Signal-triggered intervention device control

Abstract

A system (SYS) for supporting a medical procedure, comprising an interface (IN) for receiving at least one medical input signal describing a state of a target anatomical structure. A signal analyzer (SA) configured to analyze the medical input signal to determine a time window for deployment of a cardiovascular device (CL) to be deployed by a deployment device (CAT) relative to the target anatomical structure.

Description

Technical Field

[0001] The present invention relates to systems for supporting medical processes, medical device layout, deployment devices, methods for computer implementation of supporting medical processes, methods for training machine learning components, computer program units, and computer-readable media. Background Technology

[0002] Heart disease remains one of the leading causes of death worldwide.

[0003] Intricate protocols have been devised to maintain or restore heart function. One point of failure in the human heart is the heart valves, such as mitral regurgitation. Some of these protocols involve image-based, minimally invasive interventions. Under image guidance, medical devices such as mitral valve clips are introduced into the heart. The clip is deployed in the beating heart to reduce or even completely eliminate regurgitation. Properly deploying the clip is a highly demanding task, requiring a high level of skill from the operator. The task can be quite lengthy: the operator often needs to make various attempts to correctly place the clip. Besides mitral valve clip placement, there are other tasks where operators face similar challenges, such as in minimally invasive heart valve replacement surgery.

[0004] Health systems around the world are under increasing pressure from the “double whammy” of budget cuts and growing demand for such services from an aging population. Summary of the Invention

[0005] Therefore, methods and systems that can help healthcare professionals execute medical procedures more effectively may be needed.

[0006] The objectives of this invention are achieved through the subject matter of the independent claims, wherein other embodiments are incorporated in the dependent claims. It should be noted that the aspects described below are equally applicable to medical devices, deployment equipment, methods for computer-implemented support of medical procedures, methods for training machine learning components, computer program units, and computer-readable media.

[0007] According to a first aspect of the present invention, a system for supporting medical procedures is provided, comprising:

[0008] An interface for receiving at least one medical input signal describing the state of a target anatomical structure; and

[0009] A signal analyzer is configured to analyze the medical input signals to determine a time window for deploying a cardiovascular device to be deployed by a deployment device relative to the target anatomical structure.

[0010] Specifically, the medical input signals include: i) medical images of the target anatomical structure acquired by an imaging device, and ii) physiological signals of the target anatomical structure acquired by a measuring device, and the signal analyzer is configured to analyze both the medical images and the physiological signals.

[0011] The cardiovascular device may include a mitral valve clamp or a heart valve or any other part that can be deployed in the human or mammalian heart.

[0012] As used in this paper, a “deployment window” can indicate a real time period for deployment with a start and a (non-zero) length. However, this paper also envisions a “degenerate” time window, i.e., a moment, with a length of zero.

[0013] In an embodiment, the deployment device is capable of transitioning between at least two states, including a released state and a held state, in which the cardiovascular device can be deployed and in the held state, the cardiovascular device cannot be deployed (and is therefore held), wherein the transition is in response to a control signal.

[0014] In an automated embodiment, a computer-implemented controller unit is provided, which is configured to issue the control signal or control signal based on a determined time window to implement the operation of the deployed device.

[0015] In one embodiment, the system further includes an output interface with a transducer configured to provide a sensory indication of whether the deployment window or time has been so determined. This sensory indication can be used to instruct a user when to operate the manual interface of the deployment device to issue a manual control signal requesting the change.

[0016] The sensory indications can be output via an output terminal. The sensory indications may include audible alarm signals (“beep”), tactile feedback (e.g., by vibrating the handle of the deployment device to indicate a pending deployment window), or optical indications (e.g., a (dedicated) flashlight on the deployment device or elsewhere), or graphical widgets displayed on a display device (e.g., on a monitor), etc.

[0017] In embodiments involving user intervention, a controller may also be used. Specifically, in an embodiment, the deployment device includes a manual interface operated by the user to issue a manual control signal attempting to implement the transition, wherein if the manual control signal is received outside the deployment window, the control signal issued by the controller at least temporarily blocks or delays the manual control signal. Specifically, the manual control signal may be delayed until a new deployment window is found. In other words, in an embodiment, two signals are involved: a manual control signal caused by the user requesting the deployment device, and a control signal issued by the control device relating to whether a deployment window / moment has been found. In other words, the controller's control signal can effectively "override" the manual control signal to ensure that the cardiovascular (or other payload) device is deployed only during the (permissible) deployment window. In an embodiment, if the user feels the deployment is not optimal, the user may request a transition from the release state back to the hold state, but this is expected to occur only exceptionally.

