Computer-assisted method for planning placement of a femoral tunnel for ligament reconstruction

Abstract

The invention relates to a computer-assisted method for planning placement of a femoral tunnel for ligament reconstruction, comprising: - obtaining a 3D model of the patient's femur; - based on said 3D model, determining a set of anatomical features of the patient's femur; - based on said set of anatomical features, positioning a coordinate system on the 3D model of the patient's femur, - based on the 3D model of the patient's femur, computing an isometry map defining a length variation of the ligament during knee flexion depending on a position of the femoral tunnel, the isometry map comprising (i) a laxity region wherein the length variation is negative, the ligament contracting during knee flexion, (ii) a tension region wherein the length variation is positive, the ligament tightening during knee flexion, and (iii) an isometry line separating the laxity region and the tension region wherein the length variation is null, the ligament undergoing no deformation during knee flexion; - obtaining at least one first input (2000-2008) related to a desired graft position in the coordinate system (2040) and at least one second input (2010-2014) related to a desired length variation in the isometry map (2050); and - automatically computing a proposed position of the femoral tunnel by combining said desired graft position and desired length variation.

Description

[0001] COMPUTER-ASSISTED METHOD FOR PLANNING PLACEMENT OF A FEMORAL TUNNEL FOR LIGAMENT RECONSTRUCTION

[0002] TECHNICAL FIELD

[0003] The invention relates to a computer-assisted method for planning placement of a femoral tunnel for ligament reconstruction and to a computer system configured to implement said planning method. The invention also relates to a navigated surgical system for ligament reconstruction.

[0004] TECHNICAL BACKGROUND

[0005] The anterior cruciate ligament (ACL) is located between the inner medial face of the distal lateral femoral condyle and the medial horn of the tibial plateau.

[0006] FIG. 1A is a sagittal view of a knee with a femur F, a tibia T and an anterior cruciate ligament ACL connecting the distal lateral condyle of the femur F and the plateau of the tibia T. FIG. 1B is a posterior view of said knee.

[0007] The main function of the ACL is to guarantee stability of the knee during anterior tibial translation and internal tibial rotation.

[0008] In case of rupture or damage of the ACL, this ligament can be replaced by a surgery called ACL reconstruction (ACLr). ACLr requires drilling of femoral and tibial tunnels and insertion of a graft in both tunnels. As shown in FIG. 1 C, the femoral tunnel FT and the tibial tunnel TT are substantially cylindrical holes, and the graft G is tied to the femur and the tibia through said tunnels.

[0009] However, such reconstruction often leads to failures. In particular, several medical articles report that more than 50% of ACLr surgeries failures account for femoral and / or tibial tunnel malposition. Although the ideal tunnel positioning for ACLr is still a matter of debate and research, there is a certain consensus pointing out that attempting to reconstruct the patient’s native anatomy is the best approach.

[0010] Historically, open surgery allowed the surgeons to identify with precision the remnants of the ruptured ligament.

[0011] The actual tendency to execute minimally invasive surgeries with an arthroscopic approach has the advantage of reducing significantly the healing time and the post-operative complication rate. However, the procedures are often more complex and difficult, leading to inaccurate tunnel positioning.

[0012] A common cause of ACL reconstruction failure is the inaccurate femoral tunnel placement. It has been observed that among the technical errors leading to an ACL graft misplacement, 80% of these errors come from improper tunnel placement. An anteriorly oriented graft will not follow the obliqueness of the ACL and therefore has a limited ability to restore normal knee stability. Cadaver studies suggest that placing the graft closer to the centerof the attachment might improve knee kinematics compared with an anterior graft placement, also by limiting the long-term development of osteoarthritis after ACL reconstruction.

[0013] One of the gold-standard methods in the scientific literature to identify the ACL femoral tunnel position post-operatively (and more recently, pre- and per-operatively) is the Bernard & Hertel (BH) grid, also referred as the Quadrant method (Sullivan et al., 2015). The quadrant method is done by obtaining a strict mid-plane sagittal view of the distal femur with the femoral condyles superimposed. A line is drawn along the intercondylar roof line with a parallel line drawn tangent to the distal lateral femoral condyle. Perpendicular lines are then drawn along the deep and shallow borders of the condyle with the proximal line (extended intercondylar roof line) crossing the condyles. Once the grid is drawn, the origin is considered the deep-high corner, by which the anatomical ACL footprint is designed as a ratio of the depth and height of the quadrant. In the above-mentioned article, the authors determined, based on their anatomical study, that the average ACL femoral footprint is located at 24.8% of quadrant length (from posterior condyle line) and at 28.5% of quadrant height (from the intercondylar roof line).

[0014] Using arthroscopic instruments to determine the ACL insertion point to place the tunnel (regardless of the drilling technique) is one of the standard techniques used nowadays. Nevertheless, depending on the entry technique, it remains challenging to accurately estimate the location of the insertion point. Several studies have shown that the apex of the deep cartilage (ADC) is a reliable landmark to locate the ACL footprint (Hart et al., 2015).

[0015] Accessible from different arthroscopic methods, it has been found that the centroid of the ACL footprint is found 3 mm (range: 1-4 mm) higher and 12 mm (range: 11-17 mm) shallower from the ADC (Hart et al., 2015). The main disadvantage is that these measurements apparently have to be scaled depending on the size of the femur, but this information is not very well documented.

[0016] Another way to identify the ACL footprint is to use magnetic resonance imaging (MRI) in a pre-operative planning strategy. In a recent work, two ACL tunnel techniques were opposed and compared: the standard technique using an arthroscopic method and a roadmap technique with the MRI pre-planning. Despite the use of the 3D MRI to depict the ACL anatomy, there was no significative improvement between the standard and the roadmap technique (Marwan et al., 2020). This might be explained by the fact that segmenting the ACL using MRI with its oblique shape is challenging. In another study, three groups of specialists (residents in orthopedy, orthopedic surgeons and radiologists) assessed the footprint of torn ACLs. The mean intraobserver variability was 3.82mm, and the mean interobserver variability was 8.67mm (Zee et al., 2021). Such large values might help conclude that the use of imaging strategies remains a challenging task to identify the ACL footprint.

[0017] Thus, there remains a need for a method to provide an automatic and optimal planning for the placement of the femoral tunnel which reduces the risks of failure of ACLr surgery.Similarly, the same need remains for other knee ligaments, i.e. the medial collateral ligament (MCL), the lateral collateral ligament (LCL) and the posterior cruciate ligament (PCL).

[0018] SUMMARY OF THE DISCLOSURE

[0019] A goal of the invention is to define a method for planning automatically the position of the femoral tunnel in knee ligament reconstruction to minimize the risk of failure of the reconstruction.

[0020] To that end, the invention provides a computer-assisted method for planning placement of a femoral tunnel for ligament reconstruction, comprising:

[0021] - obtaining a 3D model of the patient’s femur;

[0022] - based on said 3D model, determining a set of anatomical features of the patient’s femur; - based on said set of anatomical features, positioning a coordinate system on the 3D model of the patient’s femur,

[0023] - based on the 3D model of the patient’s femur, computing an isometry map defining a length variation of the ligament during knee flexion depending on a position of the femoral tunnel, the isometry map comprising (i) a laxity region wherein the length variation is negative, the ligament contracting during knee flexion, (ii) a tension region wherein the length variation is positive, the ligament tightening during knee flexion, and (iii) an isometry line separating the laxity region and the tension region wherein the length variation is null, the ligament undergoing no deformation during knee flexion;

[0024] - obtaining at least one first input related to a desired graft position in the coordinate system and at least one second input related to a desired length variation in the isometry map; and

[0025] - automatically computing a proposed position of the femoral tunnel by combining said desired graft position and desired length variation.

[0026] Thus, the method can determine an optimal graft insertion point position based on the geometric analysis of the distal femur and the aggregation of various inputs from the user, including scientific data and / or databases of previous successful surgeries. This allows to take into account several hints of the optimal placement of the femoral tunnel entry point automatically.

[0027] According to advantageous features of the method, which may be combined if appropriate:

[0028] - the method further comprises obtaining at least one third input related to a size of a graft to be implanted in the femoral tunnel and combining said third input with the desired graft position and desired length variation to automatically compute the proposed position of the femoral tunnel;- the method further comprises obtaining at least one fourth input related to a zone of the femur to be avoided and combining said fourth input with the desired graft position and desired length variation to automatically compute the proposed position of the femoral tunnel;

[0029] - the method further comprises, after computing the proposed position of the femoral tunnel, requesting and receiving a fifth input;

[0030] - the fifth input may be a validation of the proposed position of the femoral tunnel;

[0031] - alternatively, the fifth input may be a requested adjustment of the proposed position, the method comprising automatically computing an adjusted proposed position of the femoral tunnel based on the fifth input.

[0032] The first input comprises at least one of:

[0033] - a position selected by a user either on the 3D model of the femur or directly on the femur;

[0034] - a remnant segmented zone;

[0035] - a statistical position from scientific literature; and

[0036] - a statistical position from a database of prior reconstruction cases.

[0037] Advantageously, at least two first inputs are weighted with a respective weight and summed to compute the desired graft position, the sum of the weights being equal to 1.

[0038] The second input comprises at least one of:

[0039] - a length variation selected by a user on the isometry map;

[0040] - a statistical length variation from scientific literature; and

[0041] - a statistical length variation from a database of prior reconstruction cases.

[0042] Advantageously, at least two second inputs are weighted with a respective weight and summed to compute the desired length variation, the sum of the weights being equal to 1.

[0043] In some embodiments, computing the proposed position of the femoral tunnel comprises applying a mathematical function to translate the desired graft position along at least one axis of the coordinate system with respect to the desired length variation.

[0044] Obtaining the 3D model of the patient’s femur may comprise:

[0045] - obtaining at least one medical image of the patient’s femur;

[0046] - segmenting said at least one medical image to determine a 3D mesh of the femur; and - registering said 3D mesh with a reference model of the femur to orient the 3D mesh in a coordinate system of said reference model, the registered 3D mesh forming the 3D model of the patient’s femur.

[0047] In some embodiments, the set of anatomical features comprises an intercondylar roof curve.

[0048] In some embodiments, the coordinate system is a Bernard & Hertel grid.

[0049] In some embodiments, computing the proposed position of the femoral tunnel comprises: - determining from the Bernard & Hertel grid and from the first input a target position of the entry point of the femoral tunnel within a given confidence interval;- determining from the isometry map and from the second input a target length variation interval of the anterior cruciate ligament; and

[0050] - computing a combination of the confidence interval for the target position of the entry point and of the target length variation interval to determine the optimal position of the femoral tunnel and an allowed area surrounding the optimal position.

[0051] In some embodiments, determining the intercondylar roof curve comprises:

[0052] - computing a projection of the 3D model of the patient’s femur in a sagittal plane so that both condyles of the femur are superposed;

[0053] - computing a mid-sagittal plane cutting the condyles comprising the intercondylar roof curve;

[0054] - from said projection, determining a set of points oriented along the mid-sagittal plane; - from said determined set of points, extracting points belonging to the intercondylar roof curve; and

[0055] - computing a line connecting the extracted points to draw the intercondylar roof curve. Extracting the points belonging to the intercondylar roof curve may comprise:

[0056] - computing a weight value of each point of the determined set of points to the intercondylar roof curve; and

[0057] - extracting the points having a weight value greater than a determined threshold.

[0058] Advantageously, the method further comprises displaying the 3D model of the patient’s femur and at least one of: the coordinate system, the isometry map and the proposed position of the femoral tunnel.

