, wherein

    - the feeding mechanism is configured to perform a relative movement between the object (18, 29) and the imaging device to execute the single contrast run, and

    - the apparatus is configured to obtain the raw images (9, 28) by means of the imaging device.


     
    12. Storage medium containing program code configured to execute a method according to one of the claims 1 to 9 when the program code is executed by a processing unit.
     


    Ansprüche

    1. Verfahren (1) zum Erzeugen eines Satzes (31) von verarbeiteten Bildern (32) eines Objekts (18, 29) aus überlagerten Rohbildern (9, 28), wobei die Rohbilder (9, 28) mittels eines einzelnen Füllungslaufs ohne einen separaten Maskenlauf erhalten werden,

    - wobei die Rohbilder (9, 28) derart ausgerichtet werden, dass Pixel, die denselben Teil des abgebildeten Objekts (18, 29) in unterschiedlichen Rohbildern (9, 28) darstellen, derart ausgerichtet sind, dass sie einen jeweiligen Pixelstapel (15, 16) für jeden abgebildeten Teil des Objekts (18, 29) bilden, wobei

    - mindestens ein Maskenbild (26, 30) mittels einer Projektion (4) mit maximaler Intensität durch die Pixelstapel (15, 16) der ausgerichteten Rohbilder (9, 28) erzeugt wird, sodass das mindestens eine Maskenbild (26, 30) ein in dem Füllungslauf verwendetes Kontrastmittel (33) nicht zeigt,

    - der Satz (31) von verarbeiteten Bildern (32) aus den Rohbildern (9, 28) mithilfe des mindestens einen Maskenbilds (26, 30) durch Subtrahieren entsprechender Pixel oder Pixelintensitätswerte von den Rohbildern (9, 28) erzeugt wird.


     
    2. Verfahren (1) nach Anspruch 1, dadurch gekennzeichnet, dass

    - mindestens ein rauschreduziertes Maskenbild (26, 30) aus den ausgerichteten Rohbildern (9, 28) erzeugt wird (5) durch

    - Verwerfen aller Pixel von den Pixelstapeln (15, 16) mit Intensitätswerten kleiner als ein vorbestimmter Schwellenwert,

    - Berechnen eines gewichteten Durchschnitts der Intensitätswerte der verbleibenden Pixel jedes Pixelstapels (15, 16), um das mindestens eine rauschreduzierte Maskenbild (26, 30) zu erzeugen (5), und

    - das mindestens eine rauschreduzierte Maskenbild (26, 30) von jedem entsprechenden Rohbild (9, 28) subtrahiert wird (6), um den Satz (31) von verarbeiteten Bildern (32) zu erzeugen.


     
    3. Verfahren (1) nach Anspruch 2, dadurch gekennzeichnet, dass das mindestens eine Maskenbild (26, 30) als eine Referenz zum Erzeugen des mindestens einen rauschreduzierten Maskenbilds (26, 30) verwendet wird.
     
    4. Verfahren (1) nach einem der vorhergehenden Ansprüche, dadurch gekennzeichnet, dass ein Rauschreduzierungsfilter auf die Rohbilder (9, 28) vor Erzeugen des Satzes (31) von verarbeiteten Bildern (32) angewendet wird.
     
    5. Verfahren (1) nach einem der vorhergehenden Ansprüche, dadurch gekennzeichnet, dass, nachdem das mindestens eine Maskenbild (26, 30) verwendet wird, ein Rauschreduzierungsfilter zum Erzeugen des Satzes (31) von verarbeiteten Bildern (32) angewendet wird.
     
    6. Verfahren (1) nach einem der vorhergehenden Ansprüche, dadurch gekennzeichnet, dass ein zum Erfassen der Rohbilder (9, 28) verwendeter Detektor kontinuierlich während des einzelnen Füllungslaufs bewegt wird, um die Rohbilder (9, 28) zu erhalten.
     
    7. Verfahren (1) nach einem der vorhergehenden Ansprüche, dadurch gekennzeichnet, dass eine Strahlungsdosis pro Rohbild (9, 28), die durch eine Bildgebungsvorrichtung ausgegeben und durch das Objekt (18, 29) absorbiert wird, über den einzelnen Füllungslauf in Abhängigkeit von einer jeweiligen Anzahl von Malen, die ein jeweiliger Teil des Objekts (18, 29) während des einzelnen Füllungslaufs aufgenommen wird, variiert wird, sodass die Dosis an Strahlung pro Rohbilder (9, 28) bei Rohbildern (9, 28) niedriger ist, die durch mehrere andere Rohbilder (9, 28) vollständig überlagert sind.
     
