Device and method for assistance in a percutaneous tumour ablation procedure

Abstract

The present invention relates to a medical assistance device (1), comprising: - a memory (3) designed to receive pre-procedural image data of one or more zones of interest, - a tool (5) designed to receive current image data of one or more zones of interest, including at least one electromagnetic stimulation needle, obtained by cone beam computed tomography during a percutaneous ablation procedure based on electromagnetic fields, in order to register current image data and pre-procedural image data, and to return image data augmented by cone beam computed tomography, and - a simulator (7) designed to receive image data augmented by cone beam computed tomography and electromagnetic stimulation parameters and to return electromagnetic field simulation data.

Description

[0001] Description

[0002] Device and method for assisting a percutaneous tumor ablation procedure

[0003] The invention relates to the field of interventional oncology, in particular assistance in a percutaneous tumor ablation procedure based on electromagnetic fields.

[0004] When a tumor is detected in a patient, it can usually be removed by surgical ablation, called resection. However, some deep-seated tumors are unresectable, so percutaneous tumor ablation—particularly that based on electromagnetic fields—is a preferred alternative for curative purposes.

[0005] Some of the known techniques involve inserting one or more needles into the tumor percutaneously. In the case of radiofrequency ablation, for example, the needle is connected to a generator to conduct an electric current to the tumor and destroy it with thermal energy. Microwaves can also be used. Apart from thermal treatments, irreversible electroporation involves using needles to subject the tumor to pulsed electromagnetic fields in order to damage the tumor cell membrane and destroy them by coagulation necrosis or apoptosis. The increased permeability of tumor cells can also be used to deliver therapeutic products to them in the context of electrochemotherapy, which involves injecting cytotoxic molecules after applying electrical pulses.

[0006] According to current clinical guidelines, images of the target region and surrounding vital organs are collected several weeks before percutaneous tumor ablation using medical imaging techniques such as magnetic resonance imaging (MRI) or computed tomography (CT scan). On the day of ablation, the interventional radiologist inserts the needles percutaneously into the tumor and verifies their positioning using images obtained by one of these techniques or by cone beam computed tomography (CBCT). Electromagnetic energy is then delivered to the tumor via the needles.

[0007] However, the interventional radiologist does not have visual feedback during the percutaneous ablation procedure. The effectiveness of the treatment is only assessed by follow-up image acquisition (CT-Scan and / or MRI) a few days after the procedure for electroporation or a few weeks for radiofrequency ablation.

[0008] Each medical imaging technique (MRI, CT scan, or CBCT) has its advantages and disadvantages. Magnetic resonance imaging (MRI) and computed tomography (CT scan) provide high-quality images. However, an MRI machine is expensive, has low availability, and, most importantly, is incompatible with metal needles, while CT scans carry high radiation risks since they measure tissue absorption of X-rays. Conversely, cone beam computed tomography is a simple, less irradiating, inexpensive medical imaging technique that can be used during surgery. This technique, although increasingly used in the operating room, provides only low-quality images and therefore does not allow for proper visualization of the tumor.

[0009] The present invention improves the situation.

[0010] In this respect, the invention relates to a medical assistance device comprising:

[0011] - a memory arranged to receive pre-procedural image data of one or more areas of interest,

[0012] - a tool arranged to receive current image data of one or more areas of interest including at least one electromagnetic stimulation needle obtained by cone beam computed tomography during a percutaneous ablation procedure based on electromagnetic fields, to implement a registration of current image data and pre-procedural image data, and to return augmented cone beam computed tomography image data, and

[0013] - a simulator arranged to receive cone beam computed tomography augmented image data and electromagnetic stimulation parameters, and to return electromagnetic field simulation data.

[0014] In one or more embodiments, the device further comprises a display arranged to receive electromagnetic field simulation data, the simulation data comprising electromagnetic field distribution data, and to display a distribution of the electromagnetic field in one or more areas of interest.

[0015] The invention also relates to a medical assistance method implemented by the device described above and comprising:

[0016] - receive pre-procedural image data of one or more areas of interest including all or part of a tumor,

[0017] - receiving current image data of one or more areas of interest including at least one electromagnetic stimulation needle obtained by cone beam computed tomography during a percutaneous ablation procedure based on electromagnetic fields,

[0018] - implement a registration of current image data and pre-procedural image data,

[0019] - return augmented image data from cone beam computed tomography,

[0020] - receive augmented image data from cone beam computed tomography and electromagnetic stimulation parameters, and

[0021] - return electromagnetic field simulation data.

[0022] For example, the preprocedural image data is obtained by magnetic resonance imaging or computed tomography. In one or more embodiments, the preprocedural image data and the current image data relate to three-dimensional images of one or more areas of interest.