[0059] The method may further comprise displaying a graphical user interface comprising at least one user interface element configured to allow a user to provide the fifth input.

[0060] The user interface element is configured to allow the user to modify at least one weight or parameter of the mathematical function.

[0061] In some embodiments, updating the proposed position of the femoral tunnel comprises displacing the proposed position within the allowed area.

[0062] The ligament may be an anterior cruciate ligament (ACL), a medial collateral ligament (MCL), a lateral collateral ligament (LCL), a posterior cruciate ligament (PCL) or an anterolateral ligament (ALL).

[0063] Another aspect of the invention relates to a computer system for planning placement of a femoral tunnel for ligament reconstruction, comprising a control unit configured to implement the method described above, and a user interface coupled to the control unit, the user interface being configured to display the 3D model of the patient’s femur and the computed proposed position of the femoral tunnel and to receive user’s inputs.

[0064] Another aspect of the invention relates to a navigated, with or without robotics, surgical system for anterior cruciate ligament reconstruction based on the above method.

[0065] Said navigated surgical system comprises:a surgical tool comprising either a drill or a guide for guiding a drill to drill the patient’s femur,

[0066] a localization system including navigation trackers configured to be attached to the patient's femur and tibia and to the surgical tool;

[0067] the computer system as described above, wherein the control unit is coupled to the localization system, configured to:

[0068] implement the method described above to plan the proposed tunnel position and a tunnel orientation;

[0069] display the proposed tunnel position and orientation in real-time along with the surgical tool;

[0070] provide feedback to align the surgical tool with an axis of the proposed calculated tunnel placement.

[0071] Said navigated surgical system may further comprise:

[0072] a robotic arm controlled by the control unit, comprising a plurality of motorized joints and an end-effector configured to drill or to guide a drill;

[0073] the localization system further comprising a navigation tracker attached to the robotic arm to localize the end-effector in real-time.

[0074] Based on the above planning method, it is possible to implement a computer-assisted method for anterior cruciate ligament reconstruction. Said method comprises:

[0075] - planning placement of the femoral tunnel with the method described above;

[0076] - rigidly attaching at least one tracker of a localization system to the patient’s femur and to a surgical tool comprising either a drill or a guide for guiding a drill to drill the patient’s femur;

[0077] - based on the proposed position of the femoral tunnel validated by the user and on localization data of the patient’s femur relative to the robotic arm, aligning the surgical tool with the axis of the proposed calculated tunnel placement.

[0078] In some embodiments, said method can be implemented by a robotic arm holding the surgical tool.

[0079] BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Further features, embodiments and advantages will appear from the following description, based on the appended drawings, wherein:

[0081] FIG. 1A is a sagittal view of a knee, FIG. 1B is a posterior view of a knee with the original ACL and FIG. 1C is a posterior view of the knee with an ACL graft; implanted by ACL reconstruction surgery;

[0082] FIG. 2 is a block diagram illustrating an embodiment of the control unit;

[0083] FIG. 3 outlines the general steps to reach the final automatic suggestion of optimal placement of the femoral ligament graft insertion point;FIG. 4 is a schematic representation of the workflow for determining the proposed position of the femoral tunnel based on a desired position and a desired graft length change;

[0084] FIG. 5 draws out the methodology to find the intercondylar roof curve;

[0085] FIGS. 6A to 60 illustrate steps of the determination of the intercondylar roof curve; FIG. 7 shows the placement of the Bernard & Hertel grid onto the 3D model of the femur;

[0086] FIG. 8 shows the placement of a 3D cylinder grid onto the 3D model of the femur; FIG. 9 shows the placement of an example of a coordinate system for the MCL reconstruction;

[0087] FIG. 10 shows the placement of an example of a coordinate system for the PCL reconstruction;

[0088] FIG. 11 teaches the general steps to acquire a knee flexion for each of the possible femoral insertion points of the ligament until the obtention of an isometry map;

[0089] FIGS. 12A and 12B show the simulated kinematics of the patient’s knee in full extension and in flexion, and FIG. 12C shows an example of an isometry map computer from the simulation;

[0090] FIG. 13 is a representation of the visual interface that comprises user interface elements to fine-tune each of the user’s inputs for the choice of the graft position and of the change of length;

[0091] FIG. 14A and 14B show the proposed position of the femoral tunnel along with allowable ranges resulting from the fusion of the desired position data and the desired graft change of length range;

[0092] FIG. 15 is a schematic representation of the use of a gradient to adjust the position of the proposed position of the femoral tunnel;

[0093] FIG. 16A shows a navigated surgical tool while FIG. 16B shows a navigated tool guide with an attached surgical tool;

[0094] FIG. 17 illustrates the main steps of the surgery that integrates the optimal placement of the femoral insertion point into a navigated surgery system;

[0095] FIGS. 18A and 18B are schematic views of a robotic surgical system configured to assist ligament reconstruction;

[0096] FIG. 19A depicts a navigated robotic arm with a drill end effector and FIG. 19B depicts a navigated robotic arm with a passive tool guide as end effector.

[0097] DETAILED DESCRIPTION OF EMBODIMENTS

[0098] The planning method consists in the placement of the femoral tunnel for a graft ligament. The knee ligament may be the anterior cruciate ligament (ACL), the medial collateral ligament(MCL), the lateral collateral ligament (LCL), the posterior cruciate ligament (PCL) or the anterolateral ligament (ALL).

[0099] “Placement” is interpreted herein as the determination of the position of the insertion point and the determination of the orientation of the femoral tunnel for each graft to be implanted. For ACL, the insertion point is located in the inner medial face of the distal lateral femoral condyle and the femoral tunnel extends between the insertion point and the lateral outer face of the distal femur. For MCL, the insertion point is located in the medial outer face of the distal femur and the femoral tunnel extends a certain distance into the distal femur toward the lateral part of the femur. For LCL, the insertion point is located in the lateral outer face of the distal femur and the femoral tunnel extends a certain distance into the distal femur toward the medial part of the femur. For PCL, the insertion point is located in the inner lateral face of the distal medial femoral condyle and the femoral tunnel extends between the insertion point and the medial outer face of the distal femur. For ALL, the insertion point is located in the lateral outer face of the distal femur and the femoral tunnel extends a certain distance into the distal femur toward the medial part of the femur.

[0100] The planning method is carried out by a computer system which interacts with a user (typically a surgeon).

[0101] The computer system comprises at least one control unit as depicted in FIG. 2. A function of the control unit 200 is to receive at least one medical image of the patient’s femur and inputs 220 from a user and / or from an external database to plan an optimal placement of the femoral tunnel. The control unit can, for instance, comprise one or more microprocessors 201, one or more random access memories (RAM) 202 and / or one or more read-only memories (ROM) 203, and operating system 205, one or more calculators, one or more computers and / or one or more computer programs 206. The computer program(s) 206 comprise code instructions to compute the optimal placement of the femoral tunnel based on the medical image and the inputs 220. In addition, the control unit may include other devices and circuitry for performing the functions described herein such as, for example, a storage unit 204 (i.e. a hard drive), input / output circuitry, and the like. The input / output circuitry can be adapted to treat digital and / or analog signals. The control unit 200 can be linked to a network 250 which could be a local area network or a wide area network.

[0102] The system further comprises a user interface coupled with the control unit. The user interface may comprise, for example, a display 210 allowing the user to visualize the proposed placement of the femoral tunnel. The user interface may also comprise human interface devices 211 (HID), such as a mouse ora keyboard, allowing the user to enter data, for example inputs needed to compute the optimal femoral tunnel placement, or requests for correction of a proposed placement. In some embodiments, the display 210 has a tactile screen which allows the user to directly interact with the displayed content. The control unit 200 can output221 signals to control other control units or other devices. The control unit 200 can receive as input 220 signals to be controlled from other control units or other devices.

[0103] In this invention, the proposed placement of the femoral tunnel for ACL, MCL, LCL, ALL and / or PCL is obtained from methods described hereafter illustrated in FIG. 3, which schematically depicts the methodology as consecutive steps to reach the data fusion and the automatic recommendation of the optimal femoral insertion point position obtained in step 410. By “automatic” is meant in the present text that the system is configured to carry out all the steps leading to the determination of a proposed tunnel placement without requiring any intervention from the user other than inputs corresponding to a desired graft position and a desired ligament length variation. In particular, the user does not have to treat data nor to compute the position of the tunnel based on said data.

[0104] The process begins with the image acquisition of the patient’s knee in step 400. The image or the plurality of images are then segmented in order to obtain the 3D model of the patient’s knee, at least the distal femur geometry. In some embodiments, the proximal tibia geometry might be necessary. A coordinate system definition is obtained in step 406, which requires the identification of anatomical features such as landmarks in step 404. For the case of ACL reconstruction, in some embodiments, showing the Bernard & Hertel (BH) grid in step 406 requires the proper identification of the intercondylar roof in step 404. In some embodiments, showing the isometry map (step 407) requires a kinematic acquisition of a knee flexion for every femoral insertion point of the ligament graft 405. In some embodiments, a user’s choice on the source data of the isometry map in step 409 and a user’s choice on the source data of the coordinate system in step 408 data are merged to provide the output of the optimal femoral ligament insertion point position in step 410. This workflow works for the distal femoral ligaments such as ACL, PCL, MCL, LCL and ALL.

[0105] The different user’s input and how these inputs are combined are illustrated in FIG. 4. The source of the data containing said optimal placement of the femoral tunnel is based on a single or multiple choice 2098 among five options: the choice of a piece or a plurality of pieces of scientific literature 2006, the choice of a user’s manual input using a user interface 2000, a user’s manual input using a marked zone directly from a tracked palpation probe linked to a computer-aided surgical system (CASS) 2002, the choice of a femoral insertion point issued from a user’s database that may contain historical surgery data 2008 or the choice of the remnant of the ruptured ligament 2004. This selection is integrated in the methods to calculate the optimal placement of the femoral tunnel. The inputs 2098 that can be used to compute the optimal position of the femoral tunnel may be of different types, combined as a desired weighted sum of positions 2040, each input 2000-2008 associated with a weight W0a-W4a.

[0106] Further description of a first type of information 2006 comes from scientific literature, in particular papers reporting ligament reconstruction cases for the concerned ligament andcorrelating a level of success of the reconstruction and the position of the femoral tunnel. The user may decide, for each patient, which piece(s) of scientific literature to rely on.

[0107] Further description of a second type of information 2000 comes from the manual choice of the surgeon, based on his / her own experience using a user interface for this effect.

[0108] Further description of a third type of information 2008 comes from an internal proprietary database which refers to surgeries already performed with the system, considering only the current user, the users of a particular institution (e.g. a hospital or a clinical center) or all other users of the system. In some embodiments, the surgeon may have gathered data relating to surgeries he / she has already performed for the concerned ligament and the results he / she achieved in each case. In some embodiments, artificial intelligence enables the most relevant data to be used to assist the surgeon in the planning; said internal proprietary database might be issued from the pre-operative, intra-operative and post-operative data from other users using the same system described herein, and the artificial intelligence have access to said data useful to train the artificial intelligence. In some embodiments, said artificial intelligence can be declined as an artificial neural network (i.e. a Recommender system). In some embodiments, said artificial intelligence can be declined as a single or a plurality of machine learning algorithms (i.e. a Decision support system). In some embodiments, the database can be partially or fully generated (e.g. synthetic) by reconstruction cases from the own surgeon, from the own institution or publicly available using neural networks described in the prior art.

