ROBOT CATHETER, AUTOMATIC NAVIGATION SYSTEM, AND METHOD FOR AUTOMATIC NAVIGATION OF AN EXTENDED FLEXIBLE MEDICAL INSTRUMENT

Abstract

The invention relates to a catheter robot comprising an automatic navigation system for an elongated flexible medical instrument (28) with a free distal end (29), implementing an automatic navigation method comprising successively: a step of creating a model of the elongated flexible medical instrument (28), a blood circulation network (21), and their interaction; a step of determining a trajectory to follow between the starting point and the arrival point (27); a step of planning a sequence of commands for moving the elongated flexible medical instrument (28), obtained by a tree search procedure; and a step of executing the planned command sequence, with compensation for deviations from the determined trajectory along the modeled network (21), by closed-loop control. Figure 2

Description

Title of the invention: ROBOT CATHETER, AUTOMATIC NAVIGATION SYSTEM, AND NAVIGATION METHOD AUTOMATIC OPERATION OF AN ELONGATED FLEXIBLE MEDICAL INSTRUMENT FIELD OF THE INVENTION

[0001] The invention relates to a catheter robot, an automatic navigation system, and a method for automatic navigation of an elongated flexible medical instrument which is a catheter or a catheter guide, with a free distal end. TECHNOLOGICAL BACKGROUND OF THE INVENTION

[0002] According to a first prior art, for example described in patent application US2020 / 0129740, an automatic navigation system in a robot for an elongated flexible medical instrument is known. However, in this document, automatic navigation for this elongated flexible medical instrument is only envisaged, and it is not specified which algorithm can control the movements of this elongated flexible medical instrument.

[0003] According to a second prior art, for example described in the English-language article entitled "Automatic control of cardiac ablation catheter with deep reinforcement learning method", an automatic navigation system is known in a robot for an elongated flexible medical instrument, but this elongated flexible medical instrument is of a particular type which has a distal end which is steerable ("steerable catheter" according to English terminology), which is in fact an ablation catheter with a steerable tip, which of course makes automatic navigation easier to achieve for this elongated flexible medical instrument having a steerable distal end.Furthermore, to determine the trajectory to be followed by this elongated flexible medical instrument having a steerable distal end, the navigation method uses a deep network reinforcement learning (DQN) method, which method has two drawbacks, which are firstly the need for prior training and secondly the difficulty of generalisation to any type of topology and geometry of a patient's arteries.

[0004] According to a third prior art, for example described in the article in English entitled “Reinforcement learning for guidewire navigation in coronary phantom”, there is known an automatic navigation system in a robot for an elongated flexible medical instrument, but this navigation is only carried out in a two-dimensional (2D) space, the possibility of adapting it for navigation in automatic navigation in a three-dimensional (3D) space remains purely hypothetical, which does not make this automatic navigation a good candidate for evolving in three dimensions (3D) in a real blood circulation network. In addition, to determine the trajectory to be followed by this elongated flexible medical instrument having a steerable distal end, the navigation method uses a deep network reinforcement learning method (DQN, for "Deep Q-Learning" in English, or DDPG for "Deep Deterministic Policy Gradient" in English), which method has two drawbacks which are firstly the need for prior training and secondly the difficulty of generalization to any type of topology and geometry of a patient's arteries.

[0005] In the whole of the prior art considered, the invention notes that there is no automatic navigation system capable of both:

[0006] • to be effective for a wider family of elongated flexible medical instruments, and in particular those whose distal end is free, i.e. is not steerable (unlike, for example, endoscopes whose tip is often steerable, in particular by several cables attached to this tip), • to be easily usable and applicable to a wide range of blood circulation networks for many patients, or even to a wide range of portions of a patient's blood circulation network, • to ensure efficient, rapid and secure control of an optimized trajectory, within a patient's blood circulation network, which is intrinsically interwoven, vast and complex. OBJECTS OF THE INVENTION

[0007] The aim of the present invention is to provide a catheter robot, a navigation system and a navigation method which at least partially overcome the aforementioned drawbacks.

[0008] More particularly, the invention aims to provide, to this end, a navigation system for an elongated flexible medical instrument of a catheter robot implementing an automatic navigation method, as well as the associated catheter robot and the associated automatic navigation method, which performs:

[0009] • a particular model, adapted to elongated flexible medical instruments without a steerable distal end, based on careful monitoring of the evolution of this free, non-steerable distal end, in particular during its potential interaction with a blood vessel wall in the patient's blood circulation network, • in order to obtain automatic navigation which is efficient (rapid and short route in the blood circulation network) and secure (without risk of crossing the wall of a vessel blood), • planning of a sequence of commands for moving the elongated flexible medical instrument which ensures effective tracking of a trajectory determined in an optimized manner in the cleverly modeled network, based on a method which has proven to be particularly simple and effective for this type of route in a vast network with multiple branches, derivations and re-crossings, which is the tree search, • in such a way as to obtain automatic navigation which is efficient (rapid and short route in the blood circulation network) and stable (no permanent or too frequent need for error correction, or even complete recalculation due to divergence).

[0010] According to the invention, there is provided a catheter robot comprising an automatic navigation system, of an elongated flexible medical instrument which is a catheter or a catheter guide, with a free distal end, implementing an automatic navigation method successively comprising: a step of creating a model: of the elongated flexible medical instrument, of a blood circulation network of a patient, along which the elongated flexible medical instrument is intended to move, of the interaction between this network and this elongated flexible medical instrument, by modeling the contact between this elongated flexible medical instrument and one or more walls of blood vessel(s) in this network, from one or more images of the real blood circulation network of this patient, a step of determining: the position of a starting point in the modeled network, the position of an arrival point in the modeled network, a trajectory to follow,by the elongated flexible medical instrument, between the starting point and the arrival point, by determining a path, along the modeled network, between the starting point and the arrival point, preferably with obligatory passage points in the modeled network, a step of planning a sequence of commands for moving the elongated flexible medical instrument, obtained by a tree search procedure which, using the modeling of the elongated flexible medical instrument and the modeling of said contact: simulates different possible movements of the elongated flexible medical instrument in the modeled network, evaluates the results of the simulation of these different possible movements, selects the most promising simulated movements from among these different possible movements, to follow the trajectory best determined with respect to one or more given criteria, a step of executing the planned sequence of commands,along the actual blood circulation network of said patient, with compensation for deviations from the trajectory determined along the modeled network, by closed-loop regulation.

[0011] A catheter with a free distal end is a catheter with a distal end that is not steerable in a given direction. When the distal end of this catheter arrives at a branch in the blood circulation network, this distal end cannot be deformed by an actuator to be directed to one side to go into the branch chosen at the branch, but can only be rotated around its longitudinal axis, which is also its axis of progression. The free distal end of the catheter is also called the tip of the catheter. A catheter with a free distal end is devoid of any mechanism for deforming its distal end.

[0012] A catheter guide with a free distal end is a catheter guide with a distal end that is not steerable in a given direction. When the distal end of this catheter guide arrives at a branch in the blood circulation network, this distal end cannot be deformed by an actuator to be directed to one side to go into the chosen branch at the branch, but can only be rotated around its longitudinal axis, which is also its axis of progression. The free distal end of the catheter guide is also called the tip of the catheter guide. A catheter guide with a free distal end is devoid of any mechanism for deforming its distal end.

[0013] Conversely, a steerable catheter is a catheter whose distal end can be controlled by the practitioner so that he can steer the distal end of the catheter in the desired manner. The practitioner controls the catheter by deforming the distal end via a deformation mechanism, for example cables which are pulled or released depending on the desired deformation.

[0014] The images of the actual blood circulation network of this patient may include or be, for example, radiographs, X-ray images, and in particular be two-dimensional (2D) images. The images of the actual blood circulation network of the patient may also include a 3D Coroscan (for "Coronary Computed Tomography Scan" according to English terminology), and / or an MRI. The images of the actual blood circulation network of the patient may include three-dimensional (3D) images. The three-dimensional images may be constructed from two-dimensional images.

