Device to assist in planning a minimally invasive intervention on a bone
The data processing device aids in selecting optimal minimally invasive intervention trajectories by generating a 3D anatomical model, applying validity criteria, and calculating cost functions, addressing the limitations of existing methods in planning efficiency and constraint consideration.
Patent Information
- Application Number
- FR2022013698
- Authority / Receiving Office
- FR · FR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2022-12-16
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2042-12-16
AI Technical Summary
Existing methods for planning minimally invasive medical interventions lack the ability to quickly and optimally select trajectories that consider multiple constraints and patient-specific needs, relying heavily on practitioner expertise and two-dimensional image analysis.
A data processing device that utilizes a processor and computer memory to generate a three-dimensional anatomical model, identify target points, sample candidate entry points, apply validity criteria, calculate cost functions, and display optimal trajectories for selection, allowing for user adjustment and refinement.
Facilitates rapid and optimized exploration of candidate trajectories, considering various constraints, enabling practitioners to choose trajectories that may be suboptimal according to predefined criteria but preferable for patient-specific needs, enhancing intervention planning efficiency and safety.
Smart Images

Figure 00000028_0000 
Figure 00000028_0001 
Figure 00000029_0000
Abstract
Description
Title of the invention: Device for assisting in planning a minimally invasive intervention on a bone Field of invention
[0001] The present application belongs to the field of planning a minimally invasive medical intervention, possibly assisted by a medical robot. In particular, a data processing device is proposed for implementing a method for assisting in the selection of an optimal trajectory for a minimally invasive medical intervention within an anatomy of interest of a patient. State of the art
[0002] To prepare a minimally invasive medical intervention aimed at reaching a target anatomical area in a patient's anatomy of interest with a medical instrument, a practitioner generally plans the intervention based on a preoperative or intraoperative medical image. The minimally invasive medical intervention may in particular aim to perform the ablation or biopsy of a tumor in an organ or in a bone, to perform vertebroplasty or cementoplasty, or to stimulate a particular anatomical area. The anatomy of interest may correspond, for example, to a lung, a kidney, the liver, the brain, a vertebra, the tibia, the knee, etc. The medical instrument may be a needle, an electrode, a probe, etc.
[0003] The preoperative or intraoperative medical image is for example obtained by computed tomography, magnetic resonance imaging, ultrasound, or positron emission tomography.
[0004] During planning, the practitioner defines a target point in a region of the anatomy of interest to be treated. The practitioner also defines an entry point for the medical instrument on the patient's skin. These two points then define a trajectory that the medical instrument must follow to perform the medical intervention.
[0005] Depending on the type of intervention, certain constraints must be respected. For example, it may be important that the medical instrument does not pass through a risky anatomical structure (organ, bone or blood vessel for example).
[0006] The trajectory is generally defined empirically based on the practitioner's knowledge and the information he sees on the preoperative image. However, this trajectory is not necessarily optimal to ensure effective treatment with minimal risk to the patient.
[0007] For a given intervention, there is theoretically an infinite number of possible trajectories. The practitioner must, however, select a trajectory in a limited time, sometimes very short, particularly when the planning of the intervention is performed just before the intervention.
[0008] This task must also be carried out using volumetric images of the patient, and therefore in three dimensions, whereas the conventional methods for presenting these images are two-dimensional sections of this volume, which makes the search for the optimal trajectory even more difficult for the practitioner. In addition, when several trajectories must be considered for the intervention (for example to insert several needles), planning is even more complex for the practitioner to carry out mentally.
[0009] Ultimately, the quality of trajectory optimization is largely dependent on the practitioner and the time he can allocate to this task.
[0010] There are methods for evaluating a trajectory proposed by a practitioner, for example based on the risk of crossing a risk region with the medical instrument. Such a solution does not allow the practitioner to quickly find an optimal or near-optimal trajectory for the medical intervention considered.
[0011] There are also planning methods for which the entry point, the target point and / or the trajectory are automatically proposed to the practitioner. However, these solutions generally do not give the user flexibility to visualize the optimality of a large number of trajectories, and to choose a different trajectory that may be suboptimal, but which is preferable for reasons that may vary from one patient to another and which are not necessarily taken into account by the optimality criteria considered. Statement of the invention
[0012] The solutions proposed in the present application aim to remedy all or part of the drawbacks of the prior art, in particular those set out above.
[0013] For this purpose, and according to a first aspect, there is provided a data processing device comprising a processor, a computer memory and a display screen. The computer memory comprises program code instructions which, when executed by the processor, configure the processor to implement a method for assisting in the selection of at least one optimal trajectory for a minimally invasive medical intervention for treating a bone of a patient. The processor is configured to: - obtain a three-dimensional anatomical model of the patient from medical images previously acquired on the patient, the anatomical model including a representation of the body envelope and the bone to be treated, - identify in the anatomical model at least one target point to be reached in the bone to be treated, - perform at least one iteration for which the processor is configured to: • determining a sampling region and a sampling resolution for sampling candidate entry points in the anatomical model at the patient's body envelope, each candidate entry point belonging to the sampling region, the sampling resolution being representative of a minimum distance separating two candidate entry points, each pair formed by the target point and a candidate entry point forming a candidate trajectory, • eliminate candidate trajectories based on at least one predetermined validity criterion, • for each remaining candidate trajectory, quantifying each of a plurality of predetermined characteristics, including a minimum distance between a portion of the candidate trajectory located inside the bone to be treated and a cortical wall of the bone, and calculating at least one cost function from the quantified characteristics, • display on the display screen, superimposed on the anatomical model, at least part of the remaining candidate trajectories with, for each candidate trajectory displayed, a means of quickly visualizing a value of the cost function, • obtain the selection, by a user, for said at least one target point, of a trajectory considered optimal among the candidate trajectories displayed.
[0014] Such arrangements make it possible to propose to a practitioner a set of optimal or near-optimal trajectories for a minimally invasive intervention from a three-dimensional anatomical model of the patient. The proposed trajectories take into account constraints of different types (for example practical constraints, performance constraints and / or safety constraints). The practitioner has the possibility of choosing a trajectory that may be suboptimal in the sense of a predefined criterion, but preferable for constraints not captured by this criterion. The invention also allows a rapid and optimized exploration of all the solutions to determine candidate trajectories. It allows a rapid visualization by the user of different candidate trajectories as well as their optimality criteria. The proposed solution is flexible since it allows adaptation to specific needs of the intervention.The practitioner also has the possibility of updating optimality criteria (cost function) based on user feedback (evaluation of a trajectory selected for the intervention).
