FR2861571A1 - Medical imaging device for reconstructing tomography images, has collimator positioned between X-ray source and examining zone for surrounding X-ray beam and delimiting X-ray beam coming from radiation sensor - Google Patents

Abstract

The device has a radiation sensor (300) that is moved around a zone by an acquisition unit (100) to carry out a series of 2D exposures in the zone. A collimator (310) is positioned between an X-ray source and the examining zone for surrounding the X-ray beam. The collimator delimits the X-ray beam coming from the sensor during the series of exposure. An independent claim is also included for a medical imaging process.

Description

2861571 1

  The invention relates to the reconstruction of 3D tomographic images from x-ray images obtained in 2D by X-rays. The use of X-ray 2D views is well known in interventional imaging, and in particular by the use of vascular observation devices with a C-shaped shaft. Such views are used to guide the surgical instruments (catheter, needle, etc.) and in particular to estimate the positions of such instruments with respect to the surrounding anatomical structures (vessels, bones, ... ).

  These 2D images, obtained by X-rays, are also used to reconstruct a 3D image of the observed object, ie in the context of tomographic reconstruction.

  Tomographic reconstruction requires long calculations, demanding in immobilized resources and in memory.

  This is why we sought, when the area of interest is only a fraction of the extent of the X-ray sensor, to reconstruct the 3D image on this fraction.

  Thus, from the acquisition of a series of views using a sensor of a given extent, it was proposed to exploit only a portion of these data corresponding to a sub-area of the sensor. Thus, known systems propose to the user, from a global reconstruction of the object but approximate (because low resolution), to identify the sub-volume on which it is desired to obtain a high resolution image. The system then exploits the data concerned to produce the specific reconstruction concerned.

  To provide the image of this sub-volume, these systems take into account all the acquisition data, ie the data of a larger volume than the one chosen.

  Indeed, these systems also use the acquisition data of this sub-volume.

  2861571 2 Thus, these systems use all the data acquired, ie a large volume of data. They do not respond to the need to decrease the amount of data processing.

  The object of the invention is to reduce the amount of computation required to perform a tomographic reconstruction, when it concerns a small volume.

  The invention also aims to reduce this amount of calculation while using a relatively low implementation cost.

  These objects are achieved according to the invention by means of a medical imaging device comprising an X-ray emission source and a sensitive surface radiation sensor, means for moving this sensor around an observation zone of a patient in order to perform a series of 2D shots of this area, further comprising means for processing the shooting data to achieve by this treatment a 3D tomographic reconstruction of the area to be observed, the device being characterized in that it further comprises means for selective lateral delimitation by the X-ray beam practitioner arriving on the sensitive surface of the sensor, these beam delimiting means retaining this delimitation during the series of shots.

  Other features, objects and advantages of the invention will appear on reading the detailed description which follows, with reference to the appended figures in which: FIG. 1 is a schematic 3D view of a device 25 of FIG. acquisition of tomographic reconstruction system images according to the invention; FIG. 2a illustrates the operation of a collimation contour detection device according to a variant of the invention in the case of a spherical collimator; FIG. 2b illustrates the operation of this device, in the case of a contour recognition on a rectangular contour.

  The exemplary implementation of the invention which will be described now is part of the use of a basic material of the known type, namely here a C-shaped shaft, referenced 100, forming means of acquiring a series of 2D shots.

  These shots are rotated around the area of the patient to be observed 200, and are then processed for tomographic reconstruction of a 3D volume and the display of the latter on a user interface. viewing of the practitioner.

  This known material is further equipped with means which will be described below and which allow the practitioner to select and view specifically a sub-part of an observable area.

  Thus, the data processing means are here provided for reconstructing a volume in 3D which may correspond both to the overall extent of the rotating sensor 300 and to the extent of a sub-volume corresponding to a particular selection of the practitioner. .

  However, it is proposed to delimit an area to be reconstructed, not from ensemble data previously taken and then restricted, but from a delimitation of the exposure of the sensor.

  Thus, the data processing means are supplied by data from shots which are in smaller quantities, and allow from the outset a higher processing efficiency.

  The means proposed here for limiting this volume of data consist of means 310 for limiting the X-ray beam captured by the sensor. This is a physical collimator whose role is to laterally delimit a beam 400 emitted and having to irradiate the patient, before this beam reaches the sensor.

  To simplify the illustration, the collimator 310 has been shown (FIG. 1) by its shadow on the detector because, as a variant and in a privileged manner, the collimator 310 is placed upstream of the patient, between the X-ray source and the detector. , rather than placed in direct recovery of the sensor.

  However, alternatively and in a preferred manner, the collimator 310 is placed upstream of the patient, between the X-ray source and the patient's area to be observed.

  By choosing this collimator 310, the practitioner specifically delimits the beam according to the volume that he wants to obtain in the end on its viewing interface.