[0015] According to the invention, there is also provided a catheter robot comprising an automatic navigation system, of an elongated flexible medical instrument which is a catheter or a catheter guide, with free distal end, implementing an automatic navigation method successively comprising: a step of creating a model: of the elongated flexible medical instrument, of a blood circulation network of a patient, along which the elongated flexible medical instrument is intended to move, of the interaction between this network and this elongated flexible medical instrument, by modeling the contact between this elongated flexible medical instrument and one or more walls of blood vessel(s) in this network, from one or more images of the real blood circulation network of this patient, a step of determining: the position of a starting point in the modeled network, the position of an arrival point in the modeled network, a trajectory to be followed, by the elongated flexible medical instrument, between the starting point and the arrival point, by determining a path, along the modeled network, between the starting point and the arrival point, preferably with obligatory passage points in the modeled network, a step of planning a sequence of commands for moving the elongated flexible medical instrument, obtained by a tree search procedure which,using the modeling of the elongated flexible medical instrument and the modeling of said contact: simulates different possible movements of the elongated flexible medical instrument in the modeled network, evaluates the results of the simulation of these different possible movements, selects the most promising simulated movements among these different possible movements, to follow the trajectory determined as best as possible in relation to one or more given criteria.

[0016] According to the invention, there is also provided an automatic navigation system, of an elongated flexible medical instrument which is a catheter or a catheter guide, of a catheter robot, with free distal end, implementing an automatic navigation method successively comprising: a step of creating a model: of the elongated flexible medical instrument, of a blood circulation network of a patient, along which the elongated flexible medical instrument is intended to move, of the interaction between this network and this elongated flexible medical instrument, by modeling the contact between this elongated flexible medical instrument and one or more walls of blood vessel(s) in this network, from one or more images of the real blood circulation network of this patient, a step of determining: the position of a starting point in the modeled network, the position of an arrival point in the modeled network, a trajectory to follow,by the elongated flexible medical instrument, between the starting point and the arrival point, by determining a path, along the modeled network, between the starting point and the arrival point, preferably with obligatory passage points in the modeled network, a step of planning a sequence of commands for moving the elongated flexible medical instrument, obtained by a tree search procedure which, using the modeling of the elongated flexible medical instrument and the modeling of said contact: simulates different possible movements of the elongated flexible medical instrument in the modeled network, evaluates the results of the simulation of these different possible movements, selects the most promising simulated movements among these different possible movements, to follow the trajectory best determined by, in relation to one or more given criteria, a step of executing the planned sequence of commands, along the real blood circulation network of said patient, with compensation for deviations from the trajectory determined along the modeled network, by closed-loop regulation.

[0017] According to the invention, there is then provided an automatic navigation system, of an elongated flexible medical instrument which is a catheter or a catheter guide, of a catheter robot, with free distal end, implementing an automatic navigation method successively comprising: a step of creating a model: of the elongated flexible medical instrument, of a blood circulation network of a patient, along which the elongated flexible medical instrument is intended to move, of the interaction between this network and this elongated flexible medical instrument, by modeling the contact between this elongated flexible medical instrument and one or more walls of blood vessel(s) in this network, from one or more images of the real blood circulation network of this patient, a step of determining: the position of a starting point in the modeled network, the position of an arrival point in the modeled network, a trajectory to follow,by the elongated flexible medical instrument, between the starting point and the arrival point, by determining a path, along the modeled network, between the starting point and the arrival point, preferably with obligatory passage points in the modeled network, a step of planning a sequence of commands for moving the elongated flexible medical instrument, obtained by a tree search procedure which, using the modeling of the elongated flexible medical instrument and the modeling of said contact: simulates different possible movements of the elongated flexible medical instrument in the modeled network, evaluates the results of the simulation of these different possible movements, selects the most promising simulated movements among these different possible movements, to follow the trajectory best determined with respect to one or more given criteria.

[0018] According to the invention, there is also provided a method for automatic navigation of an elongated flexible medical instrument which is a catheter or a catheter guide, with a free distal end, of a catheter robot, successively comprising: a step of creating a model: of the elongated flexible medical instrument, of a blood circulation network of a patient, along which the elongated flexible medical instrument is intended to move, of the interaction between this network and this elongated flexible medical instrument, by modeling the contact between this elongated flexible medical instrument and one or more walls of blood vessel(s) in this network, from one or more images of the real blood circulation network of this patient, a step of determining: the position of a starting point in the modeled network, the position of an arrival point in the modeled network, a trajectory to be followed, by the elongated flexible medical instrument, between the starting point and the arrival point, by determining a path, along the modeled network, between the starting point and the arrival point, preferably with obligatory passage points in the modeled network, a step of planning a sequence of commands for moving the elongated flexible medical instrument, obtained by a tree search procedure which, using the modeling of the elongated flexible medical instrument and the modeling of said contact: simulates different possible movements of the elongated flexible medical instrument in the modeled network, evaluates the results of the simulation of these different possible movements, selects the most promising simulated movements from among these different possible movements, to follow the trajectory best determined with respect to one or more given criteria, a step of executing the planned sequence of commands,along the actual blood circulation network of said patient, with compensation for deviations from the trajectory determined along the modeled network, by closed-loop regulation.

[0019] According to the invention, there is finally provided a method for automatic navigation of an elongated flexible medical instrument which is a catheter or a catheter guide, with a free distal end, of a catheter robot, successively comprising: a step of creating a model: of the elongated flexible medical instrument, of a blood circulation network of a patient, along which the elongated flexible medical instrument is intended to move, of the interaction between this network and this elongated flexible medical instrument, by modeling the contact between this elongated flexible medical instrument and one or more walls of blood vessel(s) in this network, from one or more images of the real blood circulation network of this patient, a step of determining: the position of a starting point in the modeled network, the position of an arrival point in the modeled network, a trajectory to be followed, by the elongated flexible medical instrument,between the starting point and the arrival point, by determining a path, along the modeled network, between the starting point and the arrival point, preferably with obligatory passage points in the modeled network, a step of planning a sequence of commands for moving the elongated flexible medical instrument, obtained by a tree search procedure which, using the modeling of the elongated flexible medical instrument and the modeling of said contact: simulates different possible movements of the elongated flexible medical instrument in the modeled network, evaluates the results of the simulation of these different possible movements, selects the most promising simulated movements from among these different possible movements, to follow the trajectory best determined with respect to one or more given criteria.

[0020] According to preferred embodiments, the invention comprises one or more of the following features which may be used separately or in combination: partial between them or in total combination between them, with any of the aforementioned objects of the invention.

[0021] Preferably, said automatic navigation method is capable of being implemented in real time, said tree search procedure of the planning step, simultaneously performs in parallel, for several different possible movements of the elongated flexible medical instrument in the modeled network, for at least 2 or at least 3 or at least 4 different possible movements, the following planning cycle: simulate a possible movement of the elongated flexible medical instrument in the modeled network, evaluate the result of the simulation of this possible movement, select the simulated movement, if it is one of the most promising among the different possible movements, to follow the determined trajectory, at best with respect to one or more given criteria.

[0022] Thus, not only will the automatic navigation used be efficient, in the sense that a trajectory determined in an optimized manner will be followed as closely as possible, but also, this route will be carried out in a significantly shorter time than expected, thanks to the parallelization of the route through the tree structure, which will certainly involve a little extra work but will save significant and particularly valuable time for an intervention on a patient to be carried out in real time. The tree search procedure lends itself particularly well to this parallelization of the route through the tree structure, which is not necessarily the case for a certain number of other techniques for determining and traveling an optimized trajectory.

[0023] Preferably, if at the end of said execution step, one or more deviations to be compensated are evaluated as too large, with respect to one or more given criteria: then a recalibration of the modeling is carried out, and on the recalibrated modeling, the following are then carried out: first a new step of determining a new trajectory, then a new step of planning a new sequence of commands for moving the elongated flexible medical instrument, from the new determined trajectory, finally, if necessary, a new step of executing the new planned sequence of commands.

[0024] Thus, in the case of either a particularly long trajectory or a journey through a particularly tangled portion of the patient's blood circulation network, this combination of a recalibration of the modeling and an iteration of certain of the steps in the automatic navigation process will ensure:

[0025] • on the one hand, convergent and tight guidance along the determined trajectory, • and on the other hand, to significantly limit the quantity of calculation and simulation to be carried out, by allowing the use of simplifications in the simulation or in the exploration of the tree search algorithm.

[0026] Thus, a good compromise can be achieved between a relative simplicity of the model used and a relatively low frequency of recalibrations, while reducing, or even eliminating, the risk of seeing the system diverge (due to excessively large gaps between reality and its modeling).