[0015]
[0016]
[0017]
[0018]
[0019]
[0020]
[0021]
[0022]
[0023]
[0024] In particular embodiments, the device may further comprise one or more of the following features, taken individually or in any technically possible combination. In particular embodiments, the processor is configured to perform at least two iterations. For an iteration of rank n, n being an integer strictly greater than one: - the sampling region of the iteration of rank n is determined to be smaller than the sampling region determined for the iteration of rank (n-1), and to include the candidate entry point corresponding to the trajectory selected at the iteration of rank (n-1), and - the sampling resolution of the iteration of rank n is determined to be finer than the sampling resolution determined for the iteration of rank (n-1). In particular embodiments, the sampling resolution of the iteration of rank n is determined based on a variability of the cost function calculated for the iteration of rank (n-1) in the vicinity of the trajectory selected at the iteration of rank (n-1). In particular embodiments, the sampling region of the iteration of rank n is determined based on a variability of the cost function calculated for the iteration of rank (n-1) in the vicinity of the trajectory selected at the iteration of rank (n-1). In particular embodiments, to sample the candidate input points, the processor is configured to determine in the anatomical model, at the patient's body envelope, a set of points defined by a spherical coordinate system centered on the target point. Each point is defined by a distance r^ from the target point and by two angles a; and [3j, the angles a; and [3j being defined such that, for indices i and j corresponding to strictly positive integers: [Math.l] aiJ - + 7“ And [Math.2] / 7.. = / 7..,+^ where R is representative of a minimum distance separating two candidate entry points at the level of the patient's body envelope. In particular embodiments, a validity criterion makes it possible to verify, for a given candidate trajectory, at least one of the following elements: - a validity of the length of the candidate trajectory in relation to a medical instrument envisaged for the intervention, - a possibility of configuring a robotic arm equipped with a medical instrument so that said medical instrument can follow the candidate trajectory, - the presence of objects placed on the patient and obstructing the candidate's trajectory, - an intersection of the candidate trajectory with at least one critical anatomical structure internal to the patient's body.
[0025] In particular embodiments, several validity criteria are considered to eliminate candidate trajectories, and the different validity criteria are evaluated according to an order defined as a function, for each validity criterion, of an estimated calculation time necessary to evaluate said validity criterion.
[0026] In particular embodiments, the plurality of characteristics of a candidate trajectory comprises at least one of the following characteristics: - a minimum distance between the candidate trajectory and a critical anatomical structure internal to the patient's body, - a length of the intersection between the candidate trajectory and the bone to be treated, - a minimum angle of incidence between the candidate trajectory and an interface anatomical crossed by the candidate trajectory, - a minimum distance between an envelope of the region to be treated and an envelope of an ablation region estimated for the candidate trajectory, - an angular deviation between the candidate trajectory and a major axis of the region to be treated.
[0027] In particular embodiments, several cost functions are calculated from the quantified characteristics, each cost function being calculated according to a different set of weights assigned respectively to the different quantified characteristics, and the processor is configured to obtain an indication, by a user, of the cost function to be considered for the display of the candidate trajectories.
[0028] In particular embodiments, the means for quickly displaying the value of the cost function for each displayed candidate trajectory comprises a color code associating different colors respectively with different values of the cost function.
[0029] In particular embodiments, the processor is configured to display the quantizations determined for the characteristics of the trajectory selected by the user.
[0030] In particular embodiments, the processor is configured to obtain, for at least one displayed candidate trajectory, a user evaluation of said candidate trajectory, and to update the cost function based on said evaluation.
[0031] In particular embodiments, the processor is configured to: - determine in the anatomical model, after the intervention, a real trajectory followed by a medical instrument during the intervention, from medical images acquired on the patient at a time when the medical instrument is in place, - determine a value of the cost function for the actual trajectory, - calculate a difference between the value of the cost function of the trajectory candidate selected for intervention and the value of the cost function of the actual trajectory.
[0032] In particular embodiments, the processor is configured to compare the calculated deviation with a predetermined threshold and display an indication relating to the value of the calculated deviation.
[0033] In particular embodiments, the processor is configured to: - obtain an assessment of the actual trajectory by the user, - determine, based on the calculated deviation and the evaluation obtained, a statistical value representative of the variability of the cost function, - eliminate candidate trajectories considered equivalent according to the statistical value thus obtained. Presentation of figures
[0034] The invention will be better understood on reading the following description, given by way of non-limiting example, and made with reference to Figures 1 to 7 which represent:
[0035] [Fig-1] a schematic representation of a data processing device according to the invention,
[0036] [Fig.2] a schematic representation of the main stages of a method of assisting the selection of an optimal trajectory for a medical intervention,
[0037] [Fig.3] a schematic representation of the definition of an orbital angle and a craniocaudal angle to define a sampling region,
[0038] [Fig.4] a schematic representation of a set of candidate entry points in the anatomical model at the level of the patient's body envelope,
[0039] [Fig.5] a schematic representation of the display of a set of candidate trajectories for medical intervention, with their optimality criterion visualized via a scale, represented here in gray levels,
[0040] [Fig.6] a schematic representation of additional steps to take into account a user's evaluation of a candidate trajectory,
[0041] [Fig.7] a schematic representation of additional steps to take into account takes into account a difference between the selected trajectory and the trajectory actually followed by the medical instrument,
[0042] [Fig.8] a schematic representation of a bone to be treated and two candidate trajectories to reach a target point in the bone.
[0043] In these figures, identical references from one figure to another designate identical or similar elements. For reasons of clarity, the elements represented are not necessarily on the same scale, unless otherwise stated.
[0044] Detailed description of at least one embodiment of the invention
[0045] [Fig.l] represents a data processing device 10 according to the invention. The data processing device 10 comprises at least one processor 11, at least one computer memory 12 and at least one display screen 13.
[0046] The computer memory 12 comprises program code instructions which, when executed by the processor 11, configure the processor 11 to implement a method for assisting in the selection of at least one optimal trajectory that a medical instrument will have to follow to perform a minimally invasive medical intervention within an anatomy of interest of a patient.