[0023] In one or more embodiments, the registration comprises an initial operation of implementing a nearest neighbor algorithm applied to at least one pair formed of a preprocedural image and a current image.

[0024] Advantageously, the preprocedural image and the current image are respectively decomposed into a set of local elements - or patches -, and the nearest neighbor algorithm comprises a propagation during which a distance between a patch of the preprocedural image and a patch of the current image is calculated as follows: where: - 1 is a preprocedural image and J is a current image,

[0025] - VP and VC are the respective centers of the patches of the preprocedural image I and the current image J,

[0026] - r is a local variable corresponding to a displacement vector whose variation allows VP+r and VC+r to cover the entire patch r considered, and

[0027] - V; and V7 are the respective gradients of the preprocedural image and the current image.

[0028] In one or more embodiments, the registration comprises determining a geometric transformation enabling passage from a preprocedural image to a current image, or vice versa, by minimizing an energy function depending on a criterion estimated on the basis of respective gradients of the preprocedural image and the current image.

[0029] The energy function can be defined as follows: where: - T is a geometric transformation,

[0030] - E is the energy function defined on an image domain Q, - D is the estimated criterion,

[0031] - a is a weighting coefficient,

[0032] - u, v and w are the components of the geometric transformation T.

[0033] In one or more embodiments, the energy function is minimized by solving a system of Euler-Lagrange equations obtained by differentiating the energy function with respect to each component of the geometric transformation, the value of each component being calculated by iteration.

[0034] The criterion is for example estimated as follows:

[0035] D(T) = e~ c( ^ where: - T is a geometric transformation,

[0036] - D is the criterion calculated on a patch T,

[0037] - 1 is a preprocedural image and J is a current image,

[0038] - V is a sub-element of patch T, and

[0039] - V; and Vj are the respective gradients of the preprocedural image and the current image.

[0040] In one or more embodiments, a Sobel filter is applied to the pre-procedural image and the current image to determine the magnitude and orientation of their respective gradients.

[0041] In one or more embodiments, the electromagnetic stimulation parameters include the intensity of the electrical current flowing through at least one electromagnetic stimulation needle, and the simulation data includes electromagnetic field distribution data. The electromagnetic field distribution is determined from an effective electrical conductivity of tissues in one or more areas of interest calculated as a function of the intensity of the electrical current flowing through the at least one electromagnetic stimulation needle.In one or more embodiments, one or more pairs of electromagnetic stimulation needles are used to implement the percutaneous ablation procedure based on electromagnetic fields, such that two electromagnetic stimulation needles of a same pair act as active electrodes, and the effective electrical conductivity of the tissues is calculated as follows for such a pair of active electrodes: _ Imes e. ff 9 + f EL + d n u + (x)dx + g~ f EL _ d n u~(x)dx where: - (Jeff is the effective electrical conductivity of the tissues,

[0042] - Imes is the measured electric current intensity,

[0043] - EL + and EL are the active electrodes,

[0044] - g + and g- are the respective electrical voltages of the active electrodes EL + and EL',

[0045] - x is a local variable running through each of the active electrodes EL + and EL', and

[0046] - d n u + , respectively d n u', is the normal component of the potential gradient along the electrode EL + , respectively EL'.

[0047] The invention also relates to a computer program comprising instructions for implementing the method described above, when the instructions are executed by at least one processor.

[0048] Other features, details and advantages will become apparent upon reading the detailed description below, and upon analyzing the attached drawings, in which:

[0049] [Fig. 1] schematically illustrates a device for assisting a percutaneous tumor ablation procedure according to the invention;

[0050] [Fig. 2] illustrates a method of assisting a percutaneous tumor ablation procedure according to the invention; and

[0051] [Fig. 3] illustrates an operation of registering current image data and pre-procedural image data of one or more areas of interest. [Fig. 1] schematically illustrates a device 1 for assisting a percutaneous tumor ablation procedure based on electromagnetic fields, hereinafter referred to as device 1.

[0052] A percutaneous tumor ablation procedure based on electromagnetic fields is particularly suitable when a patient's tumor is unresectable, i.e., when surgery is not possible to remove the tumor completely. This may occur when the tumor is too deep or the patient has contraindications due to their health condition.

[0053] Known techniques of percutaneous tumor ablation - for example radiofrequency or microwave ablation, or irreversible electroporation - involve introducing one or more electromagnetic stimulation needles into the tumor.

[0054] The device 1 comprises a memory 3, a tool 5 and a simulator 7. In the example illustrated in [Fig. 1], the device 1 also comprises a screen 9.

[0055] Memory 3 is arranged to receive pre-procedural image data of one or more areas of interest.

[0056] A "pre-procedural image" here means an image obtained before the percutaneous tumor ablation procedure. Pre-procedural images are therefore obtained before the interventional radiologist performs the percutaneous ablation of the patient's tumor and, more specifically, before any needle is inserted into the patient's tumor.