[0027] Preferably, said modeling creation step performs a representation of the physical system encompassing both the elongated flexible medical instrument and the blood circulation network of a patient, along which the elongated flexible medical instrument is intended to move, by finite element simulation.

[0028] Thus, this type of method works particularly well for the type of physical system to be modeled, and in particular the type of many small successive movements of the flexible medical instrument lying in the patient's blood circulation network.

[0029] Preferably, the elongated flexible medical instrument is modeled by Kirchhoff beams or by Cosserat beams.

[0030] Thus, this type of method works particularly well for modeling an elongated flexible medical instrument, as long, flexible and thin as a catheter or a catheter guide.

[0031] Preferably, the patient's blood circulation network is modeled: either by point clouds, or by meshes, or by central lines associated with their respective diameters, or by implicit surfaces.

[0032] Thus, this type of method works particularly well for modeling blood vessels in a patient's blood circulation network.

[0033] Preferably, the modeling of the patient's blood circulation network integrates the movements affecting this patient's blood circulation network, and preferentially integrates the deformations of the patient's heart when it beats and / or the deformations induced by the patient's breathing.

[0034] Thus, the automatic navigation carried out will be dynamic, and no longer just static, by integrating changes in the geometry of the patient's blood circulation network.

[0035] Preferably, in said determining step: said starting point in the modeled network is positioned at the exit of the guide catheter of the catheter robot, at the ostium, and said arrival point in the modeled network is positioned in the patient, at the lesion to be treated.

[0036] Thus, the elongated flexible medical instrument can be easily guided throughout the entire path it will travel.

[0037] Preferably, in said determining step: the trajectory to be followed is determined by interpolation between said starting point and said arrival point, and preferably by using a Dijkstra graph traversal algorithm applied to the central lines of the blood vessels of the blood circulation network.

[0038] Thus, this type of method works particularly well for determining a trajectory within a particularly vast and complex network of possible pathways such as a patient's blood circulation network.

[0039] Preferably, said tree search procedure evaluates the results of the simulation of these different possible movements by assigning scores to each branch of the tree, then by calculating a value for each node of the tree from the scores assigned to the branches leading to this node, selects the most promising simulated movements from among these different possible movements which are the simulated movements leading to the node of the highest value, to guide its exploration in the tree.

[0040] Thus, the tracking of an optimized trajectory in a shortened time will be further improved.

[0041] Preferably, said tree search procedure assigns the scores as follows: if the elongated flexible medical instrument advances along the determined trajectory, then the corresponding possible displacement of the elongated flexible medical instrument in the modeled network receives a positive score, if the elongated flexible medical instrument advances elsewhere than along the determined trajectory or if the elongated flexible medical instrument moves backward or if the elongated flexible medical instrument stagnates along the determined trajectory, then the corresponding possible displacement of the elongated flexible medical instrument in the modeled network receives a zero or negative score.

[0042] Preferably, said tree search procedure assigns the scores in the following manner: if the possible movement of the elongated flexible medical instrument along the determined trajectory is considered to be similar to that which would result from the action of a doctor chosen as a reference, then this possible movement receives a positive score, if the possible movement of the elongated flexible medical instrument along the determined trajectory is considered to be different from that which would result from the action of a doctor chosen as a reference, then this possible movement receives either a zero score or a negative score.

[0043] Preferably, said tree search procedure assigns, in part, the scores in the following first way: if the elongated flexible medical instrument advances along the determined trajectory, then the corresponding possible displacement of the elongated flexible medical instrument in the modeled network receives a positive score, if the elongated flexible medical instrument advances elsewhere than along the determined trajectory or if the elongated flexible medical instrument moves backward or if the elongated flexible medical instrument stagnates along the determined trajectory, then the corresponding possible displacement of the elongated flexible medical instrument in the modeled network receives a zero or negative score, also assigns, in part, the scores of the second following manner: if the possible displacement of the elongated flexible medical instrument along the determined trajectory is considered to be similar to that which would result from the action of a doctor chosen as a reference, then this possible displacement receives a positive score, if the possible displacement of the elongated flexible medical instrument along the determined trajectory is considered to be different from that which would result from the action of a doctor chosen as a reference, then this possible displacement receives either a zero score or a negative score, uses, with preferentially predetermined weightings, possibly different from each other, respectively the first manner and the second manner.

[0044] Thus the wealth of criteria for evaluating possible movements for the elongated flexible medical instrument will improve the probability of having an even more optimized trajectory and of traveling it even more quickly.

[0045] Preferably, the value of each node is calculated by accumulating the scores of each branch leading to this node, or the value of each node is calculated according to the UCB method (UCB means “Upper Confidence Bound” in English).

[0046] Preferably, said tree search procedure returns to pass through a node of lesser value, after a node of better value has in fact turned out to be a dead end. A node which turns out to be a dead end is a node which, promising at the start, does not keep its promises and it is in fact not interesting for the elongated flexible medical instrument to follow the corresponding path.

[0047] Thus, the guidance is made even more convergent and tight along the determined trajectory, without the risk of remaining blocked and then being obliged to restart a new trajectory calculation which could have been avoided.

[0048] Preferably, said tree search procedure is an algorithm:

[0049] • either Monte Carlo search tree (MCTS, for “Monte Carlo Tree Search » in English), • either the upper limit of the confidence interval applied to trees (UCT for “Upper Confidence Bound applied to Trees” in English) • either optimistic planning for deterministic systems (OPD, for “Optimistic Planner for Deterministic Systems” in English), • or open loop optimistic planning (OLOP, for “Open Loop Optimistic Planner” in English).

[0050] Preferably, said step of executing the planned sequence of commands is carried out by direct command carrying out the movements of the actuators controlling the movements of the elongated flexible medical instrument.

[0051] Thus, the order is more simply calculated.

[0052] Preferably, said step of executing the planned sequence of commands is carried out by inverse control calculating the movements of the actuators from the movements of the elongated flexible medical instrument.

[0053] Thus, the order is more precisely calculated.

[0054] Preferably, the blood vessels of the blood circulation network are arteries of the blood circulation network. These blood vessels could however be veins, or arteries and veins.

[0055] Preferably, the catheter robot comprises an automatic navigation system, of several elongated flexible medical instruments among which there are at least one catheter and a catheter guide associated with this catheter, these elongated flexible medical instruments having a free distal end.

[0056] Thus, the guidance provided by this automatic navigation is more complete.

[0057] The catheter may be a so-called "stent" or "balloon" catheter which is hollow. The guide catheter can be a metal wire that is solid.

[0058] Other characteristics and advantages of the invention will appear on reading the following description of a preferred embodiment of the invention, given by way of example and with reference to the appended drawings. Brief description of the drawings

[0059] [Fig.l] [Fig.l] schematically represents an example of a system including a catheter robot with a patient installed in a medical imaging device.

[0060] [Fig.2] [Fig.2] schematically represents an example of simulation of the navigation of a catheter guide in a model of coronary arteries of a patient, in an automatic navigation method according to an embodiment of the invention.

[0061] [Fig.3] [Fig.3] schematically represents an example of representation of 5 nodes of the tree connected by the 4 possible actions, in an automatic navigation method according to an embodiment of the invention.

[0062] [Fig.4] [Fig.4] schematically represents an example of a first step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0063] [Fig.5] [Fig.5] schematically represents an example of a second step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0064] [Fig.6] [Fig.6] schematically represents an example of a third step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0065] [Fig.7] [Fig.7] schematically represents an example of a fourth step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0066] [Fig.8] [Fig.8] schematically represents an example of a fifth step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0067] [Fig.9] [Fig.9] schematically represents an example of a sixth stage of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0068] [Fig. 10] [Fig. 10] schematically represents an example of a seventh step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0069] [Fig. 11] [Fig. 11] schematically represents an example of an eighth step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0070] [Fig. 12] [Fig. 12] schematically represents an example of a ninth step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0071] [Fig. 13] [Fig. 13] schematically represents an example of a tenth step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0072] [Fig. 14] [Fig. 14] schematically represents an example of an eleventh step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0073] [Fig. 15] [Fig. 15] schematically represents an example of a twelfth step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0074] [Fig. 16] [Fig. 16] schematically represents an example of a step of a second type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0075] [Fig. 17] [Fig. 17] schematically represents an example of a step which follows the step of [Fig. 16].

[0076] [Fig. 18] [Fig. 18] schematically represents an example of a step which follows the step of [Fig. 17].