  The collimator 310 has for example a circular shape, or substantially rectangular, for example mainly horizontal or vertical (Figures 2a and 2b). This collimator is preferably positioned in such a way that the X-ray beam passing through it is centered on the geometric axis 500 of rotation of the sensor. The axis of rotation 500 of the sensor then defines the central geometric axis of the volume 200 which will be finally reconstructed.

  The volume 200 which is specifically reconstructible turns out to be the one that is irradiated with each shot, ie the volume 200 defined as the largest volume whose projection on each successive sensor position falls within the collimator 310.

  In other words, the delineation of the beam 400 defines, in each shot, an observation field 320 (Field Of View) whose reconstruction leads to a volume of interest (Volume Of Interest). This volume 200 is the one whose substantially parallel projections of its borders on each 3D shooting corresponds to the limits of the collimator 310.

  For circular collimation (Figure 2a), the volume of interest (VOI) is a sphere. For rectangular collimation (Figure 2b) the volume of interest (VOI) is a cylinder.

  Advantageously positioning the plane of the sensor parallel to the axis of rotation 500 of the latter during successive shots. The sensor plane is also positioned in such a way that this axis of rotation 500 projects to the center of the sensor 300. It is also preferentially provided that the position of the collimator 310 is constant in the beam 400 during the rotation of the sensor. 300 sensor.

  In practice, positioning deflections of the sensor 300 may appear, the latter not being perfectly parallel to the geometrical axis of rotation 500 during the displacement of the C 100 shaft. The image processing means are advantageously equipped with means 2861571 for correcting these deviations, for example means for recalculating the data acquired in order to refocus these data with respect to the axis of rotation, means for reshaping these data in order to correct the faults of parallelism relative to the axis of rotation 500.

  For example, simple projective transformations are applied to the data taken for each of the sampling planes with respect to a mean axis of rotation.

  If, in an opposite but equally possible variant, the collimator 310 is expressly positioned asymmetrically with respect to the projection of the axis of rotation 500 on the sensor, then there is no volume whose projected contour corresponds to systematically at the edges of the collimator. The reconstruction volume is then defined by the collimator border that is closest to the projection of the axis. The rotation of this single contour around this axis 500 then delimits the volume to be reconstructed.

  In other words, it is the projection of the collimator 310 which is closest to the axis of rotation which restricts the observation field 320 (FOV), and which consequently delimits the volume of interest 200 (VOI). ).

  In other words, the reconstruction volume 200 can be defined in this case as that given by collimated projections which would have for side edge with respect to the axis of rotation 500, not only the most restrictive border, but also its border placed symmetrically with respect to the projection of the axis of rotation 500 on the collimator 310.

  The processing time and the memory required are reduced by simply processing the data relating to the volume of interest 200. In order to reduce the processing applied to the data actually inside the collimator 310, the processing means are here provided to identify the volume. of interest 200 before the treatment applied.

  Figures 2a and 2b illustrate the implementation proposed here in preference for such a detection. This variant aims to correct the fact that an image processing for the purpose of determining the outline of a collimator 310 from a shot does not give a satisfactory result, insofar as the limits of collimator 310 appear particularly low contrast, especially in relation to the density of the areas appearing within the inner image 320 to the collimator 310.

  Indeed, this inner image may contain shadows of bone or metal, comparable in contrast with the borders of the collimator 310.

  To accentuate the contours, it is proposed here as a preliminary step to the identification of the collimator 310, to add all the shots 600 taken during the rotation of the sensor 300.

  The location of the collimator 310 is fixed on each of the shots, so that the sum of 700 different shots reveals a contour 330 particularly contrasted with the sum of the internal images, which they tend to compensate.

  Thus, in FIG. 2a, the summed view 700 reveals a particularly contrasted circle contour 330.

  The contour identification means, here in the form of simplified pattern recognition means, then process this added image. The recognition is carried out here by scanning successive horizontal and vertical lines, each scan showing a contrast variation plot 800 whose peaks 810 correspond to the intersection of the scan line considered with the contour 330. All of these traces 800 of contrast variation gives the spatial location of each point of the contour 330, depending on the position of the scan line, and the peaks 810 revealed on this line.

  In Figure 2b, there is shown the particular case of a rectangular collimator. In this case, a row-by-row and column-by-column scan is not necessary. Yet again, only two scan lines of the image-sum 700 are needed to identify, by their peaks 810 of contrast variation, the location of the vertical edge and the horizontal edge.

  To further extract the fruit of the uniqueness of the four edges of such a collimator rectangle 310, we add all the rows of this view before proceeding to the scan, so that the 2861571 7 scan s' performs on a sum of rows or a sum of columns of an image which is already itself a sum of images.

  Such a contrast variation plot 800 is then even more significant since resulting from two distinct additions.

  Such a summation of rows or columns is a simple technique to implement to make the contours 330 more prominent, but other techniques can also be used to accentuate the contours, by using more generally the fact that an edge has an edge. constant localization in a given direction of the collimated image.