[0018] In a preferred embodiment, two signals are used in the analysis to determine not only the state of the organ but also the geometric / mechanical relationship between the cardiovascular device and the target anatomy. This allows for accurate and robust determination of the deployment window and increases the chance of successful deployment on the first attempt, thus making the process more time-efficient.

[0019] In an embodiment, the medical image is any one or more of an MRI, X-ray, or ultrasound image, and / or the physiological signal is an ECG signal, a respiratory signal, or other suitable vital sign signal.

[0020] In an embodiment, the target anatomical structure is part of the vascular system of a human or animal, or part of the heart of a human or animal.

[0021] In embodiments, the cardiovascular or other payload device includes any one or more of the following: i) a mitral valve clamp, ii) a replacement heart valve, and / or wherein the deployment device includes a catheter and / or a robotic device.

[0022] In one embodiment, the signal analyzer includes a pre-trained machine learning component. The machine learning component may be arranged as a neural network model whose parameters have been tuned in a prior learning scheme (such as forward-backward propagation or other gradient-based techniques).

[0023] In this embodiment, the machine learning component is based on any one or more of the following: neural networks, support vector machines, and decision trees. A convolutional neural network in a deep architecture is envisioned in this embodiment. Ordinary or residual architectures can be used.

[0024] According to another aspect, an arrangement is provided, comprising: a system according to any of the foregoing embodiments, and at least one means for measuring medical input signals.

[0025] In one embodiment, the apparatus further includes a deployment device. The deployment device may include a catheter device configured for delivery to a cardiovascular device.

[0026] According to another aspect, a deployment device is provided for deploying a cardiovascular device (or other payload device) at a target anatomical structure, the deployment device being capable of transitioning between at least two states, including a released state and a held state, in which the payload device is deployable and in the held state, the payload device is held (i.e., not deployable), the deployment device including an interface configured to enable or disable (e.g., block or delay) the transition in response to a control signal issued by an embodiment of a system as disclosed herein.

[0027] The proposed system can help medical users achieve correct deployment quickly. Specifically, it can reduce the number of attempts to achieve proper deployment, which is achievable on the first attempt. In mitral valve clip delivery, it can improve the likelihood of good clinical outcomes (i.e., reducing or even completely eliminating regurgitation).

[0028] In this embodiment, the medical signals include ECG measurements. These are analyzed to detect desired cardiac phases, such as induced cardiac arrest, to prevent attempted deployment of medical devices during incorrect cardiac phases. The system only allows manual / automatic deployment during the correct phase.

[0029] Additionally, image recognition techniques (e.g., machine learning techniques that utilize neural networks trained on reference images representing good examples for deployment) are used to analyze medical image data to identify when it is appropriate to deploy the device. In some examples, the signal analyzer is configured to analyze medical image data to find the moment when the desired geometric or mechano-geometric configuration between the cardiovascular device to be deployed (such as a mitral valve clamp) and the target anatomical structure is found.

[0030] In some embodiments, a combined time window is determined based on cardiac temporal analysis and image analysis, thereby providing the best opportunity for cardiovascular device deployment.

[0031] The proposed system allows for intervention when a user attempts to manually control the delivery device to achieve deployment (e.g., mitral valve clamp closure). The system can either halt deployment based on analysis of medical signals or allow deployment to continue based on the analysis.

[0032] According to another aspect, a method for supporting the computer implementation of medical procedures is provided, comprising:

[0033] Receive at least one medical input signal describing the state of a target anatomical structure, the medical input signal including: i) a medical image of the target anatomical structure acquired by an imaging device, and ii) physiological signals of the target anatomical structure acquired by a measuring device; and

[0034] The medical input signals are analyzed to determine a time window for deploying cardiovascular equipment to be deployed by a deployment device relative to the target anatomical structure.

[0035] According to another aspect, a method for training machine learning components as used in the above embodiments of the system is provided.

[0036] According to another aspect, a computer program unit is provided, which, when run by at least one processing unit, is adapted to cause the processing unit to perform any of the methods.

[0037] According to another aspect, a computer-readable medium is provided on which the program unit is stored.