[0109] Further description of a fourth type of information 2004 comes from the segmentation of the remnants of the ruptured ligament. If a magnetic-resonance image (MRI) or an ultrasound segmentation containing the ligament tissues is available, the user can choose the remnant of the ruptured ligament as the point to position the new graft. This segmentation can produce an average center of the ligament footprint as its position, along with an average footprint area expressed as a standard deviation.

[0110] Further description of a fifth type of information 2002 comes from the manual choice of the surgeon based on his / her own experience as a direct zone selection using a tracking system. This selection can also be done using a tracked palping probe attached to the navigation system of a computer-aided surgical system (CASS) which can be used to physically palp or touch the zone where the user wishes to position the graft. This palpation can be repeated a number of times in order to model a normal distribution with an average position along with its standard deviation.

[0111] The second set of inputs consists of the desired ligament length change inputs 2099. In a similar fashion to the selection of the position of the tunnel graft, the user 2010 can choose the desired change of length on the screen using a touchscreen or an input device.

[0112] As another input, the choice can be given by a list provided by the scientific literature 2012 that can be displayed and the user can choose among the possible values.The other input consists in the choice 2014 of a possible change of length value that comes from a database of prior successful surgeries, either private to the user, to the institution or as a public global database. In some embodiments, these values of change of length can have a score and classified by a machine learning algorithm or a neural network. The change of length can be provided as a simple value or modeled as a normal distribution with a standard deviation. In some embodiments, the surgeon may have gathered data relating to surgeries he / she has already performed for the concerned ligament and the results he / she achieved in each case. In some embodiments, artificial intelligence enables the most relevant data to be used to assist the surgeon in the planning; said internal proprietary database might be issued from the pre-operative, intra-operative and post-operative data from other users using the same system described herein, and the artificial intelligence have access to said data useful to train the artificial intelligence. In some embodiments, said artificial intelligence can be declined as an artificial neural network (i.e. a Recommender system). In some embodiments, said artificial intelligence can be declined as a single or a plurality of machine learning algorithms (i.e. a Decision support system). In some embodiments, the database can be partially or fully generated (e.g. synthetic) by reconstruction cases from the own surgeon, from the own institution or publicly available using neural networks described in the prior art.

[0113] The inputs 2099 related to the desired graft length change can be combined as a desired weighted sum of change of length 2050. Each input 2010-2014 has a weight W0-W2b associated.

[0114] The weighted sum of graft position 2040, the weighted sum of change of length along with the definition of zones to avoid 2020 and the graft diameter 2030 can be combined 2060 into an optimal tunnel position proposal.

[0115] In any case, as will be explained below, the user will be allowed to check the proposed placement and to modify it if appropriate to finalize the planning.

[0116] Determination of the 3D model of the femur

[0117] The method for planning position of the femoral tunnel requires a 3D model of the patient’s distal femur obtained in step 402. The following method is carried out by the control unit.

[0118] Said 3D model can be typically obtained from a 3D image acquired preoperatively on the patient. For example, said 3D image can be acquired in step 400 using computed tomography (CT), cone-beam computed tomography (CBCT) or MRI modalities.

[0119] The 3D image can then be segmented to generate a 3D geometrical representation of the femur such as a mesh (step 402), which represents the outer surface of the femur.

[0120] Alternatively, instead of a 3D image, one or several 2D images 400 can be acquired using x-ray, fluoroscopy, ultrasonography or dual energy radiography modalities. In such case, in one embodiment, segmentation is carried out on the 2D image(s) and the 3D mesh is thenreconstructed by combining the 2D images with the deformation of a geometric template representation or with other computer vision methods to generate a 3D image from the plurality of 2D images. In other embodiments, one or several 2D images can be used to reconstruct a 3D geometrical representation of the femur using a neural network that will produce as an output a 3D mesh of the patient’s femur.

[0121] The 3D mesh model of the patient’s distal femur can be represented as a plurality of triangles and vertices, although it also can be represented in other embodiments as a heterogeneous mixture of triangles, squares and other polyhedral structures.

[0122] Various 2D or 3D segmentation and geometrical reconstruction algorithms are available and can be used by the skilled person. The invention is not limited to a specific algorithm and such algorithm will thus not be described in detail in the present text.

[0123] Registration of the 3D model of the femur

[0124] FIG. 5 shows the methodology to find the intercondylar roof curve. In some embodiments, the 3D model of the patient’s distal femur is rigidly registered with a known reference model (step 500). The distal femoral condyles of the 3D model are superposed in the lateral view, allowing to find a mid-sagittal plane (step 502). Using the mid-sagittal plane, the intercondylar contour is extracted (step 504). From said intercondylar contour, it is possible to identify the intercondylar roof curve by algorithms (step 506). In some embodiments, parametric intercondylar points associated with the said intercondylar roof curve are detected in step 508.

[0125] In one embodiment, the 3D mesh is registered with a reference model of the femur (also called “atlas”). The effect of this registration is to orient the 3D mesh in the coordinate system of the atlas. Said coordinate system is in particular defined according to standard anatomical planes of the human body. Preferably, the registration is a rigid registration, i.e. a registration consisting of translations and / or rotations to align the 3D mesh and the atlas, without deforming the 3D mesh. An example of an algorithm used to do the rigid registration is the iterative closest point (ICP), being the output of said algorithm the transformation matrix that links the femoral reference model with the patient’s distal femur. Another example of an algorithm to solve the rigid registration is the ICP combined with a Random Sample Consensus (RANSAC) method to increase the chances of convergence.

[0126] In another embodiment, a plurality of anatomical landmarks or features 404 on the distal femur are automatically identified by the control unit. The possible non-exhaustive landmarks list to be identified on the surface of the distal femur is: medial epicondyle, lateral epicondyle, medial posterior condyle, lateral posterior condyle, medial distal condyle, lateral distal condyle, medial trochlea anterior, lateral trochlea anterior, medial posterior proximal condyle, lateral condyle posterior condyle, anterior notch, medial condyle internal point, medial condyle external point, lateral condyle internal point, lateral condyle external point. With some of theidentified landmark’s list on the patient’s distal femur, a coordinate system can be calculated, represented as an orthonormal base. Having a known “atlas” orthonormal base aligned to the anatomical planes, it is therefore straightforward to calculate the transformation matrix that links the atlas orthonormal base with the patient’s distal femur orthonormal base and to register the patient’s distal femur to the atlas. Other methods can be used to identify such anatomical landmarks such as neural networks, machine learning algorithms or other methods that are described in the prior art and that are not listed herein.

[0127] It should be noted that the people skilled in the art might consider registration of the 3D femur mesh with a reference model using directly a rigid registration algorithm (such as ICP) of both 3D femur mesh representations, using a rigid registration of both orthonormal bases, a combination of both methods, or with the aid of other methods already explored in the prior art that are not listed herein.

[0128] The coordinate system may typically be formed of three orthogonal axes: a first axis oriented along the sagittal plane, a second axis oriented along the coronal plane and a third axis oriented along the transverse plane. This allows the definition of three directions: first axis as the medio-lateral normal to the sagittal plane, second axis as the antero-posterior, normal to the coronal plane and the third axis as the proximal-distal, normal to the transverse plane.

[0129] Once registered in said reference coordinate system, the 3D mesh forms a 3D model of the patient’s femur which will be used for planning the placement of the femoral tunnel.

[0130] As previously explained, the planning requires proposed positions of the femoral tunnel based on various sources, such as scientific literature or data issued from the user’s own database of results. Said positions are defined in a coordinate system based on anatomical features of the knee, as described below.

[0131] Determination of anatomical features

[0132] A set of anatomical features 404 is selected on the 3D model of the knee in order to define a coordinate system. The selected anatomical features may depend on the concerned ligament. The set of anatomical features may comprise one or more points, one or more lines or curves, and / or one or more planes. For example, the anatomical features of interest may be the sagittal plane, the epicondyles, the most distal point of the condyles, etc. If an MRI image is available, additional anatomical features may comprise insertion points of ligaments and tendons.

[0133] Although not limited to this embodiment, a convenient set of anatomical features is the intercondylar roof curve, which is the curve traced sagittally in the middle of the distal intercondylar notch.

[0134] In the present invention, a distinction is made between the intercondylar roof line, found in the prior art, and the intercondylar roof curve, which is the anatomical reference for the embodiments further discussed herein. The intercondylar roof line is the approximation as astraight line of the intercondylar fossa in the distal femur from a 2D projection (usually an x-ray or a fluoroscopy), by which only two points are enough for its description according to the mathematical definition of a line. By contrast the intercondylar roof curve is the curved zone of the intercondylar fossa in the distal femur, which can be described as a plurality of points (i.e. three points or more). This curve is then used for the calculation of the optimal insertion point for most of the distal femoral ligaments as further described.

[0135] FIGS. 6A-6C schematically show the determination of the intercondylar roof curve as an anatomical feature 404 on the 3D model. Said determination requires superposing the femoral condyles.

[0136] FIG. 6A shows the knee according to the anatomical sagittal plane SP. The arrows designate both condyles of the femur F. Due to potential asymmetry of the condyles, the anatomical sagittal plane does not guarantee that the femoral condyles are superposed.

[0137] To overcome this issue, in step 502, a mid-sagittal plane MSP is defined, which is the plane according to which, in a lateral view, the distal femoral condyles are aligned (superposed), as shown by the arrow in FIG. 6B. Said mid-sagittal plane MSP cuts the condyles by the intercondylar notch.

[0138] To that end, according to an embodiment, an algorithm finds the most prominent points in the antero-posterior direction of both condyles by dividing the original bounding box of the 3D model into two sub-bounding boxes along the medial-lateral direction. For each subbounding box, the most prominent distal and posterior points are found. The alignment of the condyles in then done with two rotation matrices, for both the distal and posterior points. The vector connecting the most prominent posterior points of both sub-bounding boxes is determined and projected onto the plane described by the rotation in the proximal-distal direction. Similarly, the vector connecting the most prominent distal points of both subbounding boxes is determined and projected onto the plane described by the rotation in the anterior-posterior direction. For both projected vectors, the rotation matrix between the projected vector and the direction to be aligned is determined. Once both rotation matrices are found and applied to the 3D model, the main transformation matrix is updated.

[0139] Then, in step 504, the contour (schematized by the thick line in FIG. 6C) obtained as the intersection between the mid-sagittal plane MSP and the distal femur F is analysed in order to assign a weight value to each point so as to identify the intercondylar roof curve C. This weight value describes the likelihood of each point to belong to the intercondylar roof curve.

[0140] In one embodiment, the mid-sagittal plane is used as a reference to run the triangleplane intersection algorithm with the 3D mesh of the patient’s distal femur, and whose output is a list of vertices lying on the mid-sagittal plane. Only the distal half of vertices (in the proximal-distal direction) are kept. The resulting list of intersecting points is represented as an unordered list of points in a matrix form. The points are then sorted according to a nearest-neighbour approach, producing an ordered list of points.In another embodiment, the mid-sagittal plane normal is used as a direction vector to obtain a digitally reconstructed radiography (DRR) (i.e. a “virtual” radiography) projected normally to the mid-sagittal plane. The visual image representation of said DRR contains the superposed condyles and the intercondylar contour as a plurality of monochromatic pixels. Then, using a curve algorithm detection such as the Hough transformation, a plurality of line representations are acquired. Said lines can be discretized as an ordered list of points.

