The invention relates to a method and a system for determining, during operation, the geometric characteristics of a three-dimensional (3D) image reconstruction system from images acquired with an imaging system. X-rays. The invention thus enables on-line calibration of an X-ray imaging system.

The geometric characteristics are estimated during the rotational acquisition of a C-arm type device.

It is common to use a mobile radiology system to perform surgical or interventional procedures. These systems, also called mobile C-arm (or block amplifier) ​​allow the surgeon to acquire X-ray images during the operation and to control in real time the positioning of the tools used by the surgeon (catheter, needle, prosthesis, etc. ...) in a minimally invasive manner. Most of these systems make it possible to obtain two-dimensional images with a video image stream at thirty images per second. The practitioner then uses these images to perform a mental reconstruction of the patient's morphology in order to position the tool precisely in the area to be operated on, in real time. More recently, more sophisticated systems have appeared and make it possible to acquire a 3D image of the tool used by a surgeon during an operation. The radiology system rotates around the patient in order to obtain a set of 2D two-dimensional images. These 2D images are then processed by a reconstruction algorithm making it possible to obtain a volume image in 3D. To reconstruct the image, the algorithm needs to know, for each 2D image, the exact geometry of the C-arm, namely the position of the detector and of the X-ray source, in relation to the patient. The systems currently in use propose to perform an offline calibration of the C-arm during the preventive maintenance phases of the device, every six months or one year. The radiology system rotates around the patient to obtain a set of 2D two-dimensional images. These 2D images are then processed by a reconstruction algorithm making it possible to obtain a volume image in 3D. To reconstruct the image, the algorithm needs to know, for each 2D image, the exact geometry of the C-arm, namely the position of the detector and of the X-ray source, in relation to the patient. The systems currently in use propose to perform an offline calibration of the C-arm during the preventive maintenance phases of the device, every six months or one year. The radiology system rotates around the patient to obtain a set of 2D two-dimensional images. These 2D images are then processed by a reconstruction algorithm making it possible to obtain a volume image in 3D. To reconstruct the image, the algorithm needs to know, for each 2D image, the exact geometry of the C-arm, namely the position of the detector and of the X-ray source, in relation to the patient. The systems currently in use propose to perform an offline calibration of the C-arm during the preventive maintenance phases of the device, every six months or one year. the exact geometry of the C-arm, i.e. the position of the detector and the x-ray source, relative to the patient. The systems currently in use propose to perform an offline calibration of the C-arm during the preventive maintenance phases of the device, every six months or one year. the exact geometry of the C-arm, i.e. the position of the detector and the x-ray source, relative to the patient. The systems currently in use propose to perform an offline calibration of the C-arm during the preventive maintenance phases of the device, every six months or one year.

Patent US6510241 describes a method for calibrating a radiology device, in which a virtual volume surrounding the object to be imaged is generated and decomposed into voxels (3D pixels). The method comprises a step of acquiring the set of numbered projected two-dimensional images and a reconstruction of the three-dimensional image from the projected image is carried out.

Patent US6320928 describes a method of image reconstruction in which several two-dimensional digital images of an object are acquired for different positions of a camera rotating around the object. The projected images are calibrated in a volume containing the object and divided into voxels whose spatial coordinates are identified within a chosen calibration reference frame.

US6049582 relates to a method of calibrating a C-arm device for 3D reconstruction in an imaging system comprising an imaging source and an imaging plane that uses a plane transformation to connect voxels in a voxel space. and pixels in the imaging plane.

Patent application EP3141187 relates to a calibration test pattern for geometrically calibrating an X-ray imaging device intended to generate three-dimensional images of an object by reconstruction from two-dimensional projections of said object. The calibration target comprises a volume support provided with markers with radiological absorption contrasted with the volume support, the markers being distributed according to a three-dimensional figure. The markers are distributed into subsets of markers distributed along respective substantially parallel lines so that cross-ratio sequences can be constructed from the respective marker subsets.