[0038] The proposed system is envisioned for mitral valve clamp delivery in the embodiments, and other minimally invasive cardiac interventions are also envisioned, such as the deployment of replacement valves for the mitral valve or for other valves, and other interventions at or around the heart, and indeed interventions related to organs outside the heart.

[0039] The "user" mentioned in this article refers to the operator who performs the intervention, especially the operator who operates the deployed equipment.

[0040] A "patient" is a person or animal (especially a mammal) who is the object of intervention.

[0041] A “machine learning (“ML”) component” is any computational unit or arrangement that implements an ML algorithm. ML algorithms are able to learn from examples (“training data”). The learning (i.e., the performance of the ML component of a task, which can be measured by performance metrics) typically improves with the training data. Some ML algorithms are based on ML models that are tuned based on the training data. In particular, the parameters of the model can be tuned in this way. Attached Figure Description

[0042] Exemplary embodiments of the invention will now be described with reference to the following figures, in which:

[0043] Figure 1 A computerized imaging apparatus for supporting medical procedures is shown;

[0044] Figure 2 shows a schematic partial cross-section through the human or mammalian heart to illustrate cardiac phases;

[0045] Figure 3 illustrates two stages of a medical device to be deployed in the heart of a human or mammal.

[0046] Figure 4 A delivery system for delivering medical devices during interventional procedures is shown;

[0047] Figure 5 shows a side view of the medical device to be deployed in two states;

[0048] Figure 6 A method for supporting the computer implementation of medical processes is illustrated;

[0049] Figure 7 An exemplary timing diagram for a deployment window for a medical device is shown;

[0050] Figure 8 This illustrates a computer-implemented method for training machine learning models; and

[0051] Figure 9 The machine learning model envisioned in the embodiment is shown. Detailed Implementation

[0052] Figure 1 An image-based support system for medical intervention is shown.

[0053] The system may include one or more imaging modalities IA1, IA2, such as an ultrasound system IA2 having an ultrasound probe TR capable of acquiring a series of images within the patient's PAT. US (ultrasound) is merely an exemplary embodiment, and other imaging modalities are also envisioned. Images generated by the imaging modal US IA2 can be plotted for viewing on a display device (e.g., a monitor) DD to support intervention.

[0054] When using medical tools (CATs) to perform actions at or near a lesion site within the body of a patient's organ of interest (OOI), the system can provide image-based support. In this embodiment, the OOI is the human or mammalian heart (H). The CAT tool may include catheters, guidewires, graspers, etc., depending on the task at hand.

[0055] Alternatively, an additional image channel may be used, for example, provided by a live fluorescence fluoroscopy image stream supplied by an X-ray imaging device IA1 having an X-ray source XR and an X-ray detector D. The patient PAT is preferably positioned on an examination table TB or other support, wherein at least a portion of the patient PAT is located between the X-ray source XR and the X-ray detector. In use, the X-ray source is excited to generate an X-ray beam XB that passes through the patient at OOI and is then recorded as intensity at the detector. Acquisition circuitry converts the intensity into an image or image stream, which is displayed on a display device DD or on a different display device. Imaging modalities other than MRI or PET / SPECT, or other imaging modalities, are also envisioned herein. The X-ray imaging device IA1 may be used in combination with, or in place of, the US (or other) imaging modal IA2.

[0056] In short, the proposed image processing system SYS operates on a multi-channel medical signal stream. The signal is analyzed by a signal analyzer SA to calculate the time window for optimal deployment of a medical device MC at the lesion site. The system SYS can operate in either of two modes, or a combination of both. In one mode, the time window may include a start time and a non-zero length to define a real time period. Alternatively, to support real-time analysis, the time window may be a "degenerate" of zero length to define a moment. The earlier mode may be referred to herein as the "predictor mode" because the system predicts the window to extend into the future with the stated time length, while the latter mode may be referred to herein as the "real-time" mode, in which the system calculates whether deployment is recommended at a given moment. Preferably, the real-time calculation of the deployment moment can be real-time relative to the frame or sampling rate of data supplied in one or two channels. In real-time mode, the degenerate time window (also referred to herein as the "deployment moment") for deployment can be recalculated for each new data in the channel, although this may not be necessary in all embodiments, as gaps may sometimes be allowed as needed.