[0077] [Fig. 19] [Fig. 19] schematically represents an example of a first step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0078] [Fig.20] [Fig.20] schematically represents an example of a second step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0079] [Fig.21] [Fig.21] schematically represents an example of a third step of a third type of use of a tree search algorithm, in a automatic navigation method according to one embodiment of the invention.

[0080] [Fig.22] [Fig.22] schematically represents an example of a fourth step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0081] [Fig.23] [Fig.23] schematically represents an example of a fifth step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0082] [Fig.24] [Fig.24] schematically represents an example of a sixth step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0083] [Fig.25] [Fig.25] schematically represents an example of a seventh step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0084] [Fig.26] [Fig.26] schematically represents an example of an eighth step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0085] [Fig.27] [Fig.27] schematically represents an example of a ninth step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0086] DETAILED DESCRIPTION OF THE EMBODIMENTS OF THE INVENTION

[0087] The invention proposes an automation of the navigation of one or more flexible elongated medical instruments, for example a catheter guide and a catheter. This automation will successively carry out the following steps. First, a first step of creating a model of the elongated flexible medical instruments, for example guide, catheter, catheter-guide, and their environment, for example the arteries and / or veins of the patient's blood network, in which the elongated flexible medical instruments will move. Then, a second step of defining a starting point, for example the ostium, and an arrival point, for example a distal part of the artery close to the lesion to be treated in the patient, in the model generated in the previous step, so as to obtain a trajectory to follow for the elongated flexible medical instrument in the patient's blood network, for example by following a path along the patient's arteries.Then, a third step of planning a sequence of commands to be executed so that the elongated flexible medical instrument follows the trajectory defined previously, this planning being carried out by a tree search of the different movements of the elongated flexible medical instrument, this planning being carried out in the model generated previously. Finally, a fourth step of executing the sequence of commands defined previously on the robot. catheter.

[0088] In the first step, the model that will be constructed during this first step is a representation of the physical system comprising the arteries of the patient's blood network, obtained from medical imaging of the patient, and the elongated flexible medical instruments that navigate in these arteries of the patient's blood network. Several methods can be used to generate this model.

[0089] In one variant used, the model is a finite element simulator, of the FEM simulation type. The arteries of the patient's blood network are represented by a 3D geometric model (in three dimensions) such as, for example, point cloud, mesh, central line and diameter, implicit surface. The elongated flexible medical instruments are modeled by beams, for example by Kirchhoff beams or by Cosserat beams.

[0090] The model can integrate movements of the environment, for example deformations of the heart as it beats.

[0091] In the second step, the starting point and the end point are 3D coordinates defined in the previously generated model. The starting point generally corresponds to the exit of the guide catheter at the ostium. The end point generally corresponds to the lesion to be treated, more precisely to a location chosen in the lesion to be treated.

[0092] The trajectory is obtained by interpolation between the starting point and the arrival point. The trajectory can for example be determined by a Dijkstra graph traversal algorithm applied to the central line of the arteries of the patient's blood network.

[0093] In the third step, the planning of the sequence of commands is obtained by tree search, which is also called Predictive Control (or “Model Predictive Control” in English).

[0094] This tree search is performed by exploring command sequences in the model to predict their effects, and thus find a suitable command sequence.

[0095] The algorithm can for example simulate each possible movement from a starting position, and each of the states of the model, a state of the model including the position of the elongated flexible medical instrument as well as the position of the arteries of the patient's blood network, resulting from these commands are noted in order to explore in priority the states with the best score. The score given by the algorithm can also be called reward.

[0096] The algorithm can for example interact with its environment by 4 discrete actions which are:

[0097] • either advance by N mm (millimeters), • either rotate by M radians, • either rotate by -M radians, • either move back by N mm, • N and M being advantageously predefined constants.

[0098] The different states of the physical system constitute the nodes of a tree to be traversed. The branches of this tree are the actions allowing one to move from one node to another, and therefore from one state of the physical system to another.

[0099] Each node of the tree is associated with a value evaluating the potential of each node such that a node with a higher value has a greater chance of being on the optimal trajectory between the starting point and the arrival point than a node with a lower value. Preferably, this value is determined in the following manner:

[0100] Vnode = Vparent + gd 1 * Reward + (gd / (1-g))

[0101] With Vnoeud which is the value of the current node, Vparent which is the value of the parent node, g which is the discount factor, d which is the depth of the node in the tree, Recompense which is the reward corresponding to the action going from the parent node to the current node.

[0102] In a first possible variant of the third step, these scores are determined in the following manner. If the elongated flexible medical instrument advances along the trajectory defined in the second step, then it receives a positive score. In all other cases, that is to say if the elongated flexible medical instrument advances in an arterial branch not corresponding to the trajectory defined in the second step, or if the elongated flexible medical instrument moves backward, then it receives a zero or negative score.

[0103] In a second possible variant of the third step, the marks (or rewards) are awarded so that the movements which are closest to those which would be carried out by a doctor obtain the best marks ("Inverse Reinforcement Learning" in English), and the movements which would not be selected by a doctor obtain zero or negative marks.

[0104] The algorithm assigns a score-based value to each node in the tree and uses this value to guide its exploration, with exploration being performed on the highest-value nodes in priority.

[0105] According to a first possible option, this value can be the cumulative reward of the node and all its parents in the tree. According to a second possible option, the value is determined according to the UCB method (for "Upper Confidence Bound" in English).

[0106] If a branch that initially seemed promising is ultimately a "dead end", we go back to branches that initially had a lower value, therefore less promising initially and which had therefore not been chosen at that time.

[0107] In a third possible variant of the third step, the exploration of the movements by the tree search is carried out not systematically, but with a predetermined algorithm, for example resulting from end-to-end learning. The exploration of the movements by the tree search can also be carried out by imitating the movements that would have been carried out by a doctor. In this third variant, the exploration of the movements is carried out for A% of the time using determined values ​​as in the first variant, and B% of the time by trying to imitate the movements of a doctor as in the second variant. The third variant amounts for example to combining 50% of the time the first variant and 50% of the time the second variant, or for example to combining 80% of the time the first variant and 20% of the time the second variant.

[0108] Examples of tree search algorithms are:

[0109] • or Monte Carlo search tree (MCTS, for “Monte Carlo Tree Search” in English), • either the upper limit of the confidence interval applied to trees (UCT for “Upper Confidence Bound applied to Trees” in English) • either optimistic planning for deterministic systems (OPD, for “Optimistic Planner for Deterministic Systems” in English), • or open loop optimistic planning (OLOP, for “Open Loop Optimistic Planner” in English).

[0110] In the tree search algorithm used here, the development of the tree is done in 3 phases:

[0111] • First a selection, where the node with the highest value is chosen, • Then an expansion, where all possible actions from this chosen node are simulated and new child nodes are created, • Then a backpropagation, where the values ​​of the newly created child nodes are calculated from the rewards associated with each action. These values ​​are then propagated first to the parent nodes, and then recursively to the root node, • These 3 steps can be repeated several times. Advantageously, the number of repetitions of these three steps is predefined, thus ensuring that the algorithm converges in a finite time. This number of repetitions is called the budget.

[0112] In the fourth step, the sequence of commands planned in the third step is executed by the catheter robot in order to move the elongated flexible medical instrument. This execution can be carried out by a direct control algorithm, in which the command is a movement of the actuators, or by a inverse control in which the control is a movement of the elongated flexible medical instrument from which the corresponding movements of the actuators are calculated.

[0113] A closed-loop control system is used to execute the command sequence. This closed loop uses the medical imaging system to track over time the position of the elongated flexible medical instrument that is moved by the catheter robot, and also to verify that the elongated flexible medical instrument follows the trajectory defined during the second step. If position deviations are observed, then there are two possibilities which are either to be able to compensate for these deviations with the closed-loop control, or if these deviations are too large, to carry out a replanning of the command sequence from the current state of the system which becomes the new starting point for the second step which second step is then repeated.

[0114] The use of the closed loop with medical imaging can make it possible not to plan all the commands to go to the arrival point, but to plan only the commands on a certain step. In this case, it is considered that, in particular because of the simplifications of the model, after several time steps it would no longer correspond entirely to reality, and that then it could be more advantageous to carry out a recalibration with medical imaging and then a replanning.

[0115] [Fig. 1] schematically represents an example of a system including a catheter robot with a patient installed in a medical imaging device.