These systems assume that the rotational acquisition is sufficiently reproducible for the geometry of the C-arm determined "offline" to be applicable on the images acquired during a surgical procedure.

The mechanics of the systems have been improved to make the arch stable during rotational acquisition of 2D images. However, these improvements (reduction of mechanical clearances, use of more rigid parts, etc.) lead to more expensive devices. In addition, it is not always trivial to make these modifications to existing devices.

Other methods of the prior art offer on-line calibration.

A first method is based on the use of markers. A calibration target is positioned on or next to a patient during the procedure. This makes it possible to estimate the geometry of the device precisely and online without worrying about the reproducibility of the measurement conditions. Patent application US 201000284601 describes such a method.

However, the methods known from the prior art which use a test pattern are not optimal in the context of surgical use or other applications with equivalent use constraints, for the reasons explained below:

• The staff must be manufactured with precision so that the 3D position of the points is known with precision, which represents a cost,

• The staff itself can be bulky and difficult to use when the patient is present,

• The staff must have been sterilized, as it is used in a sterile environment, undergo chemical / heat treatment before and after use.

A second type of image reconstruction method is based on the use of an image. These methods exploit the anatomical content of an image to perform both 3D image reconstruction and geometric calibration.

In patent EP2868277, the method uses markers, however it is necessary that the 3D positions of the markers be known precisely in order to determine the geometric parameters.

The invention is based on a new approach using self-adhesive markers without needing to know the 3D position of the markers.

The invention relates to a method for calculating during operation the geometric parameters of an X-ray imaging system, where an object or a patient to be observed is disposed between the X-ray source and an X-ray detector having passed through it. 'object or the patient, characterized in that it comprises at least the following steps:

• Use at least one marker, the marker initially having an unknown 3D position,

• Acquire several 2D images for several points of view of the imaging system,

• Detect the position of at least one marker in each of the 2D images acquired,

• Estimate the projection matrices corresponding to the projections of the object from different viewing angles and reconstruct in 3D the position of a marker from the estimation of the projection matrices.

The method may include an off-line calibration step in order to initially calculate the calibration matrices used during the final step of determining the geometry parameters.

In another variant, the initial projection matrices are calculated by using the orientation sensors or the positioning sensors of the system.

The markers can be inserted or contained in patches positioned on or near the patient or the object and the patches used are, for example, adhesive patches defined as follows:

• A sticky tape that will be affixed to an object or a patient,

• A set of radio-opaque markers distributed over the surface of the patch,

• A fluid resistant outer surface.

It is possible to distribute the markers at the level of a patch in order to cover the entire surface of the patch.

Another possibility is to use small markers distributed over the whole of a compression garment before covering the area to be reconstructed.

Another variant consists in using one or more anatomical markers or else in using radiopaque markers implanted in the patient's anatomy, for example in the bone.

The method can include a step exploiting the geometric characteristics of the system to reconstruct a 3D image.

Other characteristics and advantages of the present invention will emerge better on reading the description of exemplary embodiments given by way of illustration and in no way limiting, appended to the figures which represent:

• Figure 1, a diagram showing a patient in position,

• Figure 2, an example of a patch placed on the patient, and

• Figure 3, a succession of steps implemented by the method according to the invention.

FIG. 1 illustrates an example of a device allowing a practitioner to follow in real time the position of a tool that he uses during a surgical intervention. The device comprises an imaging system comprising an X-ray source, 10, and an X-ray detector, 1 1. The device consists of a hoop 12 or C-arm supporting a first

end 12i the X-ray source and at a second end 12 2 the X-ray detector. The arch or arched arm 12 is held on a frame 13 by a retaining part 14.

A horizontal guide 15 fixed to the frame 13 via a vertical part 16 and to the retaining part 14 of the hoop, ensures a horizontal translational movement of the hoop, arrow H.