[0057] In one channel (“physiological channel”), measurements of the physiological state of the organ of interest (OOI) are provided, while another channel (“image channel”) provides information on the mutual geometric-mechanical configuration of the medical device and the OII. Information from both channels is combined to establish a deployment window for deploying the device (MC). As will be explained in more detail below, in an embodiment, a time window is established if two conditions (one in each channel) are met. Specifically, the deployment window is established only if: i) the OII is in a predetermined physiological state according to measurements in the physiological channel and ii) a predetermined geometric relationship exists between the device and the OII according to measurements in the image channel. More relaxed embodiments are also contemplated, in which at least sometimes only one of the two (or more) channels is considered, or at least not all channels are always considered.

[0058] In one embodiment, the system includes an ECG device that provides a physiological pathway for medical signals to be analyzed by a signal analyzer SA. The ECG device includes a set of probes (not shown) that can be attached to a patient's chest. The probes are coupled to a measuring device via leads. The probes pick up electrical signals from the patient's skin caused by cardiac activity. These electrical signals are forwarded to the ECG measuring device, which processes them by modulation, digitization, or other means, and generates a stream, i.e., a time series of voltage measurements. The pattern of measured voltage values ​​in this signal can be associated with different cardiac phases, particularly with the diastolic and systolic phases, or with any other phases of interest. In this embodiment, the systolic phase is of interest, as will be explained in more detail below in Figures 3 and 4, but other phases may require attention in other contexts.

[0059] In this embodiment, another channel of the medical multichannel signal stream (image channel) is provided by images acquired by the US imager IA2. In this embodiment, the TEE modality is used. Specifically, in this embodiment, deployment or attempted deployment of the device MC is performed during constant cine image support from the TEE. The TEE-US system IA2 includes a flexible probe that is introduced into the patient's mouth down approximately halfway through the esophagus, thereby positioning the tip of the probe behind the heart (when viewing the patient's chest from the front). Once the probe is positioned in this way, relatively detailed images of the cardiac anatomy can be acquired. In this embodiment, the ultrasound system IA2 is capable of generating 3D images. At the tip of the probe, a matrix array of numerous acoustic pulse transducers is positioned, which together are used to generate a three-dimensional image volume of the heart at a given frame rate with high quality and good resolution. The TEE images are displayed live on a display device DD during the intervention.

[0060] As previously mentioned, the operating modes of the proposed system SYS can be combined. Specifically, the system SYS can operate in real-time mode on one channel and / or in predictor mode on another channel. In particular, data processing in the image channel can be performed in real-time mode, while data processing in the physiological channel can be performed in predictor mode, or vice versa.

[0061] For illustrative purposes not to be construed as limiting, the operation of the SYS system will be explained with reference to exemplary use cases (i.e., minimally invasive cardiac intervention, and more specifically, mitral valve repair in a patient’s heart H).

[0062] To explain the operation of the image processing system SYS in more detail, it will be helpful to first provide some background on mitral valve repair (especially the basic anatomy and some fundamental steps involved in mitral valve repair). It should be understood that this is strictly exemplary and purely for imaging purposes, and by no means limiting. The processor can be beneficially used for other OOIs, not necessarily for cardiac interventions, but as needed for other organs.

[0063] First, refer to Figure 2A , Figure 2B Several terms related to the relevant internal cardiac anatomy. In a preferred embodiment, OOI includes the mitral valve. This particular heart valve (in the presence of three other valves) connects the left ventricle (LV) and left atrium (LA) in the human or more generally mammalian heart (H). The mitral valve (MV) includes an outer annular base (called the (mitral) annulus) supporting a pair of mitral leaflets (L1, L2). The mitral annulus changes shape as the leaflets (L1, L2) move in sync with the cardiac phases (diastole and systole).

[0064] The valve annulus is a flexible, deformable structure that contracts and reduces its surface area during ventricular systole to facilitate complete closure of the leaflets, such as... Figure 2B As shown. During diastole, the L1 and L2 lobules move anteriorly into the LV so that the valve MV opens and allows blood to flow from the LA to the LV, as... Figure 2A As shown.

[0065] The purpose of mitral valve repair is to counteract or correct a heart defect known as mitral regurgitation (MR). Essentially, in MR, the mitral valve "leaks." That is, when the left ventricle is pressurized (i.e., during ventricular systole), its two leaflets do not completely "join" or meet at the suture zone (LC). In other words, as... Figure 3B As shown in the diagram, there is at least a localized "gap" between the two lobes, and blood refluxes, flowing back from the LV to the LA in the reflux BD. If left untreated, this reflux can cause health problems.