[0141] In an embodiment, a fuzzy logic algorithm can be used to identify the intercondylar roof curve from the said ordered list of points. The fuzzy logic algorithm requires the list of points of the intercondylar contour to be ordered in such a way that the consecutive points move from the anterior to the posterior direction, in a counterclockwise order. Once the points are sorted and in the expected direction, the fuzzy logic algorithm analyses them and assigns to each one a weight value comprised between 0.0 and 1.0, which represents how likely a point is part of the intercondylar roof curve. To determine the weight value, the following four non-limiting weight variables can be used for each point:

[0142] the angle from abscissa (i.e. the anterior-posterior direction), which is the simple angle formed from the abscissa axis to the vector pointing to the point and the previous one;

[0143] the curvature angle, which is the angle formed from three consecutive points; the position in abscissa;

[0144] the position in ordinate (i.e. the proximal-distal direction).

[0145] The total weight value of a point is the product of the individual weight function based on the above variables. Each weight value can be represented (but not limited to) as a generalized bell-shaped function, which is defined with three variables. In certain embodiments, the total number of variables for the total weight value is 12.

[0146] The total weight value of each point is then discriminated to choose whether the point belongs or not to the intercondylar roof curve. To that end, a simple threshold discrimination method can be used, in which if the total weight value of the point is greater than a threshold value (e.g. 0.5), then the point is considered part the intercondylar roof curve.

[0147] Finally, once the points of the intercondylar roof curve are discriminated, the selected points are modeled as a parametrized curve with a linear regression. To that end, a leastsquares linear regression method can be used to fit a curve into the points of the intercondylar roof curve.

[0148] In an embodiment, the identification of the intercondylar roof curve might be done using an neural network method. The unordered list of points of the intercondylar contour can be used as a input to a neural network which has already been trained to infer and discriminate the plurality of points belonging to the intercondylar roof curve from the collection of intercondylar contour points. In some embodiments, the neural network algorithm is trained for the discrimination of the intercondylar roof curve from the plurality of vertices given from thesegmentation of the patient’s femur. In some embodiments, the discriminated points from the neural network are used as an input to a linear regression model to determine the closest intercondylar roof curve.

[0149] It is to be noted that the invention is not limited to fuzzy logic nor to neural network algorithms and any other suitable type of algorithm can be used by the skilled person to determine the intercondylar roof curve.

[0150] With the intercondylar roof curve identified in step 506, it is possible to find the parametric intercondylar points that are necessary to place the different grids allowing to determine the placement of the femoral tunnel in step 508.

[0151] In general, five parametric points are automatically located in the 2D projection of the distal femur. Two points BO, B1 correspond to the shallow and deep intersecting points of the parametrized intercondylar roof curve with the 2D projection of the intercondylar notch: these points represent the first and last point of the intercondylar notch that describe the intercondylar roof curve. Three additional points CO, C1, C2 correspond to the shallow and deep intersecting points of the parametrized intercondylar roof curve with the 2D projection of the condyles, as well as the lowest intersecting point with a parallel to the intercondylar roof curve.

[0152] Certain methods of identification of the intercondylar roof line have been explored in the prior art (Morita et al., 2016, 2020; Uozurni et al., 2014). However and as previously discussed, said methods focus on the identification of a line, either by Hough transformations that output a straight line (such as in Morita et al., 2016) or by the detection of two points using fuzzy logic (as in Uozurni et al., 2014). The method detailed in the invention herein focuses on the identification of a plurality of points that describes a curve.

[0153] ACL - Positioning of the Bernard & Hertel grid

[0154] For ACL, a widely used coordinate system 406 is the Bernard & Hertel grid, although any other suitable coordinate system may be used.

[0155] As already described in the prior art, the BH grid is a grid considered as one of the gold standards of the ACL tunnel placement. The grid reguires the three points CO, C1, C2 to be defined. The point C1 is considered the origin of the BH grid. The width axis is obtained from the points CO and C1. The height axis is a projection of the point C2 minus a width component.

[0156] Once both width and height axes are defined, the BH grid can be placed on the 3D model of the femur, as shown in FIG. 7 and depicted as BH1, using the previously computed point C1 as the origin of the BH grid. In this figure, a suggested location (averaged data from Sullivan et al., 2015) is 31.61% (19.98 - 43.24) from the posterior condyle line and 31.50% (23.27 -39.73) from the intercondylar roof curve. The values in parenthesis are the 95% confidence interval, shown as the ellipse zone Z.

[0157] The selection of the position data input 2098 for an appropriate femoral ACL tunnel position is based on the method illustrated on FIG. 4, where five different choices (2000-2008)are contemplated. In some embodiments, the suggested location can be based on other pieces of scientific literature. In some embodiments, the suggested location can be manually entered by the user in the visual user interface or located with a tracking probe attached to the CASS. In some embodiments, the suggested location can be inferred from a neural network pretrained with data from other sources. In some embodiments, the suggested location can be averaged from previous surgery outcomes from the same user or from a plurality of different users. In some embodiments the position can be given from the position of the segmented remnant ligament to be replaced.

[0158] With the same logic as for the BH grid, a 3D cylinder grid BH2 can be defined using the same points CO, C1, C2 in order to establish an equivalence between the BH grid and the sagittal projection of the cylinder (see FIG. 8A). The width axis is obtained from the points CO and C1. The height axis, which also defines the diameter of the 3D cylinder grid, is a projection of the point C2 minus a width component. The 3D cylinder grid can be visually declined as the Bernard & Hertel BH grid BH3 if a lateral view following the MSP is used (see FIG. 8B). In the same spirit, the 3D cylinder grid can be visually declined as the O’clock method with an oblique antero-posterior view following the intercondylar roof curve direction.

[0159] The user may choose to use either the BH grid or the 3D cylinder grid.

[0160] Alternatively, other grids may be used. For example, circular grids can be used instead of the BH grid. One circular grid uses the points CO and C1 , its origin being the middle of the segment linking CO and C1. Another circular grid uses the points BO and B1, its origin being the middle of the segment linking BO and B1. Both grids used polar coordinates (angle and radius) instead of cartesian coordinates.

[0161] MCL - Coordinate system

[0162] In order to find the femoral position of the tunnel during the reconstruction of the medial collateral ligament (MCL), a method to find the appropriate landmarks is to use the strict sagittal view of the medial face of the femur.

[0163] With reference to FIG. 9, using this strict sagittal view, the posterior wall projection of the femoral F diaphysis is identified using algorithms already described in the scientific literature such as the Hough transform, and marked as L1. Then, the intercondylar roof curve is also characterized as previously described and the most deep, high point of the intercondylar roof curve (computed as point B1 in the above described methodology) is identified. The line perpendicular to L1 that passes through B1 is then drawn and marked as L2. This configuration provides four quadrants (anteroproximal Q1, posteroproximal Q2, I posterodistal Q3 and anterodistal Q4) in which different landmarks can be identified and located using the intersection of L1 and L2 as the origin D1, defining the coordinate system 406.

[0164] Using the reference frame of L1 and L2, it is possible to locate the potential position of the MCL tunnel. According to scientific literature, such as Wijdicks et al., 2009 (Wijdicks, C. A., Griffith, C. J., LaPrade, R. F., Johansen, S., Sunderland, A., Arendt, E. A., & Engebretsen, L.(2009). Radiographic Identification of the Primary Medial Knee Structures: The Journal of Bone and Joint Surgery-American Volume, 91(3), 521-529), the average position of the femoral insertion of the MCL is located 8.6mm ± 3.6mm anterior to L1 and 11.0 ± 2.3mm distal to L2, in the anterodistal quadrant Q4. It is to be noted that other scientific sources might propose other average positions of the femoral insertion of the MCL.

[0165] Another source of position from the scientific literature for the MCL comes from the identification of the medial epicondyle landmark. This landmark can be identified either as the most prominent bulge of the medial wall of the distal epiphysis (using a neural network or a geometric analysis of the 3D surface), or as the approximate center of rotation of the femur during a flexion-extension in the short range of 0° (extension) to 30°, just before the anterior drawer movement taking place, which transfers the center of rotation most posteriorly (using trackers already positioned in the bones and flexing manually the knee). Once the medial epicondyle is identified, scientific literature such as Wijdicks et al., 2009 locates the femoral insertion point of the MCL 6.0 ± 0.8mm posterior to the medial epicondyle. It should be noted that other scientific sources could propose different average positions of the MCL based on the medial epicondyle.

[0166] PCL - Coordinate system

[0167] The positioning of the femoral tunnel for posterior cruciate ligament (PCL) reconstruction can be based on anatomical data obtained from imaging of the knee of the patient.

[0168] Scientific literature allows finding anatomical landmarks suitable for locating the insertion point of the PCL tunnel. An example is given with the work of A. M. Johannsen, C. J. Anderson, C. A. Wijdicks, L. Engebretsen, et R. F. LaPrade, « Radiographic Landmarks for Tunnel Positioning in Posterior Cruciate Ligament Reconstructions », Am J Sports Med, vol. 41, no 1, p. 3542, janv. 2013, doi: 10.1177 / 0363546512465072.

[0169] With reference to FIG. 10, using a strict coronal view or a strict sagittal view, a distal condyle line L3 is traced and used as a reference axis. In the strict coronal view, the second axis of reference is the medial epicondyle line L4 which is a line perpendicular to L3 tangential to the medial epicondyle of the distal femur F. An apex line L5 that is perpendicular to L3 and passes through the most proximal apex of the intercondylar fossa can be used as well as a reference axis. For the strict sagittal view, a line tangential to the posterior diaphyseal wall L1 is traced. From the projected points of the intercondylar roof curve CO and C1 a straight line is traced L6 used as a first reference axis, and perpendicular to it, and tangential to the anterior condyle, an anterior cortex line L7 is traced and used as a second reference axis. The position of the anatomical insertion point of the PCL in the coronal plane can be defined within the coordinate system characterized by these two lines L3 and L4. The coordinate system 406 can be absolute (distance in millimeters, for example) or relative, based on the complete anatomy (distance between the two epicondyles, for example).In a manner similar to the BH grid, the anatomical insertion point of the PCL can be determined statistically based on the reference point defined by the intercondylar roof curve L6 and the anterior cortex line L7.

[0170] Considering 3D imaging, it is possible to combine these systems, thereby increasing accuracy. Combining the distal condyle lines from both views gives a distal condyle plane. Combining the anterior cortex line with the strict sagittal axis gives an anterior cortex plane. Targeting distances relative to these planes allows target lines to be obtained. To determine the position on this line, the distance to the intercondylar roof curve can be used as a selection criterion.

[0171] LCL - Coordinate system

[0172] Similarly to the MCL coordinate system, the definitions of the L1 and L2 can be used in the strict sagittal view of the lateral face of the distal femur, in order to define a coordinate system where the LCL graft tunnel can be positioned as described herein.

[0173] Computation of the isometry map

[0174] The isometry map can be generated using geometrical simulations, biomechanical simulations or a tracked passive flexion using a navigation system as a method to acquire the knee flexion 405.

[0175] As will be described below, for the geometric simulation, the flexion / extension movement of the knee is simulated using the 2D projection of the strict sagittal views of the femur. Using a fixed tibial point, several points fixation point on the face of the femur to be tested, and the change of length is recorded and displayed as a map of different changes of length based on the femur fixation position.

[0176] The biomechanical simulation uses methods already described in the scientific literature such as the finite-element method, the multi-body simulation, a physics-informed neural network or the like. In either case, the biomechanical simulation can take into account further details of the anatomy of the knee, such as the mechanical properties of the soft tissues (cartilages, meniscus, other ligaments, tendons, muscles), and complex interactions (several degrees of freedom coupled with the flexion, contact between ligaments and the like). The purpose, as in the geometrical simulation, is to test several potential femoral fixation points and record the change of the length of the ligament. This generates a map that displays the different change of length of the graft based on the femur fixation position. To that end, the flexion is generated either using a muscle-driven model that produces the flexion movement, or using prescribed rotations and translations that, when combined, produce the knee flexion.