[0047] The minimally invasive medical intervention may in particular aim to perform the ablation or biopsy of a tumor in an organ or in a bone, to treat a bone pathology, to perform a vertebroplasty or a cementoplasty, or to stimulate a particular anatomical zone. The anatomy of interest may correspond to an organ or a bone, for example a lung, a kidney, the liver, the brain, a vertebra, the tibia, the femur, the hip, the knee, the bones of the pelvis, the pelvis, etc. The medical instrument may be a needle, an electrode, a probe, a drill, a trocar, a screw, etc.
[0048] [Fig.2] represents the main steps of a method 100 implemented by the treatment device 10 to assist in the selection of an optimal trajectory for a minimally invasive medical intervention.
[0049] As illustrated in [Fig.2], the method 100 comprises a step 101 of obtaining a three-dimensional anatomical model of the patient. The anatomical model notably comprises a representation of the body envelope (or a part of the body envelope) and the anatomy of interest of the patient.
[0050] The anatomical model is obtained from medical images previously acquired on the patient. The medical images that make it possible to generate the anatomical model are for example obtained by computed tomography, magnetic resonance imaging, ultrasound, or positron emission tomography (any medical imaging modality that makes it possible to reconstruct a volume in three dimensions can be used). Several images are generally necessary to generate a three-dimensional anatomical model. Alternatively, it is possible to generate a three-dimensional anatomical model from a single image and a statistical model of the patient.
[0051] The anatomical model can be generated directly by the processing device 10, but it can also possibly be generated by another separate device and transmitted to the processing device 10 via communications means.
[0052] Conventionally, three-dimensional modeling is, for example, performed using segmentation methods, on medical images, of anatomical structures considered relevant by clinical practice. The results of the segmentations can be represented in the form of binary images or surfaces. For example, in the context of minimally invasive removal of a liver tumor, it may be relevant to segment the liver, the tumor, the lungs, the gallbladder, the bile ducts, the organs of the digestive system, the bones, certain blood vessels (those with a significant diameter), and the patient's body envelope.As another example, in the context of a minimally invasive intervention on a bone, it may also be relevant to segment the external envelope of the bone to be treated, its cortical part (the cortical bone corresponds to the particularly rigid peripheral part of the bone, this corresponds to the "external shell" of the bone, that is to say to a more or less thick wall of the bone), and / or its spongy part (the spongy bone corresponds to the bone tissue that forms the porous part of the bone, located below the cortical bone). This three-dimensional modeling of the anatomy can be represented mathematically as follows: .
[0053] [Math.3] M = U^s,
[0054] where M is the three-dimensional anatomical model and Si are the representations of the segmentations considered to generate this model, the different segmentations all being represented in the same anatomical reference point.
[0055] Segmentation methods are generally automatic (no user input is then necessary), semi-automatic (a single user input such as a point or a segment is then necessary), or interactive (segmentation results are iteratively corrected by a user) artificial intelligence methods. Manual segmentation could also be considered, but the time required to perform such segmentation makes this alternative of little relevance in a clinical context.
[0056] Among artificial intelligence methods, convolutional neural networks are currently the most efficient. Other methods may, however, present interesting alternatives, such as methods based on decision trees ("random forest"), on partitioning in k-means ("k-means"), on segmentation by watershed line, on a model of active contour model, on level surfaces (level-set method), etc.
[0057] As illustrated in [Fig. 2], the method 100 comprises a step 102 of identifying, in the anatomical model, a target point to be reached in the region to be treated (for example a tumor) within the anatomy of interest of the patient (for example the liver). It should be noted that one can also envisage the case where several target points must be considered (for example when several trajectories must be selected to insert several needles). In this case, this set of target points can be determined from a “virtual” target point and a geometric arrangement that the target points must respect in relation to each other and in relation to the virtual target point (for example the target points must correspond to the vertices of a particular geometric shape centered on the virtual target point, or follow a particular distribution in space around the virtual target point).The virtual target point and the geometric arrangement that the target points must respect are input data provided by the user.
[0058] The region to be ablated may in particular be segmented automatically or semi-automatically. The use of correction tools or manual segmentation is possible depending on the desired quality. Once a satisfactory segmentation of the region to be treated has been obtained, the position of a target point to be reached may be estimated. It should be noted that the region to be treated may be segmented by the treatment device 10, or it may have been previously segmented on the anatomical model by another separate device and then transmitted to the treatment device 10 via communication means.
[0059] The segmentation of the area to be treated can for example be used to determine a target point automatically (for example with a calculation of the barycenter or center of a sphere or an encompassing ellipsoid determined according to a priori known parameters). Alternatively, when the intervention involves the insertion of several medical instruments, it is possible to use the segmentation to determine a set of several target points, for example according to a virtual target point (corresponding for example to the center of the region to be treated), a geometric arrangement that the target points must respect, and / or a desired ablation coverage (it is indeed possible, for a given medical instrument, to predict an ablation zone likely to be obtained around the target point in order to ensure that it covers a desired ablation zone).
[0060] Depending on the type of treatment envisaged (radiofrequency ablation, microwave, cryotherapy or electroporation, biopsy, cementoplasty, etc.), the target point may be located inside or on the periphery of the region to be treated. This estimation of the position of the target point may possibly be carried out according to a mode identification of the ablation zone likely to be obtained when the medical instrument is positioned at this target point.
[0061] The candidate trajectories for the intervention will be determined from the target point(s) thus defined and from a set of candidate entry points at the patient's skin. The set of entry points at the skin therefore constitutes an initial domain from which an optimal trajectory must be identified. Indeed, each pair formed by a target point and a candidate entry point forms a candidate trajectory. It is therefore necessary to identify candidate entry points at the patient's skin. This can be achieved by methods such as that of Otsu, which is well suited for the classification of voxels in the case of bimodal intensity distributions. However, in continuous space there is an infinite number of candidate entry points. Even in the discrete domain, for the image resolutions used in clinical practice, there is a very large number of candidate entry points.It is therefore preferable to carry out sampling in order to reduce the number of candidate points.