[0066] In one technique for mitral valve repair, a specially designed clamp is applied at the site of the gap to locally "clamp" the two leaflets L1 and L2 together, thereby reducing or even preventing regurgitation. This... Figure 3A , 3B The diagram is shown in the figure. To achieve this, a delivery system can be used that includes a catheter (CAT) and a special mitral valve clamp (MC) positioned at the tip of the catheter. The delivery system may also include a steerable guide catheter. The catheter components form a remotely controllable (via mechanical levers) system to position and deploy the clamp at the mitral valve.

[0067] Now, in a variation of minimally invasive mitral valve repair surgery, the catheter (with a clamp MC attached to its tip) is at the entry point (in... Figure 1 The catheter is introduced into the patient's PAT at a location designated as "x" (e.g., in the groin) and enters the femoral vein. The catheter is then advanced into the vena cava. After the puncture procedure, the catheter is introduced into the LA and positioned above the mitral lobule (…). Figure 3A The catheter is then carefully advanced through the leaflets into the LV and rotated such that the plane of the arm is perpendicular to the junction line LC. The clamp has gripper arms CA1 and CA2, each with corresponding gripping junction surfaces AS1 and AS2. The gripper arms CA1 and CA2 can be actuated by the user via a delivery system including the catheter CAT, which will be discussed in more detail below. Alternatively, the clamp can be actuated by a fully or partially autonomous robotic control system. The arms can be moved about a pivot PP by the user or the robot via the delivery system, whereby the arms can move away from or toward each other, forming angles, for example, between 0° and 180° or even up to 240°. The user can control the movement of the clamp arms from the other end of the catheter CAT. At the appropriate moment established by the system SYS (explained in more detail below), the catheter tip is then rapidly retracted from the LV to the direction of LA, while the clamp is then operated to grip the two leaflets that will form a gap and then clamp them together. Figure 3B The clamp is then detached from the catheter and left in place. The CAT catheter is then removed, thus ending the intervention.

[0068] Applying clamps in this manner is no easy task. It is worth noting that clamping is performed during routine cardiac surgery. Specifically, the grasping and clamping action is carried out while the leaflets are oscillating and the mitral valve annulus is moving up and down with cardiac activity or due to respiratory movements. This requires a high degree of concentration, excellent eye-hand coordination, and extensive experience in interpreting supporting images generated by the imaging modality (spatially) for both the tool operator and / or the support (interdisciplinary) team.

[0069] Now for reference Figure 4 , Figure 4A schematic block diagram of a clamp delivery system is shown, including a clamp MC attached to the tip portion of a catheter device CAT. The catheter device broadly includes a handpiece HP at the proximal end (user end) of a flexible tube T, which terminates at its other end (proximal or tip portion) at an actuator AC to which the clamp is removably attached.

[0070] The flexible tube T of the CAT conduit includes a coupling system S that couples the manual user interface (MUI) to the actuator AC and thus to the clamp MC. The coupling can be mechanical, electrical, electromechanical, or a combination thereof. The user interface MUI may include knobs, levers, buttons, etc., preferably integrated into the handheld component HP.

[0071] The tube tip is flexible and can flex in multiple different spatial directions. The actuator can also rotate along the tube's longitudinal axis. In other words, when attached to the tube tip, the clamp CL can move in any desired spatial direction in 3D space to precisely position itself relative to the engagement line, as described above. Figure 2B The movement of the tube tip, and therefore the movement of the clamp, is in response to the user interface (MUI).

[0072] The proposed image support system SYS is communicatively coupled to the catheter delivery system CAT via a control signal interface IF. The control signal interface IF is coupled to a signal suppressor SI. The signal suppressor SI can operate to interrupt or re-establish the signal path between the manual user interface MUI and the clamp actuator AC based on control signals from the controller CON of the system SYS. In other words, the signal suppressor SI can enable or disable the deployment of the clamp CL.

[0073] As envisioned in this paper, the interventional support system SYS is computer-implemented. It calculates the correct deployment window for the fixture CL based on the aforementioned multi-channel medical signal stream.