[0177] The tracked passive flexion consists in recording the motion of the femur with respect to the tibia in order to characterize the knee kinematics. This motion allows to simulate several potential femoral fixation points in order to measure the change of length of the graft and thereby generate the isometry map.In either case, the simulated flexion-extension range could include from an extension of 0° to up a flexion of 120°. The case in which the range is from 0° to 90° is also possible.

[0178] This map 407 is displayed on the surface of the 3D model of the femur where the ligament could be located. For instance, for the ACL, the isometry map is displayed on the lateral face of the intercondylar surface. The PCL isometry map is displayed on the medial face of the intercondylar surface. The MCL isometry map is displayed on the medial external distal epiphyseal face. The LCL isometry map is displayed on the lateral external distal epiphyseal face.

[0179] Methods to produce the isometry map for ACL reconstruction have been described by Dabirrahmani et al., 2013. This map represents the change of length or strain variation of the ACL in different femoral insertion positions. The method to achieve said isometry map is illustrated in FIG. 11 and executed by the control unit. The method begins with the extraction of the rigid geometrical representations of the distal femur and the proximal tibia that are in contact (step 900). Based on patient’s data, the preparation of the simulation of a knee flexion (i.e. from 0° or fully extended to 90°) implies to know beforehand the trajectory path of the rigid geometrical representation of the tibia with a fixed rigid geometrical representation of the femur (step 902). It is also necessary to identify beforehand the plurality of theoretical femoral insertion points (step 904) in the inner medial face of the lateral distal condyle of the patient’s femur, and the single theoretical tibial insertion point. The knee flexion simulation is executed (step 906), and the length change variation per femoral insertion point is acquired and stored by the control unit. The plurality of acquired length change variation points is displayed as the isometry map (step 908). The isometry map can be produced by producing a kinematics simulation of a knee movement, typically flexion-extension, and by mapping the ligament footprint in both the femur and the tibia, so as to analyse all possible combinations of ligament insertion.

[0180] In the present invention, the position of the tibial tunnel can be determined more easily by the surgeon than the position of the femoral tunnel and does not need the implementation of the present method. Thus, to implement the present method, which focuses on the placement of the femoral tunnel, only the position of the tunnel in the femur is mapped and the simulated movement is flexion-extension up to 90°.

[0181] The knee joint, and particularly the tibio-femoral joint, is a complex joint that shows movements in six degrees of freedom, which can be modelled as three orthogonal rotations and translations. The knee kinematics depends not only on the geometry of the femur and the tibia but also on the soft tissues (e.g. cartilage), the menisci, the ligaments and, less importantly, on the tendons and muscles interactions.

[0182] In an embodiment, a simplified model of said kinematics with two degrees of freedom can be used to compute the isometry map: the antero-posterior translation and the flexion-extension rotation, based on the hypothesis that the main drivers of such movement are the distal femur condyles.

[0183] To establish the simplified kinematics, a first step comprises fitting into circles the 2D projection of the femur in order to obtain a circle fitting the posterior condyle and a circle filling the full condyles. Said circles are considered the rigid geometrical representations of distal femur.

[0184] Several hypotheses may be made to compute the simplified kinematics. First, the movement occurs only in the sagittal plane and is neglected in the other planes. Besides, the anterior-posterior translation and the flexion-extension are only driven by the circles fitted in the condyles. In addition, the tibial slope and the impact of surrounding soft tissues are neglected. At last, the tibial contact point is always at a same distance of the fitted condyle circles. Therefore, the tibial contact point is considered as the rigid geometrical representation of the proximal tibia, or the single theoretical tibial insertion point of the ACL.

[0185] The determination of the tibia motion path 902 during knee flexion is done by following the perimeter of the circles of the distal femur. The simulation of the flexion-extension starts with the representation of the tibia as a single moving point in relation to the fixed femur. The starting point is the fixed tibial point aligned with the diametral axis of the full condyles circle.

[0186] The tibial flexion-extension motion path is then described by following the full condyles circle in the posterior direction up to the tangent point with the posterior condyle circle. From the tangent point to the posterior side, it is the posterior circumference which continues driving the flexion-extension movement.

[0187] In an embodiment, a partial kinematics model of the patient’s knee is acquired using the femoral cartilage rigid surface representation, as well as the tibial lateral and medial plateau cartilage as a rigid surface representation, along with the tibial menisci. The tibial motion path 902 during a knee flexion is determined by the surface curvature of the said femur cartilage and the said tibia menisci. A motion simulation algorithm (i.e. rigid-body dynamics of finite element method) coupled with this patient-specific simulation determines the precise tibial motion path with other generic input data issued from the scientific literature, such as the muscular forces, the motion constraints such as the other ligaments placement, and the surface-to-surface contact properties (i.e. penalty-based, frictionless contact). In some embodiments, a physics-informed neural network can predict the kinematics of the tibiofemoral joint as a method to determine the tibia motion path 902.

[0188] In an embodiment, the kinematics model of the patient’s knee is acquired intra-operatively with trackers attached to the femur and the tibia along with a navigation system coupled to a computer-aided surgical system (CASS) that allows the tracking in real-time of the position and orientation (i.e. in 6 degrees of freedom) of the tibia in reference to the femur. This method also allows the determination of the tibia motion path during flexion 902. The rigid geometrical representation of distal femur and proximal tibia consists in the segmented orreconstructed 3D models obtained in step 402. The tibia motion path during knee flexion 902 is obtained with the knee flexion and extension practiced directly to the patient’s knee by the user or any other external party. The navigation system provides as an output the tibial motion path during the knee flexion in reference to the patient’s femur. The knee flexion and extension can be practiced several times and the data then averaged to provide more robust data to the kinematics model. In step 908, the isometry map is obtained by simulating a knee flexion 906 with all the possible femoral ligament graft insertion points 904 and measuring for each virtual graft the change in length between the knee in full extension (0°) and in flexion. The ligament graft insertion points 904 are all the theoretical points represented in the femoral anatomical face where the graft can be positioned. For instance, for the ACL, the insertion points 904 are projected on the lateral face of the intercondylar surface. For the PCL , the insertion points 904 are projected on the medial face of the intercondylar surface. For the , the insertion points 904 are projected on the medial external distal epiphyseal face. For the LCL and the ALL, the insertion points 904 are projected on the lateral external distal epiphyseal face. In an embodiment, for the case of the ACL, in FIG. 12A it is depicted a virtual ACL graft in a knee in full extension (0°) and in FIG. 12B, the knee in flexion (90°) along with the predicted femoral path. This variation corresponds to three zones in the isometry map shown in FIG. 12C: the laxity region in which the length variation is negative, the tension region in which the length variation is positive, and the isometry region in which the length variation is zero.

[0189] ACL — Isometry map

[0190] The isometry map obtained in step 908 is able to describe an interval of length variations values (which can be seen as a strip) depending on the physical position of the lateral inner condyle wall of the distal femur. In vitro or in silico studies (Iwahashi et al., 2008; Yoo et al., 2010) have measured the length variation of the ACL during a knee flexion. From 0° to 90°, in the middle bundle of fibres, they have found average values from 5 to 6.7 mm of length decreasing in its anatomical position. This data, along with the average coordinate values of the ACL footprint using the BH grid, can be used as targets for ACLr planning.

[0191] It should be noted that other ligament graft length change values can be proposed. MCL - Isometry map

[0192] The MCL is a large ligament and therefore covers isometric and anisometric (tension and laxity) regions, as demonstrated in the study made by Kittl et al., 2021 (Kittl, C., Robinson, J., Raschke, M. J., Olbrich, A., Frank, A., Glasbrenner, J., Herbst, E., Domnick, C., & Herbort, M. (2021). Medial collateral ligament reconstruction graft isometry is affected by femoral position more than tibial position. Knee Surgery, Sports Traumatology, Arthroscopy, 29(11), 3800-3808).

[0193] Although the scientific opinion states that the MCL should be mainly working in isometry, an acceptable MCL length change for an optimum positioning of the femoral tunnel for the MCL should be between -5% and 5% for a flexion-extension of 0°-90°.It should be noted that others scientific studies could propose different values and could be included in the scientific database proposals so the user can choose.

[0194] PCL - Isometry map

[0195] The scientific literature agrees that the PCL is not isometric in nature.

[0196] In fact, no bundle remains constant between 0° and 135°. The anterolateral bundle (ALB) is more anisometric than the posteromedial bundle (PMB). The actual length of the PCL varies greatly, but natural curvature partially compensates for these variations. Kernkamp (2019) shows that during the reconstruction of the “most isometric” PCL, the graft remains anisometric and behaves in a non-physiological manner. The management of anisometry is therefore crucial for PCL reconstruction.

[0197] Another important point to mention is that the femoral area is the main determinant of isometry. The position of the femoral tunnel has a significant influence on isometry, whereas the tibial tunnel has very little influence.

[0198] The isometry of the PCL can be controlled by choosing to be closer or further away from the insertion points of the native PMB or ALB bundles. It is important to remain within the anatomical area and not be too proximal. This would improve isometry, but the tunnel would then not be physiological.

[0199] To conclude, PCL isometry can only be achieved in part. The best physiological compromise is based on a posterodistal anatomical femoral tunnel (PMB zone) and maximal preservation of the remnant to maintain the natural curvature of the ligament.

[0200] An isometry map on the internal surface of the medial condyle of the femur can help to identify the most suitable location.

[0201] It should be noted that other values for the PCL graft length change can be found elsewhere.

[0202] LCL - Isometry map

[0203] Similarly to the MCL, the LCL works in average in isometry, and therefore, the targeted values should be with an average change of length of zero. Nevertheless, other surgical strategies exists where the change of length could be different from zero.

[0204] Additional input data

[0205] Graft diameter 2030 is a non-negligible parameter in ligament reconstruction, as it directly influences how well the graft can be integrated within the footprint of the native ligament. Ideally, the entire graft circumference should remain within the remnant native ligament fibers to better reproduce physiological load sharing and heal better. As graft diameter 2030 increases, the surface area covered on the isometry map also becomes larger, which in turn increases the likelihood that peripheral graft fibers will behave anisometrically throughout knee motion.Consequently, tunnel placement is preferably not defined independently of graft size: the optimal location may be selected according to both the graft diameter and the local isometry profile in order to achieve a graft behavior that is as close as possible to native ligament biomechanics.

[0206] Nevertheless, certain zones may be deliberately avoided 2020 to prevent impingement. The graft should not conflict with adjacent anatomical structures during flexion-extension, as this may lead to abnormal stresses, loss of motion, or graft failure. Similarly, areas too close to critical structures - such as the meniscal roots, neighboring ligaments, and tendons - should be excluded to minimize the risk of iatrogenic damage and to preserve the functional integrity of the surrounding anatomy.

[0207] Fusion of the metrics

[0208] After the acquisition of the 3D geometry of the femur, the system begins with the choice of the position of the tunnel where to pass the graft of the ligament. FIG 4 illustrates an exemplary workflow comprising a plurality of options. Of course, the workflow may use only of said options, or may use any suitable additional option.

[0209] To determine the desired position of the graft (and hence the position of the femoral tunnel), in step 2040, five types of inputs may be used:

[0210] - User selection on screen 2000

[0211] - User selection as a tracked palped zone 2002

[0212] - Remnant segmented zone 2004

[0213] - Scientific literature 2006

[0214] - Database of prior surgeries 2008.