[0062] For this purpose, and as illustrated in [Fig.2], the method 100 comprises a step 103 of determining a sampling region and a sampling resolution for sampling candidate entry points in the anatomical model at the level of the patient's body envelope.
[0063] The sampling resolution is representative of a minimum distance separating two candidate entry points. A compromise must be found between the fineness of the sampling and the computation time required to evaluate the candidate trajectories respectively associated with each candidate entry point. A “brute force” approach with a high fineness of sampling of the candidate entry points would impose a very fast computation time for the evaluation of the trajectories, which would lead to strong constraints (limitation of the optimization criteria and / or use of expensive computational means). In addition, the user would have to visualize a large quantity of information.
[0064] It may therefore be advantageous to consider a sampling resolution such that two contiguous trajectories are sufficiently different from each other. The sampling resolution can be defined by default (for example, a minimum distance between two entry points of at least five millimeters on the patient's body surface can be imposed), but it can also be adjusted by the user according to needs (compromise between calculation time and desired precision).
[0065] To sample the candidate entry points, it is notably possible to determine in the anatomical model, at the level of the patient's body envelope, a set of points defined by a spherical coordinate system centered on the target point (when several target points are considered, it is possible to process each target point successively in order to determine an optimal trajectory individually for each of the target points; alternatively, it is possible to process a set of target points in a grouped manner, and in this case it is possible to sample the candidate input points with respect to a "virtual" target point corresponding for example to a point equidistant from each of the target points). Each point can then be defined by a distance r^ from the target point and by two angles a; and [3j. The sampling region can then correspond to the intersection of the body envelope with a cone of angular opening (according to angles amax and |3max) centered on the target point. The calculation is for example initialized with a first point at the skin level corresponding to a trajectory with a0= (30 = 0; the distance between this point and the origin of the reference frame (the target point) is noted rOjO.Sampling can be obtained by calculating angles a; and [3; (in radians) such that an increment between each angle corresponds to an arc length on the patient's body envelope corresponding to a predefined resolution parameter. The angles a; and [3j can notably be defined such that, for indices i and j corresponding to strictly positive integers: .
[0066] [Math.l] aij = a;AJ + £:
[0067] and
[0068] [Math.2]
[0069] where R is a predefined resolution parameter representative of a minimum distance separating two candidate entry points at the patient's body envelope. The angles a and [3 are incremented until they cover the desired opening cone.
[0070] As illustrated in [Fig.3], the angle a may correspond to an orbital angle measured relative to an antero-posterior axis 53 in a transverse plane 51 of the patient 50. The transverse plane 51 is defined by the target point, the antero-posterior axis 53 and a transverse axis 54 of the patient. The angle [3 may correspond to a cranio-caudal angle formed by the transverse plane 51 and an oblique plane 52. The oblique plane 52 is defined by the target point, the transverse axis 54, and the candidate trajectory 55 formed by the target point and the entry point considered.
[0071] [Fig.4] schematically represents a set of candidate entry points 22 in the anatomical model at the level of the patient's body envelope 21 (the candidate entry points 22 form a "mesh" at the level of the patient's body envelope 21).
[0072] As indicated previously, it is advantageous to reduce the number of candidate trajectories to limit the overall calculation time of the method of selecting a tra optimal jectory. For this purpose, and as illustrated in [Fig.2], the method 100 comprises a step 104 of eliminating certain candidate trajectories according to one or more predetermined validity criteria.
[0073] It is advantageous to consider several different validity criteria and to evaluate them according to a predefined order. This order is defined according to an estimated calculation time necessary to evaluate each validity criterion.
[0074] This involves applying successive simple filters in order to reduce the set of candidate trajectories. These filters are simple to calculate and make it possible to quickly reject certain trajectories that are impossible to produce. These filters are applied from the simplest to the most complex, in terms of calculation time, so that the last filters are applied to the smallest possible number of candidate trajectories.
[0075] By way of non-limiting example, the following validity criteria can be successively taken into account.
[0076] Initially, if the medical intervention is assisted by a robot comprising a robotic arm equipped with a medical instrument, it is possible to eliminate candidate trajectories for which it is not possible to configure the robotic arm so that the medical instrument can follow the trajectory considered. Once the sampling has been carried out, the candidate trajectories are filtered using a priori knowledge of the robot's accessibility capabilities for the orbital and craniocaudal angles and for the target point considered. This step is dependent on the robot model used, the geometry of the tools used to guide the medical instrument, and the positioning of the target point. These parameters are known when the candidate trajectories are determined, and it is therefore possible to greatly reduce the number of candidate trajectories.
[0077] In a second step, it is possible to eliminate candidate trajectories for which the length of the trajectory is greater than the length of the medical instrument envisaged for the intervention.
[0078] In a third step, it is possible to eliminate candidate trajectories for which an object affixed to the patient is an obstacle. As illustrated in [Fig. 4], this may for example be a patient reference 23 used by an optical navigation system to guide the robotic arm. According to another example, it could also be a catheter. Using the segmentation of the patient and the objects affixed to it for the intervention, and knowing a priori the three-dimensional geometry of the robotic arm, it is possible to determine trajectories leading to collisions between the robotic arm and objects affixed to the patient for the intervention. Such trajectories can be eliminated.
[0079] In a fourth step, it is possible to eliminate the candidate trajectories for which there is an intersection of the candidate trajectory with at least one critical anatomical structure internal to the patient's body. By critical anatomical structure we mean, for example, an organ different from the anatomy of interest (for example, the lungs, the spleen, the gallbladder, or the kidneys if the anatomy of interest is the liver) or vascular structures at risk (arteries, veins, bile ducts, digestive tract). In general, segmentations are not represented parametrically, so to detect a collision between an anatomical structure at risk and the trajectory, it is possible to finely sample it and recover the segmentation value for each point of the sample. A distance between points on the trajectory equal to half the resolution of the image in the acquisition direction, which is typically of the order of a millimeter, is sufficient for this purpose.
[0080] The validity criteria listed above by way of example can be applied both to the case of a minimally invasive intervention on a soft organ and to the case of a minimally invasive intervention on a bone.
[0081] As illustrated in [Fig.2], the method 100 comprises a step 105 of calculating at least one cost function, for each remaining candidate trajectory (i.e. for each candidate trajectory which has not been eliminated in the elimination step 104) from several characteristics relating for example to the safety or performance of the intervention. These characteristics are quantified; this means that a measured and normalized value is attributed to each characteristic considered (in other words, each quantified characteristic corresponds to a metric representative of a safety or performance criterion).