[0074] Once the system SYS establishes an instance of a deployment window, SYS commands the signal suppressor SI to enable the signal path between the manual user interface (MUI) and the AC actuator to deploy the gripper CL. In other words, within the deployment window, the user USR can operate the manual user interface (MUI) to deploy the gripper when he or she is available. However, outside the deployment window, the signal suppressor SI cuts off the signal path, so that the user can no longer deploy the gripper even when operating the manual user interface. In one embodiment, if this occurs—that is, if a user requests deployment and deployment is blocked because the request was made outside the deployment window—the request is stored in a buffer (not shown). Once the next deployment window is detected, the signal suppressor enables or re-enables the signal path, and the buffered signal is forwarded to the actuator to deploy the gripper. Thus, the user can experience this as a delayed deployment. In a fully automated embodiment where the delivery system is a robotic system, the system directly controls the deployment. That is, once the system establishes a deployment window, it sends control signals for deployment to the actuator via the signal interface to achieve gripper deployment.

[0075] Referring now in more detail to the clamp MC, it comprises two opposing arms CA1, CA2. The two arms CA1, CA2 engage at one of their ends at a pivot point PP. During actuator AC operation, the two arms can pivot about the (common) pivot point PP. Upon user request via the MUI, the conduit or actuator can be operated to allow the clamp to transition between two phases (between an open and a closed phase). In the closed phase, gripper arms CA1 and CA2 are pulled towards each other to achieve a clamping action on the leaflets, thereby forcing portions of the leaflets toward each other. In the closed state, during deployment, the angle between the two arms is essentially 0°. However, clamping can also be reversible as long as the clamp does not disengage (“fully deployed”). That is, the clamp can be reopened by the actuator, thereby moving the free ends of the arms away from each other to form an angle other than 0°, such as up to 240°, 180°, or as desired by the user. The user can first request to open and then close the clamp via the manual user interface, thus transitioning the clamp between the open and closed states. For full deployment, once the user is satisfied that the clamp has been properly deployed, the catheter tip T can be detached from the clamp via an operation (or other) user interface (MUI), thus leaving the clamp in place at the valve. The catheter can then be withdrawn from the patient, thus ending the intervention.

[0076] Corresponding to the state of the gripper, the actuator of the delivery device CAT can switch between two states (holding state and deployment state). In the deployment state, the actuator causes the arm to close, while in the holding state, it causes the gripper arm to open.

[0077] Although not in Figure 4 As shown, however, the delivery system CAT may also include additional clamp holders that fall into the clamp arms during deployment to substantially push the leaflets onto the respective mating surfaces AS, AS2 of the respective clamp arms CA1, CA2.

[0078] The signal suppressor SI can be implemented purely by electrical means to simply interrupt the signal flow from the user interface MUI to the actuator AC. Therefore, the signal suppressor can be arranged as a simple switch. Alternatively or additionally, mechanical or electromechanical embodiments are also contemplated, in which the signal suppressor is arranged as a pivotable pin that releasably engages a ratchet system connected to the MUI, thereby preventing user operation, for example, a rotatable knob that may form part of a mechanical user interface.

[0079] The operation of the computerized system SYS will now be explained in more detail, and this can be arranged in hardware or software, or a combination of both. The medical support system can be implemented on a single computing device located in the operating room, such as a computing unit communicatively coupled to the delivery system CAT and / or the monitor DD. The computing system can be a workstation or operator console associated with one of the imaging modalities IA1, IA2. Alternatively, the system SYS is remotely arranged and coupled to the catheter CAT via a suitable communication network. In other embodiments, the system is fully integrated into the catheter system, wherein the interface IN is arranged to receive one or more medical signals.

[0080] Preferably, as mentioned, the two medical signals (ECG signal stream and image stream) are processed in combination within a logical AND operation. Alternatively, only a single stream can be processed, such that the system's processing is based solely on, for example, the ECG signal or solely on the image stream. The images can be provided by any suitable imaging modality, such as ultrasound (e.g., TEE), X-ray (CT or radiography), MRI, or any other imaging modality required by the medical task at hand. Physiological signals are preferably provided by an ECG device, but in alternative embodiments, for other interventions, instruments other than ECG may be more readily available to measure different quantities of interest.