The holding part 14 allows the arched arm 12 to perform an “orbital rotation” type movement, according to the arrow R.

The rotation of the part 14 and of the hoop results in an angular rotation according to the arrow A.

The vertical movement is provided by the horizontal translation of the guide and the vertical part.

The joints used between the different elements of the system allowing the aforementioned rotational movements are known to those skilled in the art and will not be detailed. Likewise, the aforementioned movements of the radiology system are known to those skilled in the art.

The device also comprises a processing device 17 comprising a processor 18 adapted to execute the steps of the method according to the invention, in order to determine the geometric characteristics of the device during the intervention of the surgeon. The device may include a screen 19 on which the surgeon can view the position of the tool in real time.

The device is also equipped with orientation 22 or positioning 23 sensors.

A patient 20 is positioned on an operating table 21. FIG. 2 shows schematically an example of implementation of the method where the markers are incorporated or integrated into a patch.

The patch 30 is positioned in the upper part of the body and includes at least one q marker. The coordinates of the patch or the position of the patches are not initially known. It is possible to use one or more patches for the implementation of the method according to the invention.

FIG. 3 illustrates an example of a succession of steps implemented by the method according to the invention, in the case where the markers are integrated into a patch.

Step 1

One or more patches 30 incorporating radiopaque markers are affixed to the skin of the patient or of the object 20, for example in the vicinity of an area to be operated on, 31. This will allow the practitioner to obtain with precision a reconstruction of the patient's body. the organ he has to operate on.

2nd step

Several 2D images are acquired for different points of view in order to carry out a 3D reconstruction, 32. The X-ray source and the detector are moved around the body 20 to be imaged so as to achieve several projections of the body from different angles of view. . The projections thus produced will be used to reconstruct a three-dimensional image of the imaged body.

Step 3

The radiopaque markers contained in the patch 30 are detected in each 2D image acquired at the level of the X-ray detector and matched, from one image to another, 33. Geometric or radiometric similarity criteria are used to carry out the analysis. pairing of markers. Step 4

A 3D reconstruction of the markers is carried out using a first estimation of the projection matrices, 34. These projection matrices Mi can be determined during a prior offline calibration or predicted from the position sensors of the system. These 4 * 3 projection matrices make it possible to match each point of the object or of the patient in 3D space, for example with respect to the terrestrial reference frame, with its projection on a 2D plane detector linked to the detector.

At the end of this pairing step, a first estimate of the 3D position of the markers is obtained.

Step 5

The 3D position of the markers and the knowledge of the projection matrices are then iteratively refined, 35. The geometric parameters, namely the projection matrices as well as the 3D position of the markers, are estimated jointly by minimizing the criterion presented below. . Let be a set of N projections and therefore of N matrices to be determined. Let be a set of L points to reconstruct, the criterion is given by:

OR

q, j denotes the 2D coordinates of a marker of number j detected in an image i obtained by the system,

X is the set of L 3D points to reconstruct, X j the point of number j,

M is the set of N projection matrices, M, the projection matrix of image i.

At the end of this step 35, it is possible to reconstruct, in a precise manner, the corresponding 3D image.

The 3D images thus obtained can be used to allow a practitioner to position his tools with precision during the operation, 36.

The general principle of beam adjustment methods is described in the document entitled “Bundle adjustment - a modem synthesis” by B.trigs, PF Mc Lauchlan, RI Hartley, International Workshop on Vision Algorithms, Corfu, Greece, September 21 -22 , 1999 Proceedings.

Any other algorithm, taking as input the coordinates of a marker, determined by the execution of the method to deduce therefrom the geometric parameters of the device, may be used.

According to an alternative embodiment, the method comprises a preliminary step of offline calibration leading to an imprecise geometry of the C-arm. The calibration matrices resulting from the “offline” calibration are used during the fourth step to perform the first reconstruction.

The markers used at the patch level are, for example, spherical markers in order to facilitate detection. They can also have shapes exhibiting symmetry of revolution around axes of symmetry of revolution.