[0082] By way of non-limiting example, all or part of the following characteristics may be taken into account for the calculation of a cost function. Unless otherwise indicated, each of these characteristics may be taken into account for a minimally invasive intervention on a soft organ or for a minimally invasive intervention on a bone.
[0083] According to a first example, it is possible to take into account a minimum distance between the candidate trajectory and a critical anatomical structure internal to the patient's body. For this, the anatomical model may include a segmentation of the internal critical anatomical structures. A distance between each point of the trajectory and a given anatomical structure can then be calculated using a distance transform (also called a distance map) of the associated segmentation. When this transformation is applied to a binary image, the result is a new image where the value of each voxel corresponds to the shortest distance between this voxel and the contour of the image. The minimum distance between the candidate trajectory and an internal critical anatomical structure corresponds to a safety margin which must take into account the inaccuracy factors of the medical robot and / or the biomechanical hazards during insertion of the medical instrument.
[0084] According to another example, it is possible to consider a length of the intersection between the candidate trajectory and the anatomy of interest. This characteristic is linked to a safety criterion of the medical procedure. Indeed, a minimum distance crossed by the medical instrument in the anatomy of interest may be required to avoid spreading tumor cells outside the anatomy of interest during its removal. This characteristic is also linked to a performance criterion because the greater the length of the intersection between the candidate trajectory and the anatomy of interest, the better the stability of the medical instrument in the anatomy of interest during the intervention. There is therefore a compromise to be found depending on the clinical objectives; this compromise may be specific to each intervention.In the case where the minimally invasive intervention aims to insert a screw into a bone, the minimum distance crossed by the screw in the bone is also relevant with regard to the stability of the screw in the bone.
[0085] According to another example, it is possible to take into account a minimum angle of incidence between the candidate trajectory and an anatomical interface crossed by the candidate trajectory. When inserting a medical instrument, for example a needle, for a minimally invasive procedure, the instrument may cross a certain number of interfaces, such as the skin or the walls of certain organs (for example the liver capsule). The angle of incidence of the instrument with respect to these interfaces may have an impact on the performance of the placement of the instrument (potential deflection of the instrument or displacement of the organ). This parameter also has impacts on safety since an angle of incidence too tangent to an anatomical interface may promote the formation of a hematoma. The calculation of the angle of incidence requires the normal to the surface at the entry point of the trajectory.There are approaches based on surface meshing but they can be expensive, which penalizes their use in a clinical context. To reduce the computation time, the normal can be calculated as the gradient of the distance transform of the segmentation evaluated at the input point. It is advisable to calculate the gradient by convolution with the derivative of a Gaussian in order to obtain estimates of the normal more robust to noise (sensitivity to noise is a disadvantage of finite difference methods). In the case of an intervention on a bone, too small an angle of incidence can increase a risk of slippage of the medical instrument on the bone.
[0086] According to another example, it is possible to take into account an estimated value representative of the stability, under the effect of its own weight, of a medical instrument positioned up to the target point according to the candidate trajectory. This value can in particular be calculated as a function of the insertion length of the trajectory considered: the shorter the insertion length, the more the medical instrument will be able to move easily under the effect of its own weight or external forces. This value can also be calculated using a biomechanical model providing information, for example, on the viscoelasticity of the organs crossed. Trajectories with the least risk of causing the medical instrument to move under its own weight will be favored. This characteristic is more relevant for a soft organ than for a bone.
[0087] According to another example, when the minimally invasive intervention corresponds to an ablation, it is possible to take into account a minimum distance between the envelope of the region to be treated and the envelope of an ablation region estimated for the candidate trajectory. This minimum distance corresponds to an estimated ablation margin. For a given medical instrument, ablation zone models are available. These models make it possible to estimate the theoretical coverage area obtained for each trajectory. It is for example advantageous to optimize the coverage of a tumor zone by ablation (maximize the margin between the ablated zone and the tumor zone) while minimizing damage to the healthy parenchyma and nearby organs. Generally, the ablation zone models are ellipsoidal and eccentric relative to the distal end of the instrument.The shape and orientation of the tumor are important aspects to consider (the tumor is not necessarily spherical). Therefore, in order to obtain better coverage of the tumor, certain trajectories can be favored depending on their orientation and the position of the distal end of the instrument at the end of the trajectory.
[0088] According to another example, it is also possible to take into account an angular deviation between the candidate trajectory and a major axis of the region to be treated. Indeed, in practice, for the ablation of a tumor, it is advisable to attack the tumor along its major axis because the effect of the treatment is more effective in the axis of the medical instrument.
[0089] According to another example, when the minimally invasive intervention corresponds to an ablation, it is also possible to take into account a coverage value of the region to be treated. This coverage value can be estimated based on a segmentation of the area to be treated and an estimation of an ablation zone likely to be obtained with the candidate trajectory considered. Minimum ablation margins may indeed be required to minimize the recurrence rate, for example in the case of the ablation of a tumor. In the case of a bone tumor, another important factor is that the ablation zone remains inside the bone (the coverage constraints are then dictated by the anatomy and the position of the lesion).
[0090] According to another example, when the anatomy of interest is a bone, it is also possible to take into account a minimum distance between a part of the candidate trajectory and the cortical wall of the bone for the part of the trajectory which is inside of the bone. It is indeed advantageous to maximize the minimum distance between the part of the candidate trajectory that is inside the bone and the cortical wall of the bone in order to minimize the risk that the medical instrument, after its insertion into the bone, will touch the cortical wall of the bone again. The actual trajectory followed by the medical instrument may indeed differ from a selected theoretical trajectory due to inaccuracies in the placement of the medical instrument or biomechanical effects (deflection of the medical instrument, displacement of the anatomy of interest due to the patient's breathing or due to effects related to the insertion of the instrument, etc.). Maximizing the minimum distance between the part of the candidate trajectory that is inside the bone and the cortical wall of the bone amounts to maximizing a safety margin for the candidate trajectory.