[0057] Generally speaking, an area of ​​interest refers to an anatomical region of the patient whose visualization is necessary or useful in preparing for or performing the operation. An area of ​​interest includes, for example, all or part of the detected tumor to determine its morphological characteristics, for example, its size or position, particularly in relation to surrounding vital organs. However, an area of ​​interest may also include only parts of the body affected or likely to be affected by the tumor or organs to which the interventional radiologist must pay particular attention during the percutaneous tumor ablation procedure. Preprocedural images of the area(s) of interest are collected up to several weeks before the operation using medical imaging techniques.The techniques used in this instance are magnetic resonance imaging (MRI) or computed tomography (CT-scan), which have the advantage of providing high-quality images and therefore precise visualization of the tumor and surrounding vital organs.

[0058] A preprocedural image can be two-dimensional or three-dimensional.

[0059] In the example shown in [Fig. 1], the pre-procedural IM_CT image shown is acquired by computed tomography.

[0060] The tool 5 is arranged to receive current image data of one or more areas of interest including at least one electromagnetic stimulation needle. It is understood here that an area of ​​interest may comprise only a portion of an electromagnetic stimulation needle.

[0061] By "current image" is meant here an image obtained during the percutaneous tumor ablation procedure, i.e., while the interventional radiologist is ablating the patient's tumor in the operating room. More specifically, a current image is an image obtained after at least one electromagnetic stimulation needle has been inserted.

[0062] Indeed, as explained previously, the percutaneous tumor ablation procedure involves introducing one or more needles into the tumor. A current image therefore has the advantage, compared to a pre-procedural image, of visualizing the position of an electromagnetic stimulation needle.

[0063] Furthermore, in the context of the invention, the current images are obtained by cone beam computed tomography (CBCT). Cone beam computed tomography can indeed be used during the intervention, unlike magnetic resonance imaging, and is less irradiating than computed tomography. As explained previously, medical imaging techniques such as magnetic resonance imaging and computed tomography are rather suited to the acquisition of pre-procedural image data. Cone beam computed tomography consists of an X-ray emitter whose conical-shaped irradiation beam allows, in a single rotation sequence - complete (360°) or not (180°) - around the patient, to scan the entire volume and to reconstruct a three-dimensional representation thereof.Conversely, CT scanning is based on a flattened X-ray which only allows a section of the volume analyzed to be explored.

[0064] The cone beam passes through the patient and is analyzed, after attenuation, by a detector. The detector is integral with the emitter so that, during rotation, the emitter releases a pulse of X-rays which pass through the patient's body to be received by the detector which rotates simultaneously. The sensor is generally a flat panel detector which makes it possible to obtain, at each angular displacement, a two-dimensional image of the volume crossed by the cone beam. The digitized flat panel images - of the order of several hundred - are then processed by a volume reconstruction algorithm for a virtual visualization of the analyzed anatomical structures. A brightness intensifier can be used as an alternative to the flat panel detector and has the advantage of having very high light sensitivity.

[0065] Similar to the pre-procedural image, the current image obtained by cone beam computed tomography can be either a two-dimensional or a three-dimensional image.

[0066] A typical IM_CBCT image, obtained by cone beam computed tomography, is illustrated in [Fig. 1],

[0067] Furthermore, the tool 5 is arranged to implement a resetting of current image data and pre-procedural image data. It is therefore understood that the tool 5 is arranged to communicate with the memory 3 and receive, from the latter, pre-procedural image data.

[0068] Image registration is the process of comparing images to match them, particularly to combine or superimpose them. Registration typically involves estimating a geometric transformation that allows for optimal and consistent superposition of common or similar features present in the processed images.

[0069] The tool 5 is here arranged to receive, on the one hand, a pre-procedural image and, on the other hand, a current image to carry out a registration of these images. This registration is complex due to the fact that the pre-procedural images are obtained by magnetic resonance imaging or computed tomography, whereas the current images are obtained by cone beam digital volume tomography.

[0070] Pre-procedural images and current images are preferably of the same size to facilitate registration.

[0071] The recalibration implemented by tool 5 will be detailed in the rest of the description, with reference to [Fig. 2] and especially to [Fig. 3].

[0072] Finally, tool 5 is arranged to return augmented image data from cone beam computed volume tomography.

[0073] Tool 5 generates images close to the pre-procedural image quality while including the electromagnetic stimulation needles visible in the current images. Such images constitute augmented images, the viewing of which, during the percutaneous tumor ablation procedure, allows the position of the electromagnetic stimulation needles to be precisely checked and adjusted. The augmented image data of cone beam computed tomography can be stored in memory 3.