[0116] The patient 1 is for example lying on a bed or an examination table 2 to which is attached a catheter robot 4 controlled by a control unit 5, from the images of the patient 1 which are obtained by a medical imaging system 3.

[0117] [Fig.2] schematically represents an example of simulation of the navigation of a catheter guide in a model of coronary arteries of a patient, in an automatic navigation method according to an embodiment of the invention.

[0118] The patient's blood network 21 comprises a main artery 22 which will gradually divide into several secondary arteries 23, 24, 25, 26. The arrival point, i.e. the patient's lesion to be treated, is point 27 located along the secondary artery 23. The catheter guide 28 has a curved distal end 29 which is located on the main artery 22 at a first branch 51. At this first branch 51, the curved distal end 29 of the catheter guide 28, which descends (in [Fig. 2]) along the main artery 22, will have to take the right branch relative to its direction of movement, towards the secondary arteries 23 and 24, and leave aside the left branch relative to its direction of movement, which would go towards the secondary arteries 25 and 26. After having progressed further along the main artery 22, from the first branch 51, the curved distal end 29 of the guide of catheter 28 will arrive at a second branch 52. At this second branch 52, the curved distal end 29 of the catheter guide 28, which continues to descend (in [Fig.2]) along the main artery 22, will have to take the right branch relative to its direction of movement, towards the secondary artery 23, and leave aside the left branch relative to its direction of movement, which would go towards the secondary artery 24. After having progressed further along the secondary artery 23, from the second branch 52, the curved distal end 29 of the catheter guide 28 will arrive at its arrival point 27 in the lesion of the patient to be treated.

[0119] [Fig.3] schematically represents an example of representation of 5 nodes of the tree connected by the 4 possible actions, in an automatic navigation method according to an embodiment of the invention.

[0120] From position 30, on which the arteries 35 of the patient's blood network are seen, and the curved distal end 37 of a catheter guide 36 which progresses along these arteries 35, towards its arrival point 38 which is located in the lesion of the patient to be treated. From this position 30, 4 actions are possible for the curved distal end 37 of a catheter guide 36:

[0121] • either advance along these arteries 35, which is shown in position 31, where we see the curved distal end 37 of a catheter guide 36 which has advanced towards its arrival point 38, • either rotate in one direction around its longitudinal axis, which is shown in position 32, where we see the curved distal end 37 of a catheter guide 36 which has remained in place relative to its arrival point 38, but which has rotated around the longitudinal axis of the catheter guide 36, in one direction, for example clockwise, for example to better approach the next branch or the next artery bend 35, or to better advance thereafter, • either turn in the other direction around its longitudinal axis, which is shown in position 33, where we see the curved distal end 37 of a catheter guide 36 which has remained in place relative to its arrival point 38, but which has turned around the longitudinal axis of the catheter guide 36, in the other direction, for example counterclockwise, for example to better approach the next branch or the next artery bend 35, or to better move back afterwards, • or move back along these arteries 35, which is shown in position 34, where we see the curved distal end 37 of a catheter guide 36 which has moved back away from its arrival point 38.

[0122] Figures 4 to 27 present a tree search algorithm which is simplified, and aim to describe the steps of: selection, expansion and reverse pro- pagation. Three variants are presented: sequential exploration (Figures 4 to 15), parallel exploration based on one criterion (Figures 16 to 18), and parallel exploration based on several criteria (Figures 19 to 27). The rules of this tree search algorithm are:

[0123] - Selection: Select the leaf node(s) (node ​​without children) with the highest value; - Expansion: Systematically open all children of the selected leaf node(s); - Backpropagation: The value of a node without children (leaf node) is the reward received when taking the action from its parent to this node, the value of a node with children is the maximum value of the subtree whose root is this node.

[0124] Figures 4 to 15 schematically represent an example of the different steps of a first type of use of a tree search algorithm, in a method of automatic navigation of an elongated flexible medical instrument within a patient's blood network according to an embodiment of the invention.

[0125] This first type of use corresponds to a sequential use of the tree search algorithm, the expansion of the tree is done in 3 phases:

[0126] • First a selection, where the leaf node with the highest value is chosen, • Then an expansion, where all possible actions from this chosen node are simulated and new child nodes are created • Then a backpropagation, where the rewards associated with each action are retrieved and propagated first to the parent nodes, and then recursively to the root node, • These 3 steps can be repeated several times. Advantageously, the number of repetitions of these three steps is predefined, thus ensuring that the algorithm converges in a finite time. This number of repetitions is called the budget.

[0127] [Fig.4] schematically represents an example of a first step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0128] The different nodes 10 are divided into different levels, here only the root level 11 is represented. Its default value is 0. This node being the only one, it is necessarily this one which is selected.

[0129] [Fig.5] schematically represents an example of a second step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0130] The different nodes 10 are divided into different levels, the root level 11, the first level of filiation 12 (also called depth 1).

[0131] The passage between two nodes, of different filiation, therefore from a parent type filiation level to a child type filiation level, for example from the root level 11 to the first filiation level 12, is an action 20 (to be chosen from the 4 actions already presented which are: move forward or turn in one direction or turn in the other direction or move backward).

[0132] From the root node, the expansion of possible actions 20 is carried out, resulting in 4 new child nodes at the first level of filiation 11.

[0133] [Fig.6] schematically represents an example of a third step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0134] Once the values ​​have been determined for these 4 new child nodes at the first level of filiation 11, of this parent node which is the root node, namely from left to right, the values ​​3, 2, 1 and 3, the highest of these values, namely the value 3 of the first node or the fourth node (starting from the left of [Fig.6]) of the first level of filiation 12, is back-propagated:

[0135] • to its parent node which is the root node, the value of which then changes from 0 to 3.

[0136] [Fig.7] schematically represents an example of a fourth step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0137] The different nodes 10 are divided into different levels, the root level 11, the first level of filiation 12 (also called depth 1).

[0138] The "leaf" node (i.e. the nodes that have no children in the tree) of the first level of filiation 12 that is selected is the rightmost node in [Fig.7], with the highest value 3, because in fact the other leaf nodes have lower values ​​(or equal, we could also have selected the leftmost node of the first level of filiation 12, which is also worth 3), which are respectively 3, 2 and 1

[0139] [Fig.8] schematically represents an example of a fifth step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0140] The different nodes 10 are divided into different levels, the root level 11, the first level of filiation 12, the second level of filiation 13.

[0141] The passage between two nodes, of different filiation, therefore from a parent type filiation level to a child type filiation level, for example from the root level 11 to the first filiation level 12, or from the first filiation level 12 to the second filiation level 13, is an action 20.

[0142] From the leaf node selected at the first level of filiation 12, the node further to the right on [Fig.8] with the highest value 3, we carry out the expansion of possible actions 20, resulting in 4 new child nodes at the second level of filiation 13.

[0143] [Fig.9] schematically represents an example of a sixth step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0144] Once the values ​​have been determined for these 4 new child nodes at the second level of filiation 13 (namely from left to right the values ​​0, 0, 2 and 0), the highest of these values ​​(namely the value 2 of the third node from the left of [Fig.9] of the second level of filiation 13) is back-propagated:

[0145] • to its parent node of the first level of filiation 12, the value of which then changes from 3 to 2, • but not towards the node of root level 11, whose value remains at 3, because there still remains the leftmost node, of the first level of filiation 12, with value 3.

[0146] [Fig. 10] schematically represents an example of a seventh step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0147] The different nodes 10 are divided into different levels, the root level 11, the first level of filiation 12 (also called depth 1), the second level of filiation 13 (also called depth 2).

[0148] The leaf that is now selected is the leftmost node in [Fig.10], with the highest value 3 located in the first level of filiation 12. Indeed, the other leaves have lower values ​​which are respectively 2, 1, 0, 0, 2, 0.

[0149] [Fig. 11] schematically represents an example of an eighth step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0150] From the leaf node selected at the first level of filiation 12, the leftmost node on [Fig.l 1], with the highest value 3, the expansion of the possible actions 20 is carried out, resulting in 4 new child nodes at the second level of filiation 13, of this parent node previously selected at the first level of filiation 12.

[0151] [Fig. 12] schematically represents an example of a ninth step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0152] Once the values ​​have been determined for these 4 new child nodes at the second level of filiation 13, namely from left to right the values ​​4, 2, 2 and 2, the highest of these values ​​(4) is back-propagated:

[0153] • to its parent node of the first level of filiation 12, the value of which passes then from 3 to 4, • as well as towards the node of the root level 11, the value of which changes from 3 to 4.