[0081] Continue to refer to Figure 1 The computerized system SYS includes a signal interface IN through which medical signals from one or more channels are received. The signals are then analyzed by a signal analyzer SA to calculate the deployment window. The signal analyzer SA can be coupled to a controller CON. Once the deployment window is located, the signal analyzer provides an indication to the controller CON via an output interface OUT. The controller then translates this indication (numerical, flag, code, or other) into control signals to control the catheter delivery system CAT, as described above in Figure 3-5.

[0082] Specifically, control signals can be disabled or enabled via the signal suppressor SI to enable the manual user interface (MUI), as described above. Alternatively or additionally, in a fully autonomous robotic environment, once the signal analyzer SA has located the deployment window, control signals are provided to the robotic device via the appropriate interface to achieve deployment automatically and directly.

[0083] Alternatively, or in lieu of facilitating user operation through the User Interface (MIU) or fully automated deployment control in a robotic system, output signaling provided via the Output Interface (OUT) may include a transducer (TR). The output signaling may be processed by the transducer (TR) to provide the user with sensory indications related to the deployment window, notifying the user whether a deployment window has been established. Sensory indications may be in the form of audible or visual signals, such as by activating / deactivating an alarm light (AL), via a speaker (SP). Additionally or alternatively, the transducer (TR) may include, or be coupled to, a visualization module capable of generating appropriate graphical indications that are displayed in a graphical display on a display device (DD) along with or instead of the current image supplied by a TEE IA2 or X-ray system IA1 or any modality used. The graphical indications may be displayed on a different display device than the one used for the image. The graphical indications may include one or more bars or lines to indicate a deployment window in predictor mode, or may include symbolic representations indicating recommended / permitted or not recommended / permitted deployment in real-time mode.

[0084] In this embodiment, no sensory indication is provided if the deployment window is not found; rather, the indication is provided only once the deployment window has been found. Alternatively or additionally, the fact that no deployment device has been found is indicated by a specific sound, color, or frequency of an alarm flashing light and / or by an indication to maintain deployment on a display device. Once the deployment window is found, the indication becomes a "progress signal" by displaying a suitable symbol on the display device DD, by changing the sound emitted through a speaker, or by changing the frequency or color of the indicator light AL, etc.

[0085] The operation of the signal analyzer (SA) in predictor mode will now be explained in more detail. The analysis of signals representing physiological states (e.g., cardiac phases) can be based on finding signal patterns. For example, an electrocardiogram (ECG) is a voltage versus time curve. Analyzing this curve over multiple cycles allows the establishment of the actual heart rate, as well as the patterns of sub-segments of the curve representing the cardiac phase of interest, and the time of each heartbeat's onset. For example, the signal analyzer (SA) can analyze the received voltage versus time signal for a specific signal pattern (such as a QRS complex). More specifically, methods for (near real-time) detection of QRS complexes can be used to derive the onset and duration of the desired cardiac phase. A deployment window can then be extrapolated based on the actual heart rate and the said duration of future cardiac cycles. More refined embodiments of the signal analyzer (SA) are also envisioned, which further analyze whether the heart rate is increasing or decreasing, and then adapt the length of the cardiac cycle and associated sub-structures accordingly. In this way, a deployment window suitable for device deployment can be provided as the actual time interval. Alternatively, in real-time mode, the methods used for real-time heartbeat classification can be used to identify arrhythmias, or more generally, to identify undesirable situations such as device deployment (or device deployment during the next heartbeat).

[0086] In addition to or replacing physiological signals, the image stream provided by US (e.g., TEE) or X-rays or by any other suitable imaging equipment is analyzed to obtain a pattern indicating the appropriate deployment window. This image-based analysis is preferably envisioned for real-time mode and references... Figure 5A and 5B To explain in more detail. More specifically, Figure 5A This is a schematic representation of a cross-sectional view of the geometry of an instance of a correctly deployed window, as discovered by the applicant. The geometry shown in Figure 5 can exist in the currently received frame of the monitored image stream, constituting an instance of the deployed window, since leaflets L1, L2 lie flat on the mating surfaces AS1, AS2 of the two arms CA1, CA2 (assuming these are flat). Typically, during contact, the shape of the leaflets substantially follows the contours of the mating surfaces AS1, AS2 of the two arms CA1, CA2. In particular, both leaflet contours L1 and L2 follow the contours of the surfaces AS1, AS2 of the respective arms. Therefore, Figure 5A The geometric configuration shown is referred to in this paper as the tight fit condition. Geometrically and mechanically, this condition represents contact with a parallel tangential plane at least on the first order. In contrast, in Figure 5B The diagram indicates moments when deployment is unsuitable, where the close fit condition is not met because leaflets L1 and L2 have ripples or folds at their locations (indicated by "x" in the diagram), i.e., they do not follow the contours of the surfaces of arms CA1 and CA2. Such ripples or other contour mismatches can be caused, for example, by momentarily occurring, unfavorably oriented blood flow or eddies.