[0074] The tool 5 may be implemented in the form of an appropriate computer program executed on one or more processors. Such a processor may be implemented in any known manner, for example in the form of a microprocessor for a personal computer, a programmable logic device (PLD) or a dedicated chip of the FPGA (Field Programmable Gate Array) or SoC (System on Chip) type, a grid of computing resources, a microcontroller, or any other specific form having the computing power necessary to implement the medical assistance method described below. One or more of these elements may also be implemented in the form of specialized electronic circuits such as an ASIC (Application-Specific Integrated Circuit).A combination of processors and electronic circuits may also be envisaged. The tool 5 may also comprise a memory or any known type of storage medium for storing instructions of a computer program whose implementation by the processor(s) results in the operation of the tool 5.

[0075] The simulator 7 is arranged to receive augmented image data from cone beam computed tomography.

[0076] The simulator 7 is for example arranged to communicate with the tool 5 to receive the augmented image data of cone beam digitized volume tomography, or with the memory 3 when the augmented image data of cone beam digitized volume tomography are stored therein.

[0077] Furthermore, the simulator 7 is arranged to receive electromagnetic stimulation parameters. These parameters include, for example, the intensity of the electric current flowing in at least one electromagnetic stimulation needle.

[0078] From the cone beam computed tomography augmented image data and the electromagnetic stimulation parameters, the simulator 7 is arranged to return electromagnetic field simulation data.

[0079] The simulator 7 makes it possible to model and translate into data the electromagnetic field to which the tumor area is exposed, whether in the context of radiofrequency or microwave ablation, irreversible electroporation or any other percutaneous tumor ablation technique based on electromagnetic fields. Such data can be used during the operation, for example by the interventional radiologist, to know the distribution of the electromagnetic field generated by the electromagnetic stimulation needle(s). Here again, the electromagnetic field simulation data can be stored in the memory 3.

[0080] Like the tool 5, the simulator 7 may be implemented in the form of a suitable computer program executed on one or more processors. Regarding the implementation of such a processor, the preceding considerations relating to the tool 5 also apply to the simulator 7. Furthermore, here again the simulator 7 may also comprise a memory or any known type of storage medium for storing instructions of a computer program whose implementation by the processor(s) results in the operation of the simulator 7.

[0081] In [Fig. 1], the memory 3, the tool 5 and the simulator 7 are represented as separate elements. However, it must be understood here that this distinction is functional and that, consequently, all or part of these three elements can be constituted by a single element adapted to implement the associated functionalities.

[0082] Finally, the screen 9 is arranged to receive electromagnetic field simulation data. The screen 9 is for example arranged to communicate with the simulator 7 to receive the electromagnetic field simulation data, or with the memory 3 when the electromagnetic field simulation data is stored therein.

[0083] The screen 9 is arranged to display a distribution of the electromagnetic field in one or more areas of interest. The screen 9 allows the interventional radiologist to view live not only the position of the electromagnetic stimulation needles but also the distribution of the electromagnetic field to which the tumor is exposed.

[0084] The screen 9 is optional and can be directly integrated into the device 1 or be removed from it.

[0085] A method for assisting a percutaneous tumor ablation procedure will now be described with reference to [Fig. 2] and [Fig. 3]. In the context of the invention, a tumor is detected in a patient and a percutaneous tumor ablation procedure based on electromagnetic fields is programmed to remove it. In order to prepare for the operation, pre-procedural images of the tumor are collected by magnetic resonance imaging or computed tomography.

[0086] During an operation 200, pre-procedural image data of one or more areas of interest including all or part of a tumor is received by the memory 3.

[0087] Pre-procedural image data may be stored directly in memory 3 or may be retained on another storage medium pending operation.

[0088] During the percutaneous tumor ablation procedure based on electromagnetic fields - for example, radiofrequency or microwave ablation, or irreversible electroporation - the interventional radiologist introduces one or more electromagnetic stimulation needles into the tumor.

[0089] Additionally, in parallel with the percutaneous tumor ablation procedure, cone beam computed tomography (CBE) is used to collect routine images. As explained above, an X-ray transmitter rotates around the patient and emits a cone beam that scans the target region. The X-rays are received by a receiver and routine image data of one or more areas of interest including at least one electromagnetic stimulation needle are then generated.

[0090] During an operation 210, the current image data obtained by cone beam computed volume tomography is received by the tool 5.

[0091] In addition to the current image data, tool 5 also receives pre-procedural image data from memory 3.

[0092] During an operation 220, the tool 5 implements a registration of current image data and preprocedural image data.

[0093] This recalibration is detailed below with reference to [Fig. 3] and can be broken down into two sub-operations: a sub-operation 300 of nearest neighbor algorithm applied to at least one pair formed of a pre-procedural image and a current image, then a sub-operation 310 of determining a geometric transformation making it possible to move from a pre-procedural image to a current image.