The patch (s) containing the markers can be stuck directly on the patient or be positioned near the area to be imaged.

In the case of adhesive patches, it is possible to use a patch defined as follows:

• A sticky band which will be affixed to the skin of a patient or an object, before the acquisition of the 2D images useful for the reconstruction of the 3D images,

• A set of radio-opaque markers distributed over the surface of the patch,

• An outer surface resistant to water, blood and friction protecting the patch from its environment. The outer surface is, for example, plastic.

Self-adhesive patches can be disposable.

The markers are for example distributed so as to cover the entire surface of the patch.

The patch and marker set has, for example, a thickness of about 1 mm and an approximate size of 4x14 cm.

The markers can be integrated into a stretchable fabric, or "medical strech suit". The markers of small sizes, for example opaque beads, are for example distributed over the entire compression garment before covering the assembly to be reconstructed, part of the patient, for example. It is then possible to carry out at the same time, on the one hand the calibration of the device using the information obtained during the fourth step and on the other hand the reconstruction of the envelope (3D surface) of the object to be reconstructed. . This envelope will serve as a priori for the 3D reconstruction.

According to an alternative embodiment, the method will use one or more anatomical markers (characteristic and radiopaque part of the human body) which correspond to points of interest present in an image. The markers will be extracted from the images using image processing known to those skilled in the art. In this variant embodiment, the method will not perform the patch positioning step 31. The first step will be to acquire RX images.

In certain cases, it will be possible to combine markers contained in patches and anatomical markers, the latter possibly being implanted in the patient's body, for example in a bone.

The method according to the invention also makes it possible to calibrate a C-arm device using one or more markers, the position of these markers not initially being known.

CLAIMS

1 - Method for calculating during operation the geometric parameters of an X-ray imaging system, an object or a patient (20) to be observed being placed between the X-ray source and an X-ray detector having passed through the 'object or the patient, characterized in that it comprises at least the following steps:

• Use at least one 3D position marker not initially known,

• Acquire several 2D images for several points of view of the imaging system, (32),

• Detect the position of at least one marker in each of the 2D images acquired, (33),

• Estimate the projection matrices corresponding to the projections of the object from different viewing angles and reconstruct in 3D the position of a marker from the estimation of the projection matrices (34, 35).

2 - Method according to claim 1 characterized in that it comprises an offline calibration step in order to calculate the initial projection matrices.

3 - Method according to claim 1 characterized in that the initial projection matrices are calculated by using the orientation or positioning sensors of the system.

4 - Method according to one of claims 1 to 3 characterized in that the markers are contained in an adhesive patch positioned on or near the patient or the object, defined as follows:

• A sticky tape,

• A set of radio-opaque markers distributed over the surface of the patch,

• A fluid resistant outer surface.

5 - Method according to one of claims 1 to 3 characterized in that the markers are distributed at a patch in order to cover the entire surface of the patch.

6 - Method according to one of claims 1 to 4 characterized in that one uses markers integrated in a stretchable fabric.

7 - Method according to one of claims 1 to 5 characterized in that one uses small markers distributed over the whole of a compression garment before covering part of a patient to be reconstructed.

8 - Method according to one of claims 1 to 3 characterized in that one uses at least one anatomical marker.

9 - Method according to one of claims 1 to 3 characterized in that one uses at least one radiopaque marker implanted in the patient's anatomy, such as in the bone.

10 - Method according to one of claims 1 to 9 characterized in that it comprises a step exploiting the geometric characteristics of the system to reconstruct a 3D image.

11 - Device for calculating during operation the geometric parameters of an X-ray imaging system, an object or a patient (20) to be observed being arranged between the X-ray source and an X-ray detector having passed through the 'object or the patient, characterized in that it comprises at least one treatment device (17) comprising a processor (18) adapted to carry out the steps of the method according to one of claims 1 to 10.