    8. Verfahren (1) nach einem der vorhergehenden Ansprüche, dadurch gekennzeichnet, dass

    - ein Wert mindestens eines Erfassungsparameters, der zum Erhalten der Rohbilder (9, 28) mittels des einzelnen Füllungslaufs verwendet wird, während des einzelnen Füllungslaufs verändert wird, und

    - die jeweiligen Rohbilder (9, 28) entsprechend den unterschiedlichen Werten des Erfassungsparameters gruppiert werden, und

    - jede Gruppe von Rohbildern (9, 28) separat verarbeitet wird.


     
    9. Verfahren (1) nach einem der vorhergehenden Ansprüche, dadurch gekennzeichnet, dass ein Registrierungsalgorithmus verwendet wird, um die Rohbilder (9, 28) auszurichten.
     
    10. Einrichtung zum Erzeugen eines Satzes (31) von verarbeiteten Bildern (32) eines Objekts (18, 29) aus überlagerten Rohbildern (9, 28), wobei die Rohbilder (9, 28) mittels eines einzelnen Füllungslaufs ohne einen separaten Maskenlauf erhalten werden, und wobei die Einrichtung umfasst

    - eine Ausrichtungseinheit, die ausgestaltet ist, die Rohbilder (9, 28) derart auszurichten (3), dass Pixel, die denselben Teil des abgebildeten Objekts (18, 29) in unterschiedlichen Rohbildern (9, 28) darstellen, derart ausgerichtet sind, dass sie einen jeweiligen Pixelstapel (15, 16) für jeden abgebildeten Teil des Objekts (18, 29) bilden,

    - eine Maskenerzeugungseinheit, die ausgestaltet ist, mindestens ein Maskenbild (26, 30) mittels einer Projektion (4) mit maximaler Intensität durch die Pixelstapel (15, 16) der ausgerichteten Rohbilder (9, 28) zu erzeugen, sodass das mindestens eine Maskenbild (26, 30) ein in dem Füllungslauf verwendetes Kontrastmittel (33) nicht zeigt, und

    - eine Bildverarbeitungseinheit, die ausgestaltet ist, den Satz (31) von verarbeiteten Bildern (32) aus den Rohbildern (9, 28) mithilfe des mindestens einen Maskenbilds (26, 30) durch Subtrahieren entsprechender Pixel oder Pixelintensitätswerte von den Rohbildern (9, 28) zu erzeugen.


     
    11. Einrichtung nach Anspruch 10, dadurch gekennzeichnet, dass die Einrichtung eine Bildgebungsvorrichtung und einen Vorschubmechanismus umfasst, wobei

    - der Vorschubmechanismus ausgestaltet ist, eine relative Bewegung zwischen dem Objekt (18, 29) und der Bildgebungsvorrichtung durchzuführen, um den einzelnen Füllungslauf durchzuführen, und

    - die Einrichtung ausgestaltet ist, die Rohbilder (9, 28) mittels der Bildgebungsvorrichtung zu erhalten.


     
    12. Speichermedium, das Programmcode enthält, der ausgestaltet ist, ein Verfahren nach einem der Ansprüche 1 bis 9 auszuführen, wenn der Programmcode durch eine Verarbeitungseinheit ausgeführt wird.
     


    Revendications

    1. Procédé (1) de production d'un jeu (31) d'images (32) traitées d'un objet (18, 29) à partir d'images (9, 28) brutes se chevauchant, les images (9, 28) brutes étant obtenues au moyen d'une opération unique avec contraste, sans une opération distincte avec masque,

    - on aligne les images (9, 28) brutes de manière à ce que des pixels, représentant la même partie de l'objet (28, 29) dépeint dans des images (18, 29) brutes différentes, soient alignés en formant une pile (15, 16) respective de pixels pour chaque partie dépeinte de l'objet (18, 29),
    dans lequel

    - on produit au moins une image (26, 30) de masque au moyen d'une projection (4) d'intensité maximum dans les piles (15, 16) de pixels des images (9, 28) brutes alignées, de manière à ce que la au moins une image (26, 30) de masque ne montre pas un agent (33) de contraste utilisé dans l'opération avec contraste,

    - on produit le jeu (31) d'images (32) traitées à partir des images (19, 28) brutes, en utilisant la au moins une image (26, 30) de masque, en soustrayant des pixels correspondants ou des valeurs d'intensité de pixel correspondantes des images (9, 28) brutes.


     
    2. Procédé (1) suivant la revendication 1, caractérisé en ce que

    - on produit (5) au moins une image (26, 30) de masque à bruit réduit à partir des images (9, 28) brutes alignées en

    - écartant tous les pixels des piles (15, 16) de pixels ayant des valeurs d'intensité plus petites qu'une valeur de seuil déterminée à l'avance,

    - calculant une moyenne pondérée des valeurs d'intensité des pixels restant de chaque pile (15, 16) de pixels pour produire (5) la au moins une image (26, 30) de masque à bruit réduit, et

    - on soustrait la au moins une image (26, 30) de masque à bruit réduit de chaque image (9, 28) brute correspondante pour produire le jeu d'images (32) traitées.