[0215] The user selection on screen 2000 implies that the user is freely able to choose on the screen by directly touching it (touchscreen) or using another input device (mouse, keyboard, joystick, augmented reality interface, or the like) the position on the 3D model of the femur in which the graft should be attached on the femur. To model this position as a normal distribution, the system can assume a default standard deviation of the user election.

[0216] The user selection can also be done using a tracked palping probe 2002 attached to the navigation system of a computer-aided surgical system (CASS) which can be used to physically contact the zone where the user wishes to position the graft. This palpation can be repeated a number of times in order to model a normal distribution with an average position along with its standard deviation.

[0217] If a magnetic-resonance image (MRI) or an ultrasound segmentation containing the ligament tissues is available, the user can choose the remnant of the ruptured ligament 2004 as the point to position the new graft. Segmentation can produce an average center of the ligament footprint as its position, along with an average footprint area expressed as a standard deviation.In other examples, the user may choose to refer to the scientific literature 2006 available in the system where anatomical studies suggest positions with respect to landmarks that are already identified by the system. The user can display only a list of different studies along with the proposed tunnel positions and its standard deviation. The displayed list can be shown on a screen and be selected using an input device.

[0218] In another example, the system can propose a user selection 2008 of tunnels positions based on a database of prior surgeries or synthetic (e.g. generated by a neural network), private from the own user, common to an institution or widely common among all the users of the system. This database can display a list of average positions along with their standard deviation, or a list of suggested positions in which a machine learning or a neural network algorithm is able to score and classify for its level of success, where level of success is understood as the surgeries that created the least adverse effects.

[0219] Each of the five inputs for the choice of the tunnel graft position 2000-2008 is associated with a weight W0a-W4a whose sum is unity (1). Therefore, each weight W0a-W4a assigns the value of priority of each one of the five inputs in order to produce a single tunnel graft position. As each one of the five inputs is modeled as a normal distribution, the weighted sum of the normal distributions is another normal distribution. For a given coordinate system that can be in either two dimensions (2D projections) or three dimensions (3D cartesian, spherical, cylindrical), each input 2000-2008 provides an average center p0ax-p4ax, p0ay-p4ay and p0az-p4az and a standard deviation oOax- o4ax, oOay- o4ay and oOaz- o4az for each of the two or three dimensions. The weighted sum of the average center for each dimension results in the average position pa of the graft tunnel, and the square root of the weighted sum of squares of the standard deviation for each dimension as the standard deviation oa of the position of the graft tunnel. This results in the weighted sum of graft positions 2040. In some embodiments, the tunnel position could be simply modeled as a coordinate value without standard deviation, applying only and weighted sum of the centers for each dimension for the resulting position of the graft tunnel position. With this method, up to five inputs can be combined, and as the user manipulates the weights, the number of inputs to be considered for the resulting position of the tunnel graft could be reduced up to one.

[0220] The choice of the weights W0a-W4a is defined by the user by either using an interactive screen and an input device (for example, a series of scrollers directly manipulated in the screen), or by an automatic decision system managed by a machine learning algorithm or a neural network that can choose, based on the number of inputs, the weights W0a-W4a for each one.

[0221] The choice of the tunnel graft position can be done using these methods for other ligaments in a multi-ligament operation. The user can select either independently among the choices 2000-2008 or at the same time in order to prioritize the selection of common tunnels, for instance.The system provides three different choices for the selection of the appropriate desired change of ligament length based on the isometry map 2050:

[0222] - User selection 2010

[0223] - Choice given by the scientific literature 2012

[0224] - Choice from a database of prior selections 2014

[0225] In a similar fashion to the selection of the position of the tunnel graft, the user 2010 can choose the desired change of length on the screen using a touchscreen or an input device.

[0226] As another input, the choice can be given by a list provided by the scientific literature 2012 that can be displayed and the user can choose among the possible values.

[0227] The other input consists in the choice 2014 of a possible change of length value that comes from a database of prior successful surgeries or synthetic surgery cases (e.g. generated by a neural network), either private to the user, to the institution or as a public global database. In some embodiments, these values of change of length can have a score and classified by a machine learning algorithm or a neural network. The change of length can be provided as a simple value or modeled as a normal distribution with a standard deviation.

[0228] Each one of the three inputs for the change of length 2010-2014 has an associated weight W0b-W2b whose sum is unity (1). The value of change of length being scalar, each value of the change of length for each input p0b-p2b associated with its weight can give the weighted sum pb of change of length 2050. If standard deviation is available, in a similar fashion the squared root of the weighted sum of squares of each of the standard deviation oOb- o2b can provide the average standard deviation ob of the change of length 2050.

[0229] The choice of the weights W0b-W2b is defined by the user by either using an interactive screen and an input device (for example, a series of scrollers directly manipulated in the screen), or by an automatic decision system managed by a machine learning algorithm or a neural network that can choose, based on the number of inputs, the weights W0b-W2b for each one.

[0230] The weighted average desired graft position 2040 is a proposal of the position of the graft tunnel, in certain embodiments along with the standard deviation of such average position. The weighted average desired change of graft length 2050 corresponds to a region in the isometry map along with, in certain embodiments, the standard deviation of such average.

[0231] The view of both the desired graft position 2040 and the desired change of length 2050 can be done using a screen disposed at the surgical theater, in virtual reality glasses or projected on a surface, such as a flat wall or directly on the patient’s anatomy. The view can superpose the coordinate system used to identify the proposed graft tunnel position, in either two or three dimensions, and the isometry map, in two or three dimensions.

[0232] The fusion results in a proposed tunnel position as a union of both the desired graft position 2040 and the desired change of length 2050 that satisfies certain criteria:TJ

[0233] - the desired position vector 2040 can be modified up to a certain degree defined by a weight Wax, Way and, if three dimensions, Waz;

[0234] - the desired change of length 2050 can be modified up to a certain degree defined by a weight Wb;

[0235] - the proposed tunnel position is limited by the defined graft diameter 2030 (e.g. the position along with its diameter cannot be beyond a minimum wall thickness when drilling);

[0236] - if available, the proposed tunnel position placement is bounded by the defined zones to avoid 2020.

[0237] Each one of the weights of the of the positional inputs 2098 W0a-W4a and the isometry inputs W0b-W2b, along with the desired position weights Wax, Way, Waz and the desired graft length change weight Wb can be modified by the user with a visual interface, using a touchscreen or a screen and an input device, with the form of sliders or any other manner to manipulate such values dynamically, fully manually, as a standard presets or semi-automatically based on machine learning algorithms. For instance, if an input 2098 is not present, by default the system will compute the associated weight to zero, or if an input 2099 has a large standard deviation, its associated weight is reduced. An example of such visual interface is depicted in FIG. 13, with user interface elements in the form of sliders. At the center of the interface, the 3D model of the femur is displayed along with the coordinate system and proposed areas for positioning the tunnel based on anatomical considerations and based on the isometry map. On the left side of the interface, sliders for each type of user input related to the desired graft position are provided. On the right side of the interface, sliders for each type of user input related to the desired length change are provided. In the bottom of the interface, sliders for adjusting the weights Wax, Way (and if available Waz) and Wb are also provided. In some embodiments, initial values for each input can be set by presets or by machine learning.

[0238] The user may thus interactively modify the position inputs weights W0a-W4a, the length change inputs weights W0b-W2b and / or the weights Wax-Waz and Wb by moving the respective sliders. As a result, the representation of the proposed position of the femoral tunnel is adjusted so that the user can see the updated position of the femoral tunnel.

[0239] When the user is satisfied by the proposed position, he may validate it via a suitable user interface element, such as a button (not represented here).

[0240] In the same way, the user can define the tunnel diameter and the zones to avoid as previously described.

[0241] According to the invention, both the desired change of length 2050 of the ligament graft and the desired position of the graft 2040 are target metrics combined to automatically propose to the user a confidence interval in which the femoral tunnel can be positioned if both the desired length change and graft position lies within the same zone (see FIG.14A).As a result of the metrics combination, an allowable area in which the femoral tunnel position is recommended can be displayed, corresponding to the intersection I of the ranges of values of each metrics (R1 being the proposed location in the coordinate system (e.g. BH grid BH1) along with its confidence interval, and R2 being the range of graft length variation).

[0242] In an embodiment, a way of combining the target metrics is to consider each range as a fuzzy set: the average desired graft position 2040 along with its standard deviation being the first set, an average desired change of length 2050 along with its standard deviation from the isometry map being the second set. Based on both fuzzy sets, the automatic proposed position is a mathematical function composed of an arithmetic operation (i.e. product or addition) of both fuzzy variables and then the defuzzified values for each coordinate in the reference coordinate system. The method of defuzzifiying variables seen as a distribution (i.e. normal distribution of a confidence interval) can be seen as solving the definite integral of the functions of both the distribution of the ligament graft length range and the distribution of the desired positional data, for the axes of the coordinate system (Liu & Guan, 2016). The merged defuzzified data can be displayed in the coordinate system superposed to the other data.

[0243] In an embodiment, the fusion of the metrics concerning the average desired graft position 2040 and the average desired change of length 2050 data consists in a machine learning method that takes the normal distribution of the average position 2040, the normal distribution of the “strip” of allowable values of the ligament graft 2050 length range from the isometry map, and then a Boolean mathematical function provides the intersection of both distributions, being the latter the allowable optimal area to place the femoral tunnel.

[0244] In the case where the desired graft position 2040 is not within the desired length change range 2050, as another embodiment for the fusion to be done automatically, the first step is to transform the isometry area from the average desired change of length 2050 in a zone defined as a collection of positions in the reference coordinate system of the tunnel graft position. To that end, the average weighted change of length value pb 2050 describes an isocontour of the isometry map with the score of 1 , as well as the standard deviation ob of the average weighted change of length value 2050 if available describes isocontours, each with the standard deviation value as score. This is depicted in FIG. 14B where a coordinate system (e.g. a BH grid BH1) is used to show the desired position R1 and the desired length change R2, along with zones to avoid ZA and a mathematical function G that converges to a position R1 closer to a length change R2 and that takes into account the graft diameter defined of R1 and the zones to avoid ZA.

[0245] In some embodiments, as shown in FIG. 15, the mathematical function G stipulates the next operation that moves the average tunnel graft position pa 2040 to be closer to the average change of length value pb 2050, via a gradient and the weights Wa and Wb. The sum of these weights is unity (1). The gradient operation is done for each one of the two or three dimensionsx, y, z in which the graft tunnel position is defined. This gradient can be defined in several manners.

[0246] In an embodiment, the gradient is defined as the distance of dx = (pb - pax) Wax, dy = (pb - pay) Way, dz = (pb - paz) Waz respectively, where the distance dx, respectively dy, dz is the displacement required in one dimension for pax, respectively pay, paz to get closer to pb. In another embodiment, the gradient can be computed via a fuzzy logic algorithm assorted with the weights Wax-Waz and Wb in order to compute the distance to translate dx, dy, dz. In another embodiment, the gradient can be defined using another machine learning algorithm or neural network. In either case, the distance is limited to the already defined diameter 2030 or zones to avoid 2020. T ranslating pax, pay and / or paz will produce a new proposed fused tunnel position Px, Py and / or Pz.

[0247] The mathematical function is not necessarily based on such gradients and the skilled person may use any other suitable function. As for the weights, parameters of the mathematical function can be adjusted via the graphical user interface.

[0248] At the end, the user is requested to validate the femoral tunnel position or to modify the weights W0a-W4a and the isometry inputs W0b-W2b, along with the desired position weights Wax, Way, Waz and the desired graft length change weight Wb so to propose automatically a new tunnel graft position.