  Other collimator geometries 310 are also possible, for example an elongated rectangle shape whose extremal edges are rounded, made by mixing a circle and a rectangle.

  The identification of the contour 330 of the collimated image makes it possible, in addition to the spatial limitation of the reconstruction operation, to implement a pretreatment, which will be described hereinafter, aimed at limiting the appearance of artifacts at during the reconstruction.

  Indeed, in a manner known per se, a shooting at the strict edges gives, after reconstruction, the appearance of artifacts.

  This effect is known in the case of reconstructions from shots taken by the entire sensitive surface of a sensor, which itself has strict edges.

  It is proposed, in order to avoid these artifacts due to strict edges, to carry out a pretreatment of the shots 600 which takes advantage of the knowledge of the contour 330 of the collimated image 320. Knowing this contour 330, the processing means perform a extrapolating data beyond these collimation contours, to develop an image with a specifically arranged border to prevent the appearance of artifacts. A suitable technique for this is that of the contour mirror effect.

  The extrapolation is then followed by linear filtering, also of known type. Also, the limitation of the acquisition of useful data is here preferably followed by a technique of extrapolation and filtering of this limited image.

  Such a reconstruction algorithm using linear filtering has already been proposed in particular for tomographic reconstruction from data corresponding to the entire detection surface of the rotating sensor, the strict contour shots processed in this case being the shots corresponding to the entire sensitive surface of the sensor.

  The identification of the positioning of this physical limitation is necessary for these processing means to apply this extrapolation. As described above, this positioning is, in this variant, not provided directly by the user.

  Thus, the artifacts due to the application of a linear filter on a truncated profile are notably attenuated by the extrapolation performed at the collimation edges 340 which are directed orthogonal to the filtering direction. Such extrapolation pretreatment may, for example, be restricted to extrapolation in the direction perpendicular to the axis of rotation of the sensor.

  The extrapolation by mirror effect may, in other words, be restricted to the edges 340 disposed symmetrically to the projection of the axis of rotation 500, that is to say the edges substantially parallel to this projection.

  The edges perpendicular to the projection of the axis of rotation 500 do not necessarily have to be the subject of such pretreatments, because the presence of strict contours in these directions does not affect the reconstruction.

  Advantageously, the device includes automatic detection means 700, 800, 810 of the form of the delimiting means implemented.

  Advantageously, the device includes means for performing the tomographic reconstruction of a volume 200 which is defined as the largest volume inscribed in each of the delimited beams 400 corresponding to each of said successive shots 600.

Claims (1)

9 CLAIMS   A medical imaging device comprising an X-ray emission source (400) and a sensitive surface radiation sensor (300), means (100) for moving the sensor (300) around an area to be observed of a patient to perform a series of 2D shots of this area, further comprising means for processing the shooting data to achieve by this treatment a 3D tomographic reconstruction of the area to be observed, the device being characterized in that it further comprises means (310) for selective lateral delimitation by the X-ray beam practitioner arriving on the sensitive surface of the sensor (300), said beam delimiting means retaining this delimitation (320) at the during the series of shots.   2. medical imaging device according to claim 1 characterized in that the delimiting means (310) are constituted by at least one collimator (310) surrounding the X-ray beam (400).   3. Medical imaging device according to claim 2, characterized in that the device includes a series of collimators (310) having different shapes and arranged to be selectable by the user.   4. Medical imaging device according to claim 2 or claim 3, characterized in that the series of collimators (310) includes a collimator of rectangular shape.   5. Medical imaging device according to any one of claims 2 to 4, characterized in that the series of collimators (310) includes a circular shaped collimator.   6. medical imaging device according to any one of claims 2 to 5, characterized in that the collimator (310) is positioned between the X-ray source and the area (200) of the patient to be observed.   7. medical imaging device according to any one of the preceding claims, characterized in that the device includes automatic detection means (700, 800, 810) of the form of the delimiting means implemented.   8. medical imaging device according to the preceding claim, characterized in that it includes means capable of summing a group of 2D shots (700), and perform a contour identification (800, 810) on the image constituted by the sum of these shots.   9. medical imaging device according to any one of the preceding claims, characterized in that the device includes means for performing the tomographic reconstruction of a volume (200) which is defined as the largest volume in each of the beams delimited (400) corresponding to each of said successive shots (600).   10. medical imaging device according to any one of the preceding claims, characterized in that the device comprises means for performing on each shot an image extrapolation beyond the bounded contour of this shooting.   11. A medical imaging method using an X-ray emission source (400) and a sensitive surface radiation sensor (300), in which the sensor (300) is moved around an area to be observed. a patient in order to carry out a series of 2D shots of this area, and in which image data is processed to perform by this treatment a 3D tomographic reconstruction of the area to be observed, the method being characterized in that it further comprises a step of selective lateral delimitation by the practitioner of the X-ray beam arriving on the sensitive surface of the sensor (300), this beam delimitation being retained during the series of shots.