[0094] In the following description, the registration of a pre-procedural image and a current image is considered. However, it should be understood here that, advantageously, the tool 5 performs a registration for several pairs formed respectively of a pre-procedural image and a current image.

[0095] Sub-operation 300 corresponds to a “PatchMatch” type algorithm for establishing a correspondence between the patches of the pre-procedural image and their nearest neighbors in the current image. This algorithm makes it possible to obtain a correspondence table, often referred to by the English acronym NNF for “Nearest Neighbor Field”. This correspondence table is equivalent to a function which, to each patch of the pre-procedural image, associates a displacement vector which, starting from the patch of the current image occupying the same position as the considered patch of the procedural image, makes it possible to reach the patch of the current image detected as being the nearest neighbor.

[0096] A "patch" here refers to a local element of the image domain. In the case of a two-dimensional image, a patch is a local square-shaped surface centered on a pixel. In the case of a three-dimensional image, a patch is a local cubic volume centered on a voxel.

[0097] A patch can thus be decomposed into sub-elements which are pixels for a two-dimensional image or voxels for a three-dimensional image.

[0098] The "PatchMatch" algorithm can be broken down into three steps: initialization, propagation, and random search. Hereinafter, the preprocedural image and the current image of the pair on which tool 5 applies a registration are considered to be three-dimensional.

[0099] Initialization consists of randomly associating each patch of the preprocedural image with a patch of the current image. We consider below that the preprocedural image and the current image have been decomposed into the same number of patches, so that we can randomly construct a bijection between the preprocedural image and the current image.

[0100] As explained previously, the pre-procedural image and the current image are three-dimensional. Each patch can therefore be located by three-dimensional coordinates of its central voxel. For example, we use any reference frame O with an abscissa axis X, a ordinate axis Y and a dimension Z. For the sake of simplification, we can provide these axes with an arbitrary graduation so that each patch has integer coordinates.

[0101] We consider below that the preprocedural image and the current image are decomposed into MxNxP patches, where M, N and P are natural numbers. Each patch has coordinates (x,y,z), where x is an integer between 1 and M, y is an integer between 1 and N, and z is an integer between 1 and P.

[0102] Propagation involves examining each patch of the preprocedural image one by one to improve the initial random correspondence with the current image. Typically, the propagation order is to examine the initial correspondence table in ascending order of the axes, from left to right, from top to bottom, and from front to back. This is again a convention and it is possible to proceed differently.

[0103] For each patch examined, we consider all the neighboring patches already examined. With the exception of the patches on the periphery of the preprocedural image, each patch has exactly six neighbors, three of which have already been examined. Each of these three patches is associated with a displacement vector. We then consider, on the current image, the patch whose position corresponds to that of the patch being examined on the preprocedural image. We apply four displacement vectors to this patch on the current image: the displacement vector already associated, during initialization, with the patch examined and the three displacement vectors associated respectively with the neighboring patches of the patch examined on the preprocedural image. We thus arrive at four patches on the current image: the patch already associated, during initialization, with the patch examined and three patches reached by the application of each of the three remaining vectors.We then calculate a distance between the patch under examination and each of these four patches and we choose, among these four patches, the one that allows us to minimize this distance. The patch under examination is then associated with the corresponding displacement vector and therefore, in an equivalent manner, with the corresponding patch of the current image.

[0104] The propagation can be formalized as follows:

[0105] A patch examined on the preprocedural image is centered on a voxel VPi with coordinates (Xi,yi,Zi). The patch examined here is not on the periphery of the preprocedural image, hence: Xi> 1 , yi> 1 , Zi> 1 . The six neighboring patches are centered on voxels with respective coordinates: (Xi-1 ,yi,Zi), (Xi,y 1 ,Zi), (Xi,yi,Zi-1 ), (Xi-1 ,y 1 ,Zi), (Xi-1 ,yi,Zi-1 ) and (Xi,yi-1 ,Zi-1 ). Among these six patches, three have already been examined and are centered on voxels with the following respective coordinates: (Xi-1 ,yi,Zi), (Xi,yi-1 ,Zi) and (Xi,yi,Zi-1 ).

[0106] The examined patch centered on VPi was associated, during initialization, with a displacement vector Ui(0,0,0). The left patch, already examined during propagation, was associated with a displacement vector Ui(-1,0,0). The top patch, already examined during propagation, was associated with a displacement vector Ui(0,-1,0). The front patch, already examined during propagation, was associated with a displacement vector Ui(0,0,-1).