[0249] Tunnel orientation and diameter

[0250] The last step of the ACL femoral tunnel placement is the definition of the orientation of the femoral tunnel. For this, the user is invited to select the orientation manually, either via a visual user interface presented in a touchscreen or with input devices such as keyboards and a mouse, or via a tracked palpation probe coupled to a CASS in which the user can mark physically in the femur a second position (the first being the desired graft position defined previously) for which an orientation can be defined. The user can also refer to a database of prior surgeries, private to the user, local to the institution (e.g. hospital or clinical center) or available for all the users of the system, in which the user can pick a particular orientation configuration.

[0251] The tunnel diameter is defined intra-operatively based on the planning of the surgical technique and the size of the ligament graft 2030.

[0252] Navigated surgical system

[0253] The navigated surgical system comprises a planning system, a localization system, a navigated device and a control unit.

[0254] According to a preferred embodiment, the localization system relies on electromagnetic technology. To that end, electromagnetic transducers are rigidly fixed to the patient’s femur,the patient’s tibia and on the navigated device. Among the transducers, at least one is an emitter configured to generate a magnetic field, and at least one is a receiver configured to measure the magnetic field and generate localization data in response to the measured magnetic field.

[0255] The localization system may rely on any other suitable technology, such as optical technology. In some embodiment, optical brackets or visual markers can be rigidly installed in the bodies to be tracked such as the femur, the tibia, and the navigated device. A stereocamera is disposed appropriately in the surgical room and a navigation system captures the image flux from said stereo-camera to process it and track the movement in 6 degrees of freedom of each body to be tracked. In some embodiments, the localization system might comprise inertial measurement unit I MU to track the movement of said bodies in 6 degrees of freedom. In some embodiments, the localization system might comprise ultrasonography trackers to track the movement of said bodies at least 3 degrees of freedom. It is worth noting that those skilled in the art might be disposed to use another localization technology not discussed herein. Regardless of the localization technology, the devices used to track the position of the bones and the navigated device are designated by the general term “trackers” in the following description.

[0256] As shown in FIG.16A and 16B, the navigated device comprises either the surgical tool with a tracker 1201 attached directly to the body of the surgical tool 1202 (FIG. 16A), or a guide 1203 with a handle, a tracker 1201 attached directly to the body and a cavity in which an external surgical tool 1202 can be rigidly attached to (FIG. 16B). In a preferred embodiment, the surgical tool comprises a drill.

[0257] In some embodiments, the surgical tool or the guide can be held in the surgeon’s hand. In other embodiments that will be described below, the surgical tool can be held by a robotic arm of a robotic surgical system.

[0258] The navigated surgical system comprises at least one control unit, which may be the same as the control unit for the planning described above, or a distinct control unit which may be able to communicate with the previous one.

[0259] The entire system is installed at the beginning of ante ri or cruciate ligament reconstruction surgery. The process of the surgery is described below and is shown on FIG. 17. Only steps that interact with the system are described.

[0260] First, the surgeon makes incisions in the patient’s leg to attach the trackers 1300. At least two trackers are necessary, one on the femur, the other on the tibia. The medical team connects and sets up the navigation system around the patient 1302.

[0261] In order to carry out planning on a 3D model that is tracked in real time, a registration step 1304 is required.

[0262] In an embodiment, the registration step is performed using an X-ray imaging system. The X-ray (or fluoroscopy) imaging system comprises an X-ray source and an X-ray detector,arranged on a C-arm comprising several degrees of freedom in rotation and in translation relative to the patient. The X-ray imaging system is used to acquire 2D X-ray images of the patient. The 2D X-ray images are then used to reconstruct a 3D model of the patient’s bones in the coordinate system of the localization system. This reconstruction is performed by the control unit. In an embodiment, this reconstruction is performed using computed tomography algorithms known by the skilled person. In that case, several 2D X-ray images are necessary. In a preferred embodiment, the 3D reconstruction is performed using few 2D X-ray images, which reduces the radiation dose for the patient, by inferring the obtained images with a machine learning model. Such a method is increasingly developed and accessible to the skilled person (Yim et al., 2021).

[0263] In an embodiment, the registration is also performed using X-ray imaging system that acquires 2D X-ray images of the patient’s knee and an already known 3D models of the bones from pre-operatory imaging (i.e. MRI for diagnosis). The 2D X-ray images come along with a calibration target, which in some embodiments could be a radiotransparent frame with radiopaque speckles or spheres of small diameter that forms a known pattern. The control unit performs a registration between the 3D model of the patient’s femur and the 2D X-ray images acquired by the X-ray imaging system. Such registration is called 2D / 3D registration.

[0264] The registration may rely on any other methods, such as a femur or tibia palpation with a navigated anatomical probe. The palpation method consists in touching with a tracked point probe a plurality of surface patches on the femur or tibia. Then, using a rigid registration algorithm such as the ICP, a surgical reference frame can be deducted, and the transformation matrix from the segmentation (step 402) reference frame to said surgical reference frame is found, therefore completing the registration. Various other registration algorithms are available for the skilled person and will thus not be described in more detail in the present text.

[0265] In the next step 1306, the control unit receives the navigated 3D model of the patient’s femur along with the planning validated by the surgeon. The planning includes at least the position and orientation of the femoral tunnel.

[0266] Then, the navigation step 1308 of the surgical tool 1202, or the surgical tool guide 1203, enables to precisely respect the planning when drilling the femur. To this end, the control unit 200 displays in real time the drilling axis deduced from the navigated tool and the theoretical drilling axis given by the planning. This feedback enables the surgeon to fine-tune its procedure so that the two axes correlate. Once the two axes are aligned, the surgeon can proceed with drilling.

[0267] Once the drilling of the femoral side is achieved, the next step may comprise the drilling of the tibial tunnel, if not achieved previously, according to the methods of the ligament reconstruction known by those skilled in the art.

[0268] Then, the surgeon inserts the graft ligament into the femoral tunnel and the tibial tunnel.The final step concerns the removal of trackers from the patient’s anatomy and the system uninstallation 1310.

[0269] Robotic surgical system

[0270] The planning can be advantageously used as an input for a robotic surgical system configured to assist anterior cruciate ligament reconstruction.

[0271] With reference to FIGS. 18A and 18B, the robotic surgical system comprises an X-ray imaging system 1401, a planning system 1402, a localization system 1403 and a robotic arm 1404 with an end effector 1405.

[0272] As depicted in FIG. 19Aand 19B, the robotic arm 1501 comprises a plurality of motorized joints and an end effector which supports a tool to drill 1503 the femoral tunnel (FIG. 19A), also called an active end effector, or a tool guide configured to guide the drilling tool 1504 in translation (the drilling tool being operated by the surgeon) (FIG. 19B), also called a passive end effector. In either case, a tracker 1502 is attached to the robotic arm, to provide the link with the localization system 1403.

[0273] The robotic surgical system comprises at least one control unit as depicted in FIG. 2, which may be the same as the control unit 200 for the planning described above, or a distinct control unit which may be able to communicate with the previous one via the inputs 220 and the outputs 221, or via the network 250. A function of the control unit of the robotic surgical system is to compute and instruct movement(s) to the motorized joints of the robotic arm which holds the surgical tool or the tool guide. The control unit can, for instance, comprise one or more microprocessors 201, one or more random access memory (RAM) 202 and / or one or more read-only memory (ROM) 203, an operating system 205 one or more calculators, one or more computers and / or one or more computer programs 206. The computer program(s) 206 comprise code instructions to compute the needed instructions to be sent to the motorized joints of the robotic arm. In addition, the control unit may include other devices and circuitry for performing the functions described herein such as, for example, a storage unit 204 (i.e. a hard drive), input / output circuitry, and the like. The input / output circuitry can be adapted to treat digital and / or analog signals. The control unit 200 can be linked to a network 250 which could be a local area network or a wide area network.

[0274] The processing workflow of the robotic surgical system follows the steps of the navigated surgical system described above and illustrated on FIG. 17, considering that the navigated tool is the robotic arm.

[0275] As a result, the robotic arm, can be navigated in the 3D model of the patient’s femur. Based on the planning, which defines the position and orientation of the femoral tunnel with respect to the 3D model of the patient’s femur, the control unit computes instructions to the motors of the robotic arm, to align the tool (active end effector) or tool guide (passive end effector) with the target position and orientation of the femoral tunnel.Said alignment can be maintained even if the patient slightly moves, which is a robotic operating mode called servoing.

[0276] Once the target position and orientation of the tool guide is achieved, the surgeon can insert the drilling tool within the tool guide to drill the femoral tunnel. The drilling of the tibial tunnel can be done manually by the surgeon without assistance from the robotic arm according to the methods of ligament reconstruction known by those skilled in the art.

[0277] Then, the surgeon inserts the graft ligament into the femoral tunnel and the tibial tunnel. The robotic surgical system thus provides a greater accuracy in the drilling of the femoral tunnel, which reduces the risk of bad recovery of the knee mobility.

[0278] REFERENCES

[0279] A. Amis, A. M. J. Bull, C. M. Gupte, I. Hijazi, A. Race, et J. R. Robinson, « Biomechanics of the PCL and related structures: posterolateral, posteromedial and meniscofemoral ligaments », Knee Surg Sports Traumatol Arthrosc, vol. 11, no 5, p. 271 281, sept. 2003, doi: 10.1007 / S00167-003-0410-7.

[0280] Dabirrahmani, D., Christopher Hogg, M., Walker, P., Biggs, D., & Mark Gillies, R. (2013). Comparison of isometric and anatomical graft placement in synthetic ACL reconstructions: A pilot study. Computers in Biology and Medicine, 43(12), Article 12. https: / / doi.Org / 10.1016 / j.compbiomed.2013.10.008

[0281] Forsythe, B. et al., « Dynamic Three-Dimensional Computed Tomography Mapping of Isometric Posterior Cruciate Ligament Attachment Sites on the Tibia and Femur: Single-Versus Double-Bundle Analysis », Arthroscopy, vol. 36, no 11, p. 2875 2884, nov. 2020, doi: 10.1016 / j.arthro.2020.06.006.

[0282] T. Ted Funahashi, K. R. Kaufman, et D. M. Daniel, « Isometry and graft placement in posterior cruciate ligament reconstructive surgery », Operative Techniques in Sports Medicine, vol. 1, no 2, p. 110 114, avr. 1993, doi: 10.1016 / S1060-1872(10)80039-5.

[0283] Hart, A., Han, Y, & Martineau, P. A. (2015). The Apex of the Deep Cartilage: A Landmark and New Technique to Help Identify Femoral Tunnel Placement in Anterior Cruciate Ligament Reconstruction. Arthroscopy, 37(9), Article 9. https: / / doi.Org / 10.1016 / j.arthro.2015.03.026 Iwahashi, T., Shino, K., Nakata, K., Nakamura, N., Yamada, Y, Yoshikawa, H., & Sugamoto, K. (2008). Assessment of the “functional length” of the three bundles of the anterior cruciate ligament. Knee Surgery, Sports Traumatology Arthroscopy, 16(2), Article 2. https: / / doi.Org / 10.1007 / S00167-007-0456-z

[0284] W.-S. Jeong et al., « An Analysis of the Posterior Cruciate Ligament Isometric Position Using an In Vivo 3-Dimensional Computed Tomography-Based Knee Joint Model », Arthroscopy, vol. 26, no 10, p. 1333 1339, oct. 2010, doi: 10.1016 / j.arthro.2010.02.016A. M. Johannsen, C. J. Anderson, C. A. Wijdicks, L. Engebretsen, et R. F. LaPrade, « Radiographic Landmarks for Tunnel Positioning in Posterior Cruciate Ligament Reconstructions », Am J Sports Med, vol. 41, no 1, p. 35 42, janv. 2013, doi: 10.1177 / 0363546512465072.