[0091] [Fig. 8] schematically represents a bone 80 to be treated with the cortical wall 81 of the bone and the spongy part 82 of the bone. Reference 83 represents a target point that a medical instrument must reach to treat the bone 80. [Fig. 8] also represents two candidate trajectories 31-1 and 31-2 to reach the target point 83 respectively from two candidate entry points 22-1 and 22-2 at the body surface 21 of the patient. In the example illustrated in [Fig.8], candidate trajectory 31-1 could be preferred to candidate trajectory 31-2 because the minimum distance di which separates the part of candidate trajectory 31-1 which is inside bone 80 and the cortical wall 81 of the bone is greater than the minimum distance d2 which separates the part of candidate trajectory 31-2 which is inside bone 80 and the cortical wall 81 of the bone.
[0092] According to another example, when the anatomy of interest is a bone, it is also possible to take into account an angle of incidence between the candidate trajectory and the wall of the bone. It is indeed advantageous for the candidate trajectory to have an angle of incidence as close as possible to a right angle (90° angle) to prevent the medical instrument from sliding on the wall of the bone during its insertion (this could damage the bone during insertion of the medical instrument and / or cause the medical instrument to deviate from the selected trajectory). The term "angle of incidence" means the angle formed by the candidate trajectory and the bone at the interface where the trajectory penetrates inside the bone.
[0093] In the example illustrated in [Fig.8], candidate trajectory 31-1 could be preferred to candidate trajectory 31-2 because the angle of incidence ii formed by candidate trajectory 31-1 and the wall of bone 80 is closer to a right angle than the angle of incidence i2 formed by candidate trajectory 31-1 and the wall of bone 80.
[0094] According to another example, when the anatomy of interest is a bone, it is also possible to take into account an estimate of a risk of fracture of the bone following an intervention according to the candidate trajectory considered. For example, when the intervention aims to place a screw in a bone to consolidate it, it is possible to use a biomechanical model capable of modeling the structure of the bone (with solid areas and more fragile areas of the bone) as well as the mechanical stresses undergone by the bone (for example based on patient measurements) in order to estimate a risk of bone fracture following the addition of the screw according to the candidate trajectory considered. This risk can possibly be compared to a risk of bone fracture without intervention (i.e. without the consolidation screw), in order to estimate the reduction in the risk of fracture brought by the intervention. The risk of fracture can also be determined from a database of past interventions including trajectories that have prevented a bone fracture and trajectories that have not prevented bone fracture.
[0095] The characteristics of a trajectory described above can be quantified and used to construct a cost function for classifying trajectories according to an objective criterion. Several mathematical functions can be considered to combine the characteristics in order to properly represent the clinical objectives. The simplest function is a linear combination, for which the cost function C(T) of a trajectory T takes the form of a weighted sum of the characteristics described above:
[0096] [Math.4] C(J) =
[0097] where the weights w; make it possible to control the “balance of power” of the characteristics y;.
[0098] Different sets of weights w; with predefined values can form different cost functions that can be used as such if they prove suitable for the clinical practice of the practitioner. The practitioner also has the possibility of creating his own cost function by modifying the values of the weights Wi in order to give more or less importance to the characteristics y;, depending on the objectives targeted for the intervention considered. This can be done manually, for example through a user interface or automatically from previous interventions. Indeed, for each of the previous interventions, an evaluation (a score) could be assigned to the trajectory used and the characteristics calculated.From this information a system of equations (with a number N of equations significantly larger than the number K of features considered) can be created and solved by regression methods. Neural networks could also be used to generate a cost function.
[0099] It may be interesting to use different cost functions representative of different optimization categories (for example “performance” or “security”). This makes it possible to take into account criteria predefined or personalized by the user. A user can then provide an indication, via a user interface, of the cost function to be considered for displaying the candidate trajectories on the display screen 13.
[0100] As illustrated in [Fig.2], the method 100 comprises a display step 106, on the display screen 13, superimposed on the anatomical model, of at least part of the remaining candidate trajectories with, for each candidate trajectory displayed, a means of rapid visualization of a value of the cost function.
[0101] It should be noted that the display screen 13 can take different forms: it can be, for example, a flat liquid crystal screen of the computer screen type (LCD panel), but it can also be the screen of an augmented reality headset, in which case the information displayed can be projected directly onto the patient.
[0102] It is possible to not display candidate trajectories for which the cost function is less than a threshold value.
[0103] In particular embodiments, the means for quickly viewing the value of the cost function of a displayed candidate trajectory corresponds to a color code: different colors are respectively associated with different values of the cost function. Such a display makes it possible to quickly identify the best trajectory according to established clinical criteria and to select it using the user interface. When a trajectory is selected by the user, it is possible to display the quantifications determined for the characteristics of this trajectory in order to provide the user with a detailed view of each characteristic used to calculate the cost function of the trajectory. Of course, the system offers the possibility of choosing any candidate trajectory, which makes it possible to manage optimality criteria not captured by the cost function.
[0104] [Fig. 5] schematically illustrates the display of a set of candidate trajectories 31 on the display screen 13. Each candidate trajectory 31 is represented with a color associated with the value of the cost function calculated for the candidate trajectory 31 considered. The color code 30 used is also displayed. In the example illustrated in [Fig. 5], the candidate trajectories 31 are represented at the level of the body surface 21 of the patient. When the user selects one of these candidate trajectories 31, it is possible to display different views of the selected trajectory in the anatomical model (for example according to different section planes comprising the selected trajectory).
[0105] As illustrated in [Fig.2], the method 100 comprises a step 107 of selection, by a user, and for the target point considered, of a trajectory considered to be optimal among the trajectories displayed.
[0106] The selected optimal trajectory can for example be used to configure the movement of a robotic arm adapted to hold or guide the medical instrument. intended to be used to perform the medical procedure.
[0107] All of the steps of determining 103 a region and a sampling resolution, eliminating 104 certain candidate trajectories, calculating 105 at least one cost function, displaying 106 the candidate trajectories and selecting 107 a trajectory may be iterated several times to refine the selection of the candidate trajectory that will be used to perform the medical intervention. For this purpose, and as illustrated in [Fig.2] in step 108, the user may for example be invited to indicate, via a user interface, whether he wishes to refine the selected candidate trajectory.