     
    3. Procédé (1) suivant la revendication 2, caractérisé en ce que l'on utilise la au moins une image (26, 30) de masque comme référence pour produire la au moins une image (26, 30) de masque à bruit réduit.
     
    4. Procédé (1) suivant l'une quelconque des revendications précédentes, caractérisé en ce que l'on applique un filtre de réduction du bruit aux images (9, 28) brutes avant de produire le jeu (31) d'images (32) traitées.
     
    5. Procédé (1) suivant l'une quelconque des revendications précédentes, caractérisé en ce qu'après avoir utilisé la au moins une image (26, 30) de masque, on applique un filtre de réduction du bruit pour produire le jeu (31) d'images (32) traitées.
     
    6. Procédé (1) suivant l'une quelconque des revendications précédentes, caractérisé en ce que l'on déplace un détecteur utilisé pour acquérir les images (9, 28) brutes continuellement pendant l'opération unique avec contraste, afin d'obtenir les images (9, 28) brutes.
     
    7. Procédé (1) suivant l'une quelconque des revendications précédentes, caractérisé en ce que l'on fait varier une dose de rayonnement par image (9, 28) brute sortie par un dispositif d'imagerie et absorbée par l'objet (18, 29) sur l'opération unique avec contraste en fonction d'un nombre respectif de fois où une partie respective de l'objet (18, 29) est prise en image pendant l'opération unique avec contraste, de manière à ce que la dose de rayonnement par image (9, 28) brute soit plus petite pour des images (9, 28) brutes, qui sont chevauchées complètement par de multiples autres images (9, 28) brutes.
     
    8. Procédé (1) suivant l'une quelconque des revendications précédentes, caractérisé en ce que

    - on change, pendant l'opération unique avec contraste, une valeur d'au moins un paramètre d'acquisition utilisé pour obtenir les images (9, 28) brutes au moyen de l'opération unique avec contraste, et

    - on regroupe les images (9, 28) brutes respectives correspondant aux différentes valeurs du paramètre d'acquisition, et

    - on traite séparément chaque groupe d'images (9, 28) brutes.


     
    9. Procédé (1) suivant l'une quelconque des revendications précédentes, caractérisé en ce que l'on utilise un algorithme d'enregistrement pour aligner les images (9, 28) brutes.
     
    10. Installation de production d'un jeu (31) d'images (32) traitées d'un objet (18, 29) à partir d'images (9, 28) brutes se chevauchant, les images (9, 28) brutes étant obtenues au moyen d'une opération unique avec contraste sans une opération distincte avec masque, l'installation comprenant

    - une unité d'alignement configurée pour aligner (3) les images (9, 28) brutes, de manière à ce que des pixels, représentant la même partie de l'objet (18, 29) dépeint dans des images (9, 28) brutes différentes, soient alignés en formant une pile (15, 16) respective de pixels pour chaque partie dépeinte de l'objet (18, 29),

    - une unité de production de masque configurée pour produire au moins une image (26, 30) de masque au moyen d'une projection (4) d'intensité maximum dans les piles (15, 16) de pixels des images (9, 28) brutes alignées, de manière à ce que la au moins une image (26, 30) de masque ne montre pas un agent (33) de contraste utilisé dans l'opération avec contraste, et

    - une unité de traitement d'image configurée pour produire le jeu (31) d'images (32) traitées à partir des images (9, 28) brutes, en utilisant la au moins une image (26, 30) de masque, en soustrayant des pixels correspondants ou des valeurs d'intensité de pixel correspondantes des images (9, 28) brutes.


     
    11. Installation suivant la revendication 10, caractérisée en ce que l'installation comprend un dispositif d'imagerie et un mécanisme de chargement, dans lequel

    - le mécanisme de chargement est configuré pour effectuer un déplacement relatif entre l'objet (18, 29) et le dispositif d'imagerie, afin d'exécuter l'opération unique avec contraste, et

    - l'installation est configurée pour obtenir les images (9, 28) brutes au moyen du dispositif d'imagerie.


     
    12. Support de mémoire contenant un code de programme configuré pour exécuter un procédé suivant l'une des revendications 1 à 9, lorsque le code de programme est exécuté par une unité de traitement.
     




    Drawing














    Cited references

    REFERENCES CITED IN THE DESCRIPTION



    This list of references cited by the applicant is for the reader's convenience only. It does not form part of the European patent document. Even though great care has been taken in compiling the references, errors or omissions cannot be excluded and the EPO disclaims all liability in this regard.

    Patent documents cited in the description