[0285] H.-J. Jung et al., « The isometry of two different paths for remnant-preserving posterior cruciate ligament reconstruction », Knee Surg Sports Traumatol Arthrosc, vol. 21, no 5, p. 1029 1035, mai 2013, doi: 10.1007 / s00167-012-2111-6.

[0286] Kernkamp, W.A. et al., « Anatomic is better than isometric posterior cruciate ligament tunnel placement based upon in vivo simulation », Knee Surgery, Sports Traumatology, Arthroscopy, vol. 27, no 8, p. 1795, 2019, doi: 10.1007 / s00167-018-5233-7.

[0287] Liu, H., & Guan, J. (2016). A Model of Fuzzy Normal Distribution. Open Journal of Statistics, 6(5), Article 5. https: / / doi.org / 10.4236 / ojs.2016.65061

[0288] Marwan, Y, Bdttcher, J., Laverdiere, C., Jaffer, R., Burman, M., Boily, M., & Martineau, P. A. (2020). Three-Dimensional Magnetic Resonance Imaging for Guiding Tibial and Femoral Tunnel Position in Anterior Cruciate Ligament Reconstruction: A Cadaveric Study. Orthopaedic Journal of Sports Medicine, 8(3), Article 3. https: / / doi.org / 10.1177 / 2325967120909913 Morita, K., Kobashi, S., Kashiwa, K., Nakayama, H., Kambara, S., Morimoto, M., Yoshiya, S., &Aikawa, S. (2016). Blumensaat’s line detection for Quadrant method on MR images. 2016 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), 2144-2149. https: / / doi.org / 10.1109 / FUZZ-IEEE.2016.7737957

[0289] Morita, K., Nii, M., Koh, M.-S., Kashiwa, K., Nakayama, H., Kambara, S., Yoshiya, S., & Kobashi, S. (2020). Bone Tunnel Placement Determination Method for 3D Images and Its Evaluation for Anterior Cruciate Ligament Reconstruction. Current Medical Imaging Formerly Current Medical Imaging Reviews, 76(5), Article 5. https: / / doi.Org / 10.2174 / 1573405614666181030125846

[0290] Sullivan, J. P, Cook, S., Gao, Y, & Wolf, B. R. (2015). Radiographic Anatomy of the Native Anterior Cruciate Ligament: A Systematic Review. / 7SS Journal®, 11(2), Article 2. https: / / doi.Org / 10.1007 / S11420-014-9417-5

[0291] Uozumi, Y, Nagamune, K., Nakano, N., Nagai, K., Nishizawa, Y, Hoshino, Y, Matsushita, T., Kuroda, R., & Kurosaka, M. (2014). An automated determination of Blumensaat line using fuzzy system based on physician experience from femur CT image. 2014 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), 1831-1836. https: / / doi.org / 10.1109 / FUZZ-IEEE.2014.6891742

[0292] Wijdicks, C. A., Griffith, C. J., LaPrade, R. F., Johansen, S., Sunderland, A., Arendt, E. A., & Engebretsen, L. (2009). Radiographic Identification of the Primary Medial Knee Structures: The Journal of Bone and Joint Surgery-American Volume, 91(3), 521-529 Yim, D., Lee, S., Nam, K., Lee, D., Kim, D. K., & Kim, J.-S. (2021). Deep learning-based image reconstruction for few-view computed tomography. Nuclear Instruments and Methodsin Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 1011, 165594. https: / / doi.Org / 10.1016 / j.nima.2021.165594

[0293] Yoo, Y.-S., Jeong, W.-S., Shetty, N. S., Ingham, S. J. M., Smolinski, P., & Fu, F. (2010). Changes in ACL length at different knee flexion angles: An in vivo biomechanical study. Knee Surgery, Sports Traumatology, Arthroscopy, 78(3), Article 3. https: / / doi.org / 10.1007 / s00167-009-0932-8

[0294] Zee, M. J. M., Sulaihem, R. A., Diercks, R. L., &Van Den Akker-Scheek, I. (2021). Intra-and interobserver reliability of determining the femoral footprint of the torn anterior cruciate ligament on MRI scans. BMC Musculoskeletal Disorders, 22(1), Article 1. https: / / doi.Org / 10.1186 / s 12891-021 -04376-5

Claims

36CLAIMS1. Computer-assisted method for planning placement of a femoral tunnel for ligament reconstruction, comprising:- obtaining a 3D model of the patient’s femur;- based on said 3D model, determining a set of anatomical features of the patient’s femur; - based on said set of anatomical features, positioning a coordinate system on the 3D model of the patient’s femur,- based on the 3D model of the patient’s femur, computing an isometry map defining a length variation of the ligament during knee flexion depending on a position of the femoral tunnel, the isometry map comprising (i) a laxity region wherein the length variation is negative, the ligament contracting during knee flexion, (ii) a tension region wherein the length variation is positive, the ligament tightening during knee flexion, and (iii) an isometry line separating the laxity region and the tension region wherein the length variation is null, the ligament undergoing no deformation during knee flexion;- obtaining at least one first input (2000-2008) related to a desired graft position in the coordinate system (2040) and at least one second input (2010-2014) related to a desired length variation in the isometry map (2050); and- automatically computing a proposed position of the femoral tunnel by combining said desired graft position and desired length variation.

2. Computer-assisted method according to claim 1, further comprising obtaining at least one third input (2030) related to a size of a graft to be implanted in the femoral tunnel and combining said third input with the desired graft position and desired length variation to automatically compute the proposed position of the femoral tunnel.

3. Computer-assisted method according to claim 1 or claim 2, further comprising obtaining at least one fourth input (2020) related to a zone of the femur to be avoided and combining said fourth input with the desired graft position and desired length variation to automatically compute the proposed position of the femoral tunnel.

4. Computer-assisted method according to any one of claims 1 to 3, further comprising, after computing the proposed position of the femoral tunnel, requesting and receiving a fifth input.

5. Computer-assisted method according to claim 4, wherein the fifth input is a validation of the proposed position of the femoral tunnel.

376. Computer-assisted method according to claim 4, wherein the fifth input is a requested adjustment of the proposed position, the method comprising automatically computing an adjusted proposed position of the femoral tunnel based on the fifth input.

7. Computer-assisted method according to any one of claims 1 to 6, wherein the first input comprises at least one of:- a position (2000, 2002) selected by a user;- a remnant segmented zone (2004);- a statistical position from scientific literature (2006); and- a statistical position from a database of reconstruction cases (2008).

8. Computer-assisted method according to claim 7, wherein at least two first inputs are weighted with a respective weight (W0a-W4a) and summed to compute the desired graft position (2040), the sum of the weights being equal to 1.

9. Computer-assisted method according to any one of claims 1 to 8, wherein the second input comprises at least one of:- a length variation selected by a user on the isometry map (2010);- a statistical length variation from scientific literature (2012); and- a statistical length variation from a database of reconstruction cases (2014).

10. Computer-assisted method according to claim 9, wherein at least two second inputs are weighted with a respective weight (W0b-W2b) and summed to compute the desired length variation, the sum of the weights being equal to 1.

11. Computer-assisted method according to any one of claims 1 to 10, wherein computing the proposed position of the femoral tunnel comprises applying a mathematical function to translate the desired graft position along at least one axis of the coordinate system with respect to the desired length variation.

12. Computer-assisted method according to any one of claims 1 to 11, wherein obtaining the 3D model of the patient’s femur comprises:- obtaining at least one medical image of the patient’s femur;- segmenting said at least one medical image to determine a 3D mesh of the femur; and - registering said 3D mesh with a reference model of the femur to orient the 3D mesh in a coordinate system of said reference model, the registered 3D mesh forming the 3D model of the patient’s femur.

13. Computer-assisted method according to any one of claims 1 to 12, wherein the set of anatomical features comprises an intercondylar roof curve.

14. Computer-assisted method according to claim 13, wherein the coordinate system is a Bernard & Hertel grid.

15. Computer-assisted method according to claim 14, wherein computing the proposed position of the femoral tunnel comprises:- determining from the Bernard & Hertel grid and from the first input a target position of the entry point of the femoral tunnel within a given confidence interval;- determining from the isometry map and from the second input a target length variation interval of the anterior cruciate ligament; and- computing a combination of the confidence interval for the target position of the entry point and of the target length variation interval to determine the optimal position of the femoral tunnel and an allowed area surrounding the optimal position.

16. Computer-assisted method according to any one of claims 13 to 15, wherein determining the intercondylar roof curve comprises:- computing a projection of the 3D model of the patient’s femur in a sagittal plane so that both condyles of the femur are superposed;- computing a mid-sagittal plane cutting the condyles comprising the intercondylar roof curve;- from said projection, determining a set of points oriented along the mid-sagittal plane; - from said determined set of points, extracting points belonging to the intercondylar roof curve; and- computing a line connecting the extracted points to draw the intercondylar roof curve.

17. Computer-assisted method according to claim 16, wherein extracting the points belonging to the intercondylar roof curve comprises:- computing a weight value of each point of the determined set of points to the intercondylar roof curve; and- extracting the points having a weight value greater than a determined threshold.

18. Computer-assisted method according to any one of claims 1 to 17, further comprising displaying the 3D model of the patient’s femur and at least one of: the coordinate system, the isometry map and the proposed position of the femoral tunnel.

19. Computer-assisted method according to claim 18 in combination with claim 6, further comprising displaying a graphical user interface comprising at least one user interface element configured to allow a user to provide the fifth input.

20. Computer-assisted method according to claim 19 in combination with any one of claims 8, 10 and 11 , wherein the user interface element is configured to allow the user to modify at least one weight or parameter of the mathematical function.

21. Computer-assisted method according to claim 15 in combination with claim 6, wherein updating the proposed position of the femoral tunnel comprises displacing the proposed position within the allowed area.

22. Computer-assisted method according to any one of claims 1 to 21, wherein the ligament is an anterior cruciate ligament (ACL), a medial collateral ligament (MCL), a lateral collateral ligament (LCL), a posterior cruciate ligament (PCL) or an anterolateral ligament (ALL).

23. Computer system for planning placement of a femoral tunnel for ligament reconstruction, comprising a control unit configured to implement the method according to any one of claims 1 to 22, and a user interface coupled to the control unit, the user interface being configured to display the 3D model of the patient’s femur and the computed proposed position of the femoral tunnel and to receive user’s inputs.

24. Navigated surgical system for ligament reconstruction, comprising:a surgical tool comprising either a drill or a guide for guiding a drill to drill the patient’s femur,a localization system including navigation trackers configured to be attached to the patient's femur and tibia and to the surgical tool;the computer system of claim 23, wherein the control unit is coupled to the localization system, configured to:implement the method according to any one of claims 1 to 22 to plan the proposed tunnel position and a tunnel orientation;display the proposed tunnel position and orientation in real-time along with the surgical tool;provide feedback to align the surgical tool with an axis of the calculated proposed tunnel placement.

25. Navigated surgical system according to claim 24, further comprising:a robotic arm controlled by the control unit, comprising a plurality of motorized joints and an end-effector configured to drill or to guide a drill;the localization system further comprising a navigation tracker attached to the robotic arm to localize the end-effector in real-time.

Images