[0108] Thus, in some embodiments, at least two iterations are performed. Advantageously, for a given iteration (different from the first iteration), the new sampling region includes the entry point corresponding to the trajectory selected at the previous iteration, the new sampling region is determined to be smaller than the sampling region determined for the previous iteration, and the sampling resolution is determined to be finer than the sampling resolution determined for the iteration (this means that the minimum distance separating two candidate entry points is smaller).
[0109] In certain implementations, for a given iteration (different from the first iteration), the new sampling resolution and / or the new sampling region is determined as a function of a variability of the cost function calculated for the previous iteration in the vicinity of the trajectory selected at the previous iteration. The variability of the cost function can for example be measured as a function of the angles α and β. The greater the variability at the previous iteration, the finer the resolution should be chosen for the new iteration. The lower the variability, the more extensive the sampling region should be chosen for the new iteration.
[0110] As illustrated in [Fig.6], the method 100 may also comprise a step 110 of evaluation by the user of at least one displayed candidate trajectory, and a step 111 of updating the cost function from the evaluation obtained.
[0111] For each trajectory displayed, the user can evaluate the trajectory by assigning it a score. This score can be used to trigger a possible update of the parameters of the cost function (for example, if the cost function is a linear combination of several quantified characteristics, the weights of the linear combination can be modified in order to adapt to the score provided by the user, via, for example, regression algorithms or neural networks).
[0112] The trajectory that was actually followed by the medical instrument during the medical intervention may deviate from the selected candidate trajectory, in particular due to inaccuracies in needle placement or biomechanical effects (deflection of the needle, displacement of the anatomy of interest due to breathing or due to effects related to insertion). It is therefore relevant to determine the a posteriori cost function for the actual trajectory followed by the medical instrument, as well as a difference AC(T) between the value C(T) of the cost function calculated for the candidate trajectory selected for the intervention and the value C'(T) of the cost function of the actual trajectory.
[0113] Thus, and as illustrated in [Fig.7], the method 100 can also comprise the following steps, after the medical intervention: - a determination 120 in the anatomical model of a real trajectory which was actually followed by the medical instrument during the intervention (this real trajectory can in particular be determined from medical images acquired on the patient at a time when the medical instrument is in place, that is to say when it has reached the target point and is in position to carry out the treatment); - a determination 121 of a value of the cost function for the actual trajectory; - a calculation 122 of a difference between the value of the cost function of the candidate trajectory selected for the intervention and the value of the cost function of the actual trajectory.
[0114] Calculating the value C'(T) of the cost function of the actual trajectory requires an estimation of the position of the medical instrument after insertion from a control image. This can be done manually (by clicking on the entry point and on the end of the instrument) or automatically using image processing methods to segment the instrument and then extract the required information.
[0115] The a posteriori determination of the cost function of the actual trajectory can in particular be used as a means of alerting the user to the poor performance of the gesture (the placement of the instrument is of lower quality than planned). This alert can be triggered when the variation AC(T) of the cost function is significant. A significance threshold can be determined from a retrospective analysis of the trajectories performed in the past for which an evaluation criterion (a score) has been assigned.
[0116] Thus, and as illustrated in [Fig.7], the method 100 may comprise a step 123 of displaying, on the display screen 13, an indication relating to the value of the difference calculated in step 122.
[0117] A retrospective analysis of the statistics of the variation AC(T) of the cost function for the trajectories whose realization is in accordance with the planning can be carried out in order to determine the typical values of AC(T) in order to determine trajectory equivalence classes. Indeed, for a set of trajectories whose realization has been judged to be in accordance with the planning, a statistic of AC(T) can be derived in order to determine what is the typical variation of the cost function. This value can then be used in the method 100 of selecting an optimal trajectory in order to avoid proposing to a user trajectories which would be too similar (i.e. trajectories whose cost function does not differ sufficiently).
[0118] Thus, and as illustrated in [Fig.7], the method 100 may comprise the following steps: - obtaining 124 an evaluation, by the user, of the actual trajectory; - a determination 125, based on the calculated gap and the evaluation obtained from a statistical value representative of the variability of the cost function; - an elimination 126 of the candidate trajectories considered as equivalent according to the statistical value thus obtained.
[0119] In the method 100 for assisting in selecting an optimal trajectory, steps 101 to 108, 110 and 111 described with reference to FIGS. 2 and 6 are all implemented before the medical intervention; steps 120 to 126 described with reference to [Fig. 7] are implemented after the medical intervention. These steps therefore do not involve the execution of a surgical procedure. Steps 108, 110, 111 and 120 to 126 are optional.
[0120] The data processing device 10 described above makes it possible to propose to a practitioner a set of optimal or near-optimal trajectories for a minimally invasive intervention from a three-dimensional anatomical model of the patient. The proposed trajectories take into account constraints of different types (for example practical constraints, performance constraints and / or safety constraints). The practitioner has the possibility of choosing a trajectory that may be suboptimal in the sense of a predefined criterion, but preferable for constraints not captured by this criterion.
[0121] The proposed solution allows a rapid and optimized exploration of all the solutions to determine candidate trajectories (“multi-resolution” approach for sampling candidate entry points). It allows a rapid visualization by the user of different candidate trajectories as well as their optimality criteria. The proposed solution is flexible since it allows adaptation to specific needs of the intervention (increased need for placement precision, increased need for safety margins linked to a potential weakness in the patient’s condition, etc.). The practitioner also has the possibility of updating optimality criteria (cost function) based on user feedback (evaluation of a se- selected for the intervention).
[0122] The above description has been made mainly by considering a case with a single target point (for example for a “single-needle” intervention), or by considering a case where several target points are considered successively in order to determine an optimal trajectory individually for each of the target points (for example for a “multi-needle” intervention). In the case of an intervention where several medical instruments must be inserted (for example for a “multi-needle” intervention), it is however also possible to consider all of the target points in a grouped manner (and therefore to consider groups of candidate trajectories associated respectively with the different target points). This can in particular make it possible to take into account in the calculation of the cost function characteristics which depend on the candidate trajectories in relation to each other.For example, this can allow a maximum relative angle between the different two-by-two trajectories to be taken into account in the calculation of the cost function (it may be advantageous to minimize this relative angle). According to another example, this can allow a deviation between a distance separating the different two-by-two trajectories from a target value to be taken into account (the target value may in particular come from recommendations of the manufacturer of the medical instrument used).