[0107] We now consider on the current image the patch centered on the voxel VCi, with the same coordinates (Xi,yi,Zi) as the voxel VPi on which the examined patch is centered. We apply to the patch centered on the voxel VCi each of the four displacement vectors to reach four voxels on the current image, therefore VO + Ui(0,0,0), VO + Ui(-1,0,0), VO + Ui(0,-1,0) and VCi + Ui(0,0,-1). We calculate the distance between the patch centered on the voxel VPi examined and each of the patches centered on the voxels thus reached. We then select, as the displacement vector corresponding to the patch centered on the voxel VPi, the one allowing this distance to be minimized. Noting u'i(0,0,0) the new displacement vector associated, during propagation, with the patch centered on the voxel VPi, we therefore have: u (0,0,0) = argmin^D VP . VCi + U^OOO^. D VP^ V It is possible to define the distance D in different ways. In the context of the invention, a distance is chosen that allows for the alignment of contours between the pre-procedural image and the current image to be favored. By setting and V } the respective gradients of the preprocedural image I and the current image J, the distance D(VPi ;VCi+Ui) between a patch centered on the voxel VPi on the preprocedural image I and the corresponding patch potential centered on the voxel VCi+Ui on the current image J can be defined as follows: where the local variable r is a displacement vector whose variation allows VPi+r and VCi+r to cover the entire patch T considered. In other words, any voxel of the patch considered on the preprocedural image I, respectively on the current image J, can be reached from the central voxel VPi, respectively VCi+Ui, with a displacement vector r. Thus, for any voxel V of the patch centered on the voxel VPi of the preprocedural image, respectively centered on the voxel VCi+Ui of the current image, there exists a displacement vector r such that V=VPi+r, respectively V=VCi+Ui+r.

[0108] Finally, random search consists of improving each displacement vector associated with a patch of the preprocedural image by testing several candidates located at an exponentially decreasing distance from the nearest neighbor found during propagation. Typically, for each patch of the current image, propagation is first applied, then random search, and then the next patch is moved.

[0109] We again consider the patch of the preprocedural image centered on the voxel VPi. At the end of the propagation, this patch is now associated with the displacement vector u'i(0,0,0) and therefore, equivalently, with the patch centered on the voxel VCi+u'i(0,0,0) on the current image.

[0110] We then test a set of candidate displacement vectors Uk(0, 0, 0) of the following form: u k (0,0,0) = u'j(0,0,0) + a>p k ri where: - k is a natural integer,

[0111] - œ is a maximum search radius,

[0112] - P is a fixed ratio between the search window sizes, typically equal to 1 / 2, and

[0113] - n is a random vector in [-1, 1] 3 .

[0114] For each candidate Uk(0,0,0), we measure the distance between the patch centered on voxel VPi in the pre-procedural image and the patch centered on voxel VCi + Uk(0,0,0) in the current image. We then select, as the displacement vector corresponding to the patch centered on voxel VPi, the one allowing this distance to be minimized.

[0115] It is worth noting that propagation and random search can be repeated several times to improve the correspondence table between the preprocedural image and the current image.

[0116] Sub-operation 310 corresponds to the actual registration of the current image data and the pre-procedural image data. In practice, the registration consists of determining a geometric transformation making it possible to move from the pre-procedural image I to the current image J. Such a geometric transformation makes it possible to align the pre-procedural image I and the current image J. Equivalently, the current image J is registered on the pre-procedural image I. It should be noted that sub-operation 310 can be carried out alone, without sub-operation 300 which is therefore optional but makes it possible to facilitate and accelerate the implementation of the registration.

[0117] Of course, it is also possible to re-align the pre-procedural image I on the current image J, thus determining a geometric transformation allowing one to move from the current image I to the pre-procedural image J. It is understood that, in the rest of the description, the re-alignment can be implemented by inverting I and J.

[0118] To do this, we first determine an estimated criterion based on the respective gradients and V7 of the preprocedural image I and the current image J for a given geometric transformation T. Such a criterion can be estimated as follows by integrating over all voxels V of a patch f: D(T) = e~ c ^

[0119] Optionally, tool 5 can apply a Sobel filter to the preprocedural image I and the current image J to determine the amplitudes Mi, Mj and orientations 0i, 0j of their respective gradients at a voxel V. We then obtain the following expression: where WT(V) and A0T(V) are calculated as follows:

[0120] Once the criterion D has been determined for any geometric transformation T, the tool 5 determines the geometric transformation allowing the transition from the preprocedural image I to the current image J by minimizing an energy function depending on this criterion.

[0121] The energy function E is for example defined as follows on the image domain Q which includes all the patches T: where: - a is a weighting coefficient, and

[0122] - u, v and w are the components of the geometric transformation T.