[0154] [Fig. 13] schematically represents an example of a tenth step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0155] The leaf node that is selected is the node with value 4 located at the second level of filiation 13 and which is the leftmost in [Fig. 13]. Indeed the other leaf nodes have lower values ​​which are respectively 2, 2, 2, 2, 1, 0, 0, 2 and 0. .

[0156] [Fig. 14] schematically represents an example of an eleventh step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0157] Starting from the leaf node selected at the second level of filiation 13, the leftmost node in [Fig.14], with the highest value 4, the expansion of the possible actions 20 is carried out, resulting in 4 new child nodes at the third level of filiation 14.

[0158] [Fig. 15] schematically represents an example of a twelfth step of a first type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0159] Once the values ​​have been determined for these 4 new child nodes at the third level of filiation 14 (namely from left to right, the values ​​0, 5, 0 and 0) the highest of these values ​​(namely the value 5 of the second node from the left of [Fig. 15] of the third level of filiation 14) is back-propagated:

[0160] • to its parent node of the second level of filiation 13, the value of which then goes from 4 to 5, • as well as to its grandparent node of the first level of filiation 12, whose value goes from 4 to 5, • as well as towards the node of the root level 11, the value of which changes from 4 to 5.

[0161] Figures 16 to 18 schematically represent an example of the different stages of a second type of use of a tree search algorithm, in the context of the automatic navigation of an elongated flexible medical instrument within a patient's blood network according to an embodiment of the invention.

[0162] This second type of use corresponds to a parallel use based on the value of the leaf nodes. The operation is similar to the sequential operation, but instead of selecting only the leaf node with the highest value, several leaf nodes with the highest values ​​are selected. Parallel tree traversal allows more leaves to be explored in the same budget as traversing the tree sequentially. The value of the number of selected leaf nodes is linked to the available computing resources. Here, we will consider a selection of the 2 highest value leaf nodes, with an expansion of their 8 children, the recovery of rewards and the propagation of the value towards the root of the tree.

[0163] [Fig. 16] schematically represents an example of a step of a second type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention. The preceding steps are not re-described here for the purposes of simplification. The steps preceding [Fig. 16] may have been carried out according to a sequential exploration, or a parallel exploration.

[0164] The two selected leaf nodes are: on the one hand the leaf node with value 4 which is located at the second level of filiation 13 (the leftmost leaf node on [Fig. 16] on the second level of filiation 13); and on the other hand the leaf node with value 3 which is located at the second level of filiation 13 (the second rightmost leaf node on [Fig. 16] at the second level of filiation 13). Indeed, these two leaf nodes are the two leaf nodes with the highest values, the other leaf nodes having respectively the following values ​​2, 2, 2, 2, 1, 0, 0, 0 (from left to right).

[0165] [Fig. 17] schematically represents an example of a step of a second type of use of a tree search algorithm which follows the step illustrated in [Fig. 16], in an automatic navigation method according to an embodiment of the invention.

[0166] From the first selected leaf node (of value 4) which is at the second level of filiation 13, the expansion of the possible actions 20 is carried out, resulting in 4 new child nodes at the third level of filiation 14.

[0167] In parallel with this first expansion, from the second selected leaf node (of value 3) which is at the second level of filiation 13, the expansion of the possible actions 20 is carried out, resulting in 4 new child nodes at the third level of filiation 14.

[0168] [Fig. 18] schematically represents an example of a sixth step of a second type of use of a tree search algorithm which follows the step of [Fig. 17], in an automatic navigation method according to an embodiment of the invention.

[0169] The values ​​are determined for these 4 new child nodes at the third level of filiation 14 of the first parent node previously selected at the second level of filiation 13, the values ​​being 0, 5, 0 and 0 (from left to right). Once the values ​​have been determined for these 4 new child nodes at the third level of filiation 14, the highest of these values ​​(namely the value 5 of the second node from the left of [Fig. 18] of the third level of filiation 14) is back-propagated:

[0170] • to its parent node of the second level of filiation 13, the value of which then goes from 4 to 5, • as well as to its grandparent node of the first level of filiation 12, whose value goes from 4 to 5, • as well as to the root level node 11, whose value should change from 4 to 5 (subject to the result of the other backpropagation).

[0171] The values ​​are determined for these 4 new child nodes at the third level of filiation 14 of the second parent node previously selected at the second level of filiation 13, the values ​​being 6, 0, 0 and 0 (from left to right). Once the values ​​have been determined for these 4 new child nodes at the third level of filiation 14, the highest of these values ​​(namely the value 6 of the fifth node from the left of [Fig. 18] of the third level of filiation 14) is back-propagated:

[0172] • to its parent node of the second level of filiation 13, the value of which then goes from 3 to 6, • as well as to its grandparent node of the first level of filiation 12, whose value goes from 3 to 6, • as well as to the root level node 11, whose value goes from 3 to 6 (this value 6 being higher than the previously back-propagated value (value 5)).

[0173] Figures 19 to 27 schematically represent an example of the different steps of a third type of use of a tree search algorithm in the context of the automatic navigation of an elongated flexible medical instrument within a patient's blood network according to an embodiment of the invention.

[0174] This third type of use corresponds to a parallel expansion based on several values ​​(two values ​​in the example presented in Figures 19-27). The operation is similar to the parallel operation described in Figures 16 to 18, but instead of selecting the nodes only on a single value (the value defined previously), a first part of the nodes is selected according to a first value (corresponding to a first criterion) and a second part of the nodes according to a second value (corresponding to another criterion or several other criteria). Among the various other criteria that can be considered, there is the path recommended by a pre-trained end-to-end learning algorithm, or a node that would have been chosen by a cardiologist (then transformed into a numerical value by inverse reinforcement learning for example).Which results in the selection of the leaf node with the first highest value and the leaf node with the second highest value, with the expansion of their 8 child nodes, with the retrieval of rewards and the propagation of values ​​towards the root of the tree.

[0175] [Fig. 19] schematically represents an example of a first step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0176] The different nodes 10 are divided into different levels, here only the root level 11 is represented. Its double default value is 0-0 (first value of 0, and second value of 0).

[0177] [Fig.20] schematically represents an example of a second step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0178] The different nodes 10 are divided into different levels, the root level 11, the first level of filiation 12 (also called depth 1).

[0179] The passage between two nodes, of different filiation, therefore from a parent type filiation level to a child type filiation level, for example from the root level 11 to the first filiation level 12, is an action 20 (to be chosen from the 4 actions already presented which are: move forward or turn in one direction or turn in the other direction or move backward).

[0180] From the root node, the expansion of possible actions 20 is carried out, resulting in 4 new child nodes at the first level of filiation 11.

[0181] [Fig.21] schematically represents an example of a third step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0182] Once the double values ​​have been determined for these 4 new child nodes at the first level of filiation 11, of this parent node which is the root node, namely from left to right, the double values ​​4-4, 2-2, 1-5 and 3-6, the highest of these values, namely the first value 4 of the first node and the second value 6 of the fourth node (starting from the left of [Fig.21]) of the first level of filiation 12, are back-propagated:

[0183] • towards its parent node which is the root node, whose double value then changes from 0-0 to 4-6.

[0184] [Fig.22] schematically represents an example of a fourth step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0185] The different nodes 10 are divided into different levels, the root level 11, the first level of filiation 12 (also called depth 1).

[0186] The "leaf" nodes (i.e., nodes that have no children in the tree) that are selected are:

[0187] • on the one hand the leftmost node on [Fig.22], with the highest first value 4, compared to a first criterion, because in fact the other leaf nodes all have first lower values ​​which are respectively 2, 1 and 3, and on the other hand the rightmost node on [Fig.22], of second highest value 6, with respect to a second criterion, because in fact the other leaf nodes all have second lower values ​​which are respectively 4, 2 and 5.

[0188] [Fig.23] schematically represents an example of a fifth step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0189] From the two selected leaf nodes which are located at the first level of filiation 12, on the one hand the leftmost node on [Fig.23], with the first highest value 4, and on the other hand the rightmost node on [Fig.23], with the second highest value 6, we carry out the expansion of the possible actions 20. We thus arrive to 8 new child nodes at the second level of parentage 13, 4 new children per parent node previously selected at the first level of parentage 12. The expansions from the two selected leaf nodes are carried out in parallel.