Claims
Claims
1. A data processing device (10) comprising a processor (11), a computer memory (12) and a display screen (13), the computer memory (12) comprises program code instructions which, when executed by the processor (11), configure the processor (11) to implement a method (100) for assisting in the selection of at least one optimal trajectory for a minimally invasive medical intervention for treating a bone of a patient, the processor (11) is configured to: - obtaining (101) a three-dimensional anatomical model of the patient from medical images previously acquired on the patient, the anatomical model comprising a representation of the body envelope (21) and the bone to be treated, - identify (102) in the anatomical model at least one target point to be reached in the bone to be treated, - perform at least one iteration for which the processor is configured to: • determining (103) a sampling region and a sampling resolution for sampling candidate entry points (22) in the anatomical model at the body envelope (21) of the patient, each candidate entry point (22) belonging to the sampling region, the sampling resolution being representative of a minimum distance separating two candidate entry points (22), each pair formed by the target point and a candidate entry point (22) forming a candidate trajectory, • eliminate (104) candidate trajectories based on at least one predetermined validity criterion, • for each remaining candidate trajectory, quantifying each of a plurality of predetermined characteristics, including a minimum distance between a portion of the candidate trajectory located inside the bone to be treated and a cortical wall of the bone, and calculating (105) at least one cost function from the quantified characteristics, display (106) on the display screen (13), superimposed on the anatomical model, at least part of the remaining candidate trajectories with, for each candidate trajectory (31) displayed, a means of rapid visualization of a value of the cost function, obtaining (107) the selection, by a user, for said at least one target point, of a trajectory considered optimal among the candidate trajectories (31) displayed.
2. Device (10) according to claim 1 wherein the processor (11) is configured to perform at least two iterations, and wherein, for an iteration of rank n, n being an integer strictly greater than one: - the sampling region of the iteration of rank n is determined to be smaller than the sampling region determined for the iteration of rank (n-1), and to include the candidate entry point (22) corresponding to the trajectory selected at the iteration of rank (n-1), and - the sampling resolution of the iteration of rank n is determined to be finer than the sampling resolution determined for the iteration of rank (n-1).
3. Device (10) according to claim 2 wherein the sampling resolution of the iteration of rank n is determined as a function of a variability of the cost function calculated for the iteration of rank (n-1) in the vicinity of the trajectory selected at the iteration of rank (n-1).
4. Device (10) according to any one of claims 2 to 3 wherein the sampling region of the iteration of rank n is determined as a function of a variability of the cost function calculated for the iteration of rank (n-1) in the vicinity of the trajectory selected at the iteration of rank (n-1).
5. Device (10) according to any one of claims 1 to 4 wherein, to sample the candidate input points (22), the processor (11) is configured to determine in the anatomical model, at the level of the body envelope (21) of the patient, a set of points defined by a spherical coordinate system centered on the target point, each point being defined by a distance r^ from the target point and by two angles a; and [3j, the angles a; and [3j being defined such that, for indices i and j corresponding to strictly positive integers: n ■ ■ — n - < • + — and where R is representative of a minimum distance separating two candidate entry points (22) at the level of the body envelope (21) of the patient.
6. Device (10) according to any one of claims 1 to 5 in which a validity criterion makes it possible to verify, for a given candidate trajectory, at least one of the following elements: - a validity of the length of the candidate trajectory in relation to a medical instrument envisaged for the intervention, - a possibility of configuring a robotic arm equipped with a medical instrument so that said medical instrument can follow the candidate trajectory, - a presence of objects affixed to the patient and obstructing the candidate trajectory, - an intersection of the candidate trajectory with at least one critical anatomical structure internal to the patient's body.
7. Device (10) according to any one of claims 1 to 6 in which several validity criteria are considered to eliminate candidate trajectories, and the different validity criteria are evaluated according to an order defined as a function, for each validity criterion, of an estimated calculation time necessary to evaluate said validity criterion.
8. A device (10) according to any one of claims 1 to 7 wherein the plurality of characteristics of a candidate trajectory comprises at least one of the following characteristics: - a minimum distance between the candidate trajectory and a critical anatomical structure internal to the patient's body, - a length of the intersection between the candidate trajectory and the bone to be treated, - a minimum angle of incidence between the candidate trajectory and an anatomical interface crossed by the candidate trajectory, - a minimum distance between an envelope of the region to be treated and an envelope of an ablation region estimated for the candidate trajectory, - an angular difference between the candidate trajectory and a major axis of the region to be treated.
9. Device (10) according to any one of claims 1 to 8 in which several cost functions are calculated from the quantified characteristics, each cost function being calculated according to a different set of weightings attributed respectively to the different quantified characteristics, and the processor (11) is configured to obtain an indication, by a user, of the cost function to be considered for the display of the candidate trajectories.
10. Device (10) according to any one of claims 1 to 9 in which the means for quickly displaying the value of the cost function for each candidate trajectory (31) displayed comprises a color code associating different colors respectively with different values of the cost function.
11. Device (10) according to any one of claims 1 to 10 in which the processor (11) is configured to display the quantifications determined for the characteristics of the trajectory selected by the user.
12. Device (10) according to any one of claims 1 to 11 wherein the processor (11) is configured to obtain (110), for at least one displayed candidate trajectory (31), an evaluation by the user of said candidate trajectory, and to update (111) the cost function according to said evaluation.
13. Device (10) according to any one of claims 1 to 12 in which the processor (11) is configured to: - determine (120) in the anatomical model, after the intervention, a real trajectory followed by a medical instrument during the intervention, from medical images acquired on the patient at a time when the medical instrument is in place, - determine (121) a value of the cost function for the actual trajectory, - calculate (122) a difference between the value of the cost function of the candidate trajectory selected for the intervention and the value of the cost function of the actual trajectory.
14. Device (10) according to claim 13 in which the processor (11) is configured to compare the calculated deviation with a predetermined threshold and display (123) an indication relating to the value of the calculated deviation.
15. Device (10) according to claim 13 wherein the processor is configured to: - obtain (124) an evaluation of the actual trajectory by the user, - determine (125), based on the calculated deviation and the evaluation obtained, a statistical value representative of the variability of the cost function, - eliminate (126) candidate trajectories considered equivalent according to the statistical value thus obtained.