[0123] As explained previously, here we consider that the preprocedural image and the current image are three-dimensional and, therefore, that the geometric transformation allowing to pass from one to the other has three components. However, the same procedure can be applied when the preprocedural image and the current image are two-dimensional, in which case it is sufficient to remove one of the components of the geometric transformation. Tool 5 can minimize the energy function by solving a system of Euler-Lagrange equations obtained by deriving the energy function with respect to each component of the geometric transformation.

[0124] Such a system of equations can then take the following form using a Laplacian operator A:

[0125] The value of each component of the geometric transformation can be calculated by iterations:

[0126] The previous system corresponds to the p-th iteration. The sequences u p , v p and w p thus defined converge respectively towards the components of the desired geometric transformation.

[0127] During an operation 230, the tool 5 returns augmented image data of cone beam computed tomography. The augmented image data characterizes an augmented image obtained by registration from a pair formed of a pre-procedural image and a current image. This augmented image of cone beam computed tomography has a quality close to the pre-procedural image and comprises a region of interest including at least one electromagnetic stimulation needle identified on the current image.

[0128] The cone beam computed tomography augmented image data can be stored in memory 3. The cone beam computed tomography augmented image data allows the position of the electromagnetic stimulation needles to be adjusted, if necessary, to ensure that they are correctly introduced into the tumor without damaging a nearby body part or organ.

[0129] During an operation 240, the simulator 7 receives cone beam computed tomography augmented image data and electromagnetic stimulation parameters.

[0130] The augmented cone beam computed volume tomography image data may be received by the simulator 7 directly from the tool 5 or retrieved from the memory 3.

[0131] As explained above, the electromagnetic stimulation parameters may include the intensity of the electrical current flowing through at least one electromagnetic stimulation needle. Preferably, these parameters include the intensity of the electrical current flowing through each electromagnetic stimulation needle.

[0132] During an operation 250, the simulator 7 returns electromagnetic field simulation data.

[0133] To do this, simulator 7 calculates an effective electrical conductivity of the tissues based on the intensity of the electrical current flowing in each electromagnetic stimulation needle.

[0134] The following is a case in which one or more pairs of electromagnetic stimulation needles are introduced into the patient's tumor. This case occurs in particular when the percutaneous tumor ablation technique used is irreversible electroporation. Indeed, irreversible electroporation operates in bipolar mode - also referred to as multi-bipolar mode when several pairs of needles are introduced - so that an electromagnetic field is created between each pair of electromagnetic stimulation needles. The electromagnetic stimulation needles forming a pair then act as active electrodes to deliver pulses of electromagnetic fields. The effective electrical conductivity of the tissues can be calculated as follows for a pair of active electrodes - or electromagnetic stimulation needles -:

[0135] _ Imes e ff 9 + f EL + d n u +(x)dx + g~ f EL _ d n u~(x)dx where: - (Jeff is the effective electrical conductivity of the tissues,

[0136] - Imes is the measured electric current intensity,

[0137] - EL + and EL- are the active electrodes,

[0138] - g + and g- are the respective electrical voltages of the active electrodes EL + and EL',

[0139] - x is a local variable running through each of the active electrodes EL + and EL', and

[0140] - d n u + , respectively d n u', is the normal component of the potential gradient along the electrode EL + , respectively EL'.

[0141] The simulator 7 then determines, from the effective electrical conductivity of the tissues, electromagnetic field distribution data. This data makes it possible to know the distribution of the electromagnetic field in the tumor area and therefore, here too, to adjust the position of the electromagnetic stimulation needles or to vary the intensity of the electrical current in each electromagnetic stimulation needle to carry out percutaneous tumor ablation in the most efficient way possible.

[0142] Electromagnetic field simulation data can be stored in memory 3. The simulation data is for example stored in DICOM format (acronym for “Digital Imaging and COmunication in Medicine”) which is a classic standard for data from medical imaging.

[0143] Finally, during an optional operation 260, the screen 9 receives electromagnetic field simulation data.

[0144] The electromagnetic field simulation data may be received by the screen 9 directly from the simulator 7 or retrieved from the memory 3. Advantageously, the electromagnetic field simulation data includes electromagnetic field distribution data, such that the screen 9 displays a distribution of the electromagnetic field in one or more areas of interest, and in particular in the tumor area. Displaying the electromagnetic field distribution on the screen 9 allows the interventional radiologist to have visual feedback during the percutaneous tumor ablation procedure and to adjust the position of the electromagnetic stimulation needles.

[0145] Of course, feedback other than visual feedback is possible so that the simulation data returned by the simulator 7 can be rendered in another form so as to inform the interventional radiologist of the position of the electromagnetic stimulation needles and the electromagnetic field to which the tumor area is exposed.