[0190] [Fig.24] schematically represents an example of a sixth stage of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0191] Once the double values ​​have been determined for these 8 new child nodes at the second level of filiation 13, namely from left to right, the double values ​​4-1, 2-4, 2-3, 2-1, 0-4, 0-1, 3-1, and 0-0:

[0192] • the highest of these first values, namely the first value 4 of the first node (starting from the left of [Fig.24]) of the second level of filiation 13, is back-propagated: to its parent node of the first level of filiation 12, whose first value remains at 4, to the root level node 11, whose first value remains at 4, the highest of these second values, namely the second value 4 of the fifth node (from the left of [Fig.24]) of the second level of filiation 13, is back-propagated: to its parent node of the first level of filiation 12, whose second value then changes from 6 to 4, but not to the node at root level 11, whose second value changes from 6 to 5, because the third leftmost node, at the first level of descent 12, has a second value of 5.

[0193] [Fig.25] schematically represents an example of a seventh step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0194] The leaf node that is selected for the first value is the leaf node located at the second level of filiation 13 which is further to the left in [Fig.26], with the highest first value 4 (its double value being 4-1).

[0195] The leaf node that is selected for the second value is the leaf node located on the first level of filiation 12 which is the third leftmost node in [Fig.25]. This leaf node in fact has the second highest value 5 (its double-value being 1-5), the other leaf nodes having second lower values ​​which are respectively 1, 4, 3, 1, 2, 4, 1, 1, 0.

[0196] [Fig.26] schematically represents an example of an eighth step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0197] From the two selected leaf nodes which are located on the one hand at the second level of filiation 13 for the first selected leaf node of first highest value 4, and on the other hand at the first level of filiation for the second selected leaf node of second highest value 5, the expansion of the possible actions 20 is carried out. This results in 8 new child nodes, including four new child nodes at the third level of filiation 14 and 4 new child nodes at the second level of filiation 13 (4 child nodes per previously selected parent node). The expansions from the two selected leaf nodes are carried out in parallel.

[0198] [Fig.27] schematically represents an example of a ninth step of a third type of use of a tree search algorithm, in an automatic navigation method according to an embodiment of the invention.

[0199] Once the double values ​​have been determined for these 4 new child nodes at the third level of filiation 14, namely from left to right, the double values ​​0-0, 5-1, 0-1 and 0-0:

[0200] • the highest of the first values, namely the value 5 of the second node (starting from the left of [Fig.27]) of the third level of filiation 14, is back-propagated: • to its parent node of the second level of filiation 13, whose double value then goes from 4-1 to 5-1, • to its grandparent node of the first level of filiation 12, whose double value goes from 4-4 to 5-4, • to the root level node 11, whose first value goes from 4 to 5, • the highest of the second values, namely the value 1 of the second and third nodes (starting from the left of [Fig.27]) of the third level of filiation 14, is back-propagated only towards its parent node of the second level of filiation 13 (the second value being already 1 it remains of 1). In fact, the second value of 1 is not back-propagated to the grandparent node (and therefore not to the root node either) because it is less than 4 (the double value node 2-4 at the second level of filiation 13 stops the back-propagation of the second value 1).

[0201] Similarly, once the double values ​​have been determined for the 4 new child nodes at the second level of filiation 13, namely from left to right, the double values ​​0-0, 5-1, 0-1 and 0-0:

[0202] • the highest of the first values, namely the value 1 of the seventh node (starting from the left of [Fig.27]) of the second level of filiation 13, is back-propagated only towards its parent node of the first level of filiation 12 whose double value becomes 1-6 (the first value being already 1 it remains 1). This first value of 1 is not back-propagated towards the root node because the node with double value 5-4 of the first level of filiation 12 blocks the back-propagation, • the highest of the second values, namely the value 6 of the fifth node (starting from the right of [Fig.27]) of the second level of filiation 13, is back-propagated: • to its parent node of the first level of filiation 12, whose double value then changes from 1-5 to 1-6, • to the root level node 11, whose second value changes from 5 to 6.

[0203] Of course, the present invention is not limited to the examples and the embodiment described and shown, but it is susceptible to numerous variants accessible to those skilled in the art. In particular, the number of actions 20 is not limited to 4. For example, the number of possible actions 20 can be increased by establishing the following actions: moving forward, moving backward, turning in a first direction at an angle of X°, turning in the first direction at an angle of Y°, turning in a second direction at an angle of X°, and turning in the second direction at an angle of Y°. The first direction of rotation being opposite to the second direction of rotation, and X being less than Y.

[0204] According to a possible embodiment, an additional action may be to carry out a command according to another algorithm during N time steps, for example an inverse command of an optimization algorithm under QP (for Quadratic Programming according to the English terminology) constraint. Another additional action may also be to ask the doctor to ensure the manipulation of the instrument during N time steps.

Claims

Claims

1. Catheter robot comprising an automatic navigation system, of an elongated flexible medical instrument (28) which is a catheter or a catheter guide, with free distal end (29), implementing an automatic navigation method comprising successively: • a step of creating a model: • the elongated flexible medical instrument (28), • a blood circulation network (21) of a patient (1), along which the elongated flexible medical instrument (28) is intended to move, • of the interaction between this network (21) and this elongated flexible medical instrument (28), by modeling the contact between this elongated flexible medical instrument (28) and one or more walls of blood vessel(s) in this network (21), • from one or more images of the actual blood circulation network (21) of this patient (1), • a determination step: • the position of a starting point in the modeled network (21), • the position of an arrival point (27) in the modeled network (21), • a trajectory to be followed by the elongated flexible medical instrument (28) between the starting point and the arrival point (27), by determining a path along the modeled network (21) between the starting point and the arrival point (27), preferably with obligatory passage points in the modeled network (21), • a step of planning a sequence of commands for moving the elongated flexible medical instrument (28), obtained by a tree search procedure which, using the modeling of the elongated flexible medical instrument (28) and the modeling of said contact: • simulates different possible movements of the elongated flexible medical instrument (28) in the modeled network (21), • evaluates the results of the simulation of these different possible movements, • selects the most promising simulated movements from among these different possible movements, to follow the trajectory best determined in relation to one or more given criteria, a step of executing the planned command sequence, along the real blood circulation network (21) of said patient (1), with compensation for deviations from the trajectory determined along the modeled network (21), by closed-loop regulation.

2. Catheter robot comprising an automatic navigation system, of an elongated flexible medical instrument (28) which is a catheter or a catheter guide, with free distal end (29), implementing an automatic navigation method comprising successively: • a step of creating a model: • the elongated flexible medical instrument (28), • a blood circulation network (21) of a patient (1), along which the elongated flexible medical instrument (28) is intended to move, • of the interaction between this network (21) and this elongated flexible medical instrument (28), by modeling the contact between this elongated flexible medical instrument (28) and one or more walls of blood vessel(s) in this network (21), • from one or more images of the actual blood circulation network (21) of this patient (1), • a determination step: • the position of a starting point in the modeled network (21), • the position of an arrival point (27) in the modeled network (21), • a trajectory to be followed, by the elongated flexible medical instrument (28), between the starting point and the arrival point (27), by determining a path, along the modeled network (21), between the starting point and the arrival point (27), preferably with obligatory passage points in the modeled network (21), a step of planning a sequence of commands for moving the elongated flexible medical instrument (28), obtained by a tree search procedure which, using the modeling of the elongated flexible medical instrument (28) and the modeling of said contact: • simulates different possible movements of the elongated flexible medical instrument (28) in the modeled network (21), • evaluates the results of the simulation of these different possible movements, • selects the most promising simulated movements from among these different possible movements, to follow the trajectory best determined in relation to one or more given criteria.

3. Catheter robot according to any one of the preceding claims, characterized in that: • said automatic navigation method is capable of being implemented in real time, • said tree search procedure of the planning step simultaneously performs in parallel, for several different possible movements of the elongated flexible medical instrument (28) in the modeled network, for at least 2 or at least 3 or at least 4 different possible movements, the following planning cycle: • simulate a possible movement of the elongated flexible medical instrument (28) in the modeled network (21), • evaluate the result of the simulation of this possible movement, • select the simulated movement, if it is one of the most promising among the different possible movements, to follow the determined trajectory, as best as possible in relation to one or more given criteria.