Claims

Demands

1. Medical assistance device (1) comprising: - a memory (3) arranged to receive pre-procedural image data from one or more areas of interest, - a tool (5) arranged to receive current image data from one or more areas of interest including at least one electromagnetic stimulation needle obtained by cone-beam computed tomography during a percutaneous ablation procedure based on electromagnetic fields, to perform registration of current image data and pre-procedural image data, and to return augmented cone-beam computed tomography image data, and - a simulator (7) arranged to receive augmented image data from cone beam computed tomography and electromagnetic stimulation parameters, and to return electromagnetic field simulation data.

2. Device (1) according to claim 1, further comprising a screen (9) arranged to receive electromagnetic field simulation data, said simulation data comprising electromagnetic field distribution data, and to display an electromagnetic field distribution in one or more areas of interest.

3. A medical assistance method implemented by the device (1) according to claim 1 or 2 and comprising: - receive (200) pre-procedural image data from one or more areas of interest including all or part of a tumor, - receive (210) current image data of one or more areas of interest including at least one electromagnetic stimulation needle obtained by cone beam computed tomography during a percutaneous ablation procedure based on electromagnetic fields, - implement a registration (220) of current image data and pre-procedural image data, - return (230) augmented image data from cone beam computed tomography, - receive (240) augmented image data from cone beam computed tomography and electromagnetic stimulation parameters, and - return (250) electromagnetic field simulation data.

4. A method according to claim 3, wherein the pre-procedural image data are obtained by magnetic resonance imaging or computed tomography.

5. A method according to claim 3 or 4, wherein the pre-process image data and the current image data are related to three-dimensional images of one or more areas of interest.

6. A method according to any one of claims 3 to 5, wherein the registration (220) comprises an initial operation (300) of implementing a nearest neighbor algorithm applied to at least one pair consisting of a pre-procedural image and a current image.

7. A method according to claim 6, wherein the pre-process image and the current image are respectively decomposed into a set of local elements - or patches -, and wherein the nearest neighbor algorithm includes a propagation during which a distance between a patch of the pre-process image and a patch of the current image is calculated as follows: where: - 1 is a pre-procedural image and J is a current image, - VP and VC are the respective centers of the patches of the pre-procedural image I and the current image J, - r is a local variable corresponding to a displacement vector whose variation allows VP+r and VC+r to cover the entire patch r under consideration, and - V; and V7 are the respective gradients of the pre-procedural image and the current image.

8. A method according to any one of claims 3 to 7, wherein the registration (220) comprises the determination (310) of a geometric transformation enabling the conversion from a pre-procedural image to a current image, or vice versa, by minimizing an energy function depending on a criterion estimated on the basis of respective gradients of the pre-procedural image and the current image.

9. A method according to claim 8, wherein the energy function is defined as follows: where: - T is a geometric transformation, - E is the energy function defined on an image domain Q, - D is the estimated criterion, - a is a weighting coefficient, - u, v and w are the components of the geometric transformation T.

10. A method according to claim 9, wherein the energy function is minimized by solving a system of Euler-Lagrange equations obtained by differentiating the energy function with respect to each component of the geometric transformation, the value of each component being calculated by iteration.

11. A method according to any one of claims 8 to 10, wherein the criterion is estimated as follows: where: - T is a geometric transformation, - D is the criterion calculated on a patch T, - 1 is a pre-procedural image and J is a current image, - V is a sub-element of patch T, and - V; and V7 are the respective gradients of the pre-procedural image and the current image.

12. A method according to any one of claims 8 to 11, wherein a Sobel filter is applied to the pre-processing image and the current image to determine the amplitude and orientation of their respective gradients.

13. A method according to any one of claims 3 to 12, wherein the electromagnetic stimulation parameters include the intensity of the electric current flowing through at least one electromagnetic stimulation needle, and wherein the simulation data include electromagnetic field distribution data, said electromagnetic field distribution being determined from an effective electrical conductivity of tissues in one or more areas of interest calculated as a function of the intensity of the electric current flowing through said at least one electromagnetic stimulation needle.

14. A method according to claim 13, wherein one or more pairs of electromagnetic stimulation needles are used to implement the percutaneous ablation procedure based on electromagnetic fields, such that two electromagnetic stimulation needles of the same pair act as active electrodes, and wherein the effective electrical conductivity of the tissues is calculated as follows for such a pair of active electrodes: _ Imes e ff 9 + f EL + d n u + (x)dx + g~ f EL _ d n u~(x)dx where: - Oeff is the effective electrical conductivity of the tissues, - Imes is the measured intensity of the electric current, - EL + and EL are the active electrodes, - g + and g- are the respective electrical voltages of the active electrodes EL + and EL', - x is a local variable that spans each of the active electrodes EL + and EL', and - d n u + , respectively d n u', is the normal component of the potential gradient along the electrode EL + , respectively EL'.

15. Computer program comprising instructions for carrying out the method according to any one of claims 3 to 14, when said instructions are executed by at least one processor.

Images