4. Catheter robot according to any one of claims 1 or 3, characterized in that: • if at the end of said execution step, one or more deviations to be compensated are evaluated as too large, compared to one or more given criteria: • then a recalibration of the modeling is carried out, and on the recalibrated modeling, the following are then carried out: • first a new step of determining a new trajectory, • then a new step of planning a new sequence of commands for moving the elongated flexible medical instrument (28), from the new determined trajectory, • finally, if necessary, a new step of executing the new planned sequence of commands.

5. A catheter robot according to any preceding claim, characterized in that said modeling creation step performs a representation of the physical system encompassing both the elongated flexible medical instrument (28) and the blood circulation network (21) of a patient, along which the elongated flexible medical instrument (28) is intended to move, by finite element simulation.

6. Catheter robot according to claim 4 or 5, characterized in that: • the blood circulation network (21) of the patient (1) is modeled: • either by point clouds, • or by meshes, • or by central lines associated with their respective diameters, • or by implicit surfaces.

7. Catheter robot according to claim 4 or 5 or 6, characterized in that: • the modeling of the blood circulation network (21) of the patient (1) integrates the movements affecting this blood circulation network (21) of the patient (1), • and preferentially integrates the deformations of the patient's heart (1) when it beats and / or the deformations induced by the patient's breathing (1).

8. Catheter robot according to any one of the preceding claims, characterized in that: • in said determining step: • said starting point in the modeled network (21) is positioned at the exit of the guide catheter of the catheter robot, at the ostium, • and said arrival point (27) in the modeled network (21) is positioned in the patient (1), at the lesion to be treated.

9. Catheter robot according to any one of the preceding claims, characterized in that: • in said determining step: • the trajectory to be followed is determined by interpolation between said starting point and said arrival point (27), • and preferably by using a Dijkstra graph traversal algorithm applied to the central lines of the blood vessels of the blood circulation network (21).

10. Catheter robot according to any one of the preceding claims, characterized in that said tree search procedure: • evaluates the results of the simulation of these different possible movements by assigning scores to each branch of the tree structure, then by calculating a value for each node (10) of the tree structure from the scores assigned to the branches leading to this node (10), • selects the most promising simulated movements from among these different possible movements which are the simulated movements leading to the node (10) of the highest value, to guide its exploration in the tree structure.

11. Catheter robot according to claim 10, characterized in that said tree search procedure: • assigns the scores as follows: • if the elongated flexible medical instrument (28) advances along the determined trajectory, then the corresponding possible displacement of the elongated flexible medical instrument (28) in the modeled network (21) receives a positive score, • if the elongated flexible medical instrument (28) advances elsewhere than along the determined trajectory or if the elongated flexible medical instrument (28) moves backward or if the elongated flexible medical instrument (28) stagnates along the determined trajectory, then the corresponding possible displacement of the elongated flexible medical instrument (28) in the modeled network (21) receives a zero or negative score.

12. Catheter robot according to claim 10, characterized in that said tree search procedure: • assigns the scores in the following manner: • if the possible movement of the elongated flexible medical instrument (28) along the determined trajectory is considered to be similar to that which would result from the action of a doctor chosen as a reference, then this possible movement receives a positive score, • if the possible movement of the elongated flexible medical instrument (28) along the determined trajectory is considered to be different from that which would result from the action of a doctor chosen as a reference, then this possible movement receives either a zero score or a negative score.

13. Catheter robot according to claim 10, characterized in that said tree search procedure: • awards, in part, the marks in the following manner: • if the elongated flexible medical instrument (28) advances along the determined trajectory, then the corresponding possible movement of the elongated flexible medical instrument (28) in the modeled network (21) receives a positive score, • if the elongated flexible medical instrument (28) advances elsewhere than along the determined trajectory or if the elongated flexible medical instrument (28) moves backward or if the elongated flexible medical instrument (28) stagnates along the determined trajectory, then the corresponding possible movement of the elongated flexible medical instrument (28) in the modeled network (21) receives a zero or negative score, • also awards, in part, the marks in the following second way: • if the possible movement of the elongated flexible medical instrument (28) along the determined trajectory is considered similar to that which would result from the action of a doctor chosen as a reference, then this possible movement receives a positive rating, • if the possible movement of the elongated flexible medical instrument (28) along the determined trajectory is considered to be different from that which would result from the action of a doctor chosen as a reference, then this possible movement receives either a zero mark or a negative mark, • uses, with preferentially predetermined weightings, possibly different from each other, respectively the first way and the second way.

14. Catheter robot according to any one of claims 10 to 13, characterized in that the value of each node (10) is calculated by adding together the notes of each branch leading to this node (10), or the value of each node (10) is calculated using the UCB method.

15. A catheter robot according to any one of claims 10 to 14, characterized in that said tree search procedure returns to pass through a node (10) of lesser value, after a node (10) of better value has in fact proven to be a dead end.

16. Catheter robot according to claim 1 or according to any one of claims 3 to 15 attached to claim 1, characterized in that said step of executing the planned sequence of commands is carried out by direct command carrying out the movements of the actuators controlling the movements of the elongated flexible medical instrument (28).

17. Catheter robot according to claim 1 or according to any one of claims 3 to 15 attached to claim 1, characterized in that said step of executing the planned sequence of commands is carried out by inverse control calculating the movements of the actuators from the movements of the elongated flexible medical instrument (28).

18. Catheter robot according to any one of the preceding claims, characterized in that the catheter robot comprises an automatic navigation system, of several elongated flexible medical instruments (28) among which there are at least one catheter and a catheter guide associated with this catheter, these elongated flexible medical instruments (28) being with a free distal end (29).

19. Automatic navigation system, of an elongated flexible medical instrument (28) which is a catheter or a catheter guide, of a catheter robot, with free distal end (29), implementing an automatic navigation method successively comprising: • a step of creating a model: • of the elongated flexible medical instrument (28), • of a blood circulation network (21) of a patient (1), along which the elongated flexible medical instrument (28) is intended to move, • of the interaction between this network (21) and this elongated flexible medical instrument (28), by modeling the contact between this elongated flexible medical instrument (28) and one or more walls of blood vessel(s) in this network (21), • from one or more images of the actual blood circulation network (21) of this patient (1), • a determination step: • the position of a starting point in the modeled network (21), • the position of an arrival point (27) in the modeled network (21), • a trajectory to be followed by the elongated flexible medical instrument (28) between the starting point and the arrival point (27), by determining a path along the modeled network (21) between the starting point and the arrival point (27), preferably with obligatory passage points in the modeled network (21), • a step of planning a sequence of commands for moving the elongated flexible medical instrument (28), obtained by a tree search procedure which, using the modeling of the elongated flexible medical instrument (28) and the modeling of said contact: • simulates different possible movements of the elongated flexible medical instrument (28) in the modeled network (21), • evaluates the results of the simulation of these different possible movements, • selects the most promising simulated movements from among these different possible movements, to follow the trajectory best determined in relation to one or more given criteria, • a step of executing the planned command sequence, along the real blood circulation network (21) of said patient (1), with compensation for deviations from the trajectory determined along the modeled network (21), by closed-loop regulation.

20. Automatic navigation system, of an elongated flexible medical instrument (28) which is a catheter or a catheter guide, of a catheter robot, with free distal end (29), implementing an automatic navigation method comprising successively: a step of creating a model: • the elongated flexible medical instrument (28), • a blood circulation network (21) of a patient (1), along which the elongated flexible medical instrument (28) is intended to move, • of the interaction between this network (21) and this elongated flexible medical instrument (28), by modeling the contact between this elongated flexible medical instrument (28) and one or more walls of blood vessel(s) in this network (21), • from one or more images of the actual blood circulation network (21) of this patient (1), a determination step: • the position of a starting point in the modeled network (21), • the position of an arrival point (27) in the modeled network (21), • a trajectory to be followed by the elongated flexible medical instrument (28) between the starting point and the arrival point (27), by determining a path along the modeled network (21) between the starting point and the arrival point (27), preferably with obligatory passage points in the modeled network (21), a step of planning a sequence of commands for moving the elongated flexible medical instrument (28), obtained by a tree search procedure which, using the modeling of the elongated flexible medical instrument (28) and the modeling of said contact: • simulates different possible movements of the elongated flexible medical instrument (28) in the modeled network (21), • evaluates the results of the simulation of these different possible movements, • selects the most promising simulated movements from among these different possible movements, to follow the trajectory best determined in relation to one or more given criteria. 40

Images