[0087] Deploying the clamp CL in scenario 5B may result in poor outcomes that are insufficient to combat or contain backflow. However, in situations such as Figure 5A Under the schematically illustrated conditions of close engagement, deployment will, in most cases, result in reliable occlusion of the local regurgitation. Advantageously, in the multi-channel embodiment of the proposed support system SYS, there is no sole dependence on the physiological channel, because even if the heart may be in the systolic phase and therefore the mitral valve is open, intervention is unlikely to succeed without close engagement with the surface of the grasper, as residual regurgitation may still persist. Furthermore, deployment is extremely unlikely to succeed if an attempt is made to operate outside the systolic phase. However, as proposed herein in the embodiments, by combining information from both channels (the imaging channel and the physiological channel), intervention can be shortened by enabling correct and rapid deployment.

[0088] In principle, and as envisioned in the embodiments, the close fit requirement can be established based on conventional image analysis techniques, including segmenting the image frame for lobes and arms. Tangent planes or (depending on the image dimensions) can be established for the two segmented structures to assess the closeness between curves or surfaces.

[0089] However, instead of using such segmentation-based techniques, alternative embodiments using machine learning techniques (such as those leveraging neural networks or other deep learning methods) are envisioned. These machine learning methods assume the existence of a latent mapping between images and labels, indicating suitability or unsuitability for deployment. This implicit latent mapping is itself unknown but can be learned from training data. In particular, training data can be acquired in supervised learning schemes. Similar analyses based on machine learning models can also be used to analyze ECG signals to correctly identify the desired phase of interest, systolic or diastolic phases, or other phases. The following will discuss... Figure 8 , 9 These ML-based implementations will be explained in more detail below.

[0090] In real-time mode, this analysis by the analyzer SA is preferably performed for each image frame or for each nth frame (n<1) or randomly, and at each such moment it is determined whether to deploy.

[0091] Alternatively or additionally, in predictor mode, a deployment window (real time period) can be calculated given a specific image and / or ECG reading. In this way, the deployment window can be formed by indicating a real time period rather than a specific moment. Prediction can be performed by initially analyzing multiple frames and / or ECG readings within one or more cardiac cycles using the analyzer SA and then predicting the time window. Due to this delay, the system's response may be slower.

[0092] More specifically, the predictor pattern is based on the insight that the received image sequence of the mitral valve provides the timing of mitral valve opening (i.e., the start of left atrial contraction (LA)) and mitral valve closure (i.e., the end of left atrial contraction (LA)). This information about the opening and closing times of the valve MV forms patient-specific information about the length of the cardiac cycle (repeated valve openings) provided by tandem physiological signals (e.g., ECG signals) and the individual signal structure of the patient. In other words, the individual heart rate of a given patient can be estimated. If these events (i.e., valve opening or closure) have already been identified in previous frames, the heart rate can be predicted by extrapolation as described above, and thus a window can be deployed. In embodiments, a state-space model can be used, where states represent different cardiac phases. Alternatively or additionally, extrapolation can be performed by a machine learning model pre-trained on training data including such historical image sequences and / or ECG data.

[0093] Now for reference Figure 6 , Figure 6 This illustrates supporting medical procedures (such as those shown above). Figure 1 The flowchart below describes a computerized method for the medical process described in section -5. However, it should be understood that the methodological steps described below are not necessarily limited to this. Figure 1 , 4 The framework, and the following steps can themselves be understood as instruction.

[0094] At step S610, one or more medical signals are received. The medical signals may include images of an anatomical structure of interest and signals representing the physiological state of that structure. In embodiments, the anatomical structure includes a mammalian (human or animal) heart, particularly its valves, such as the mitral valve or other valves. The medical procedure may be one of minimally invasive valve repair procedures. For example, embodiments include images of a human heart, particularly showing portions of the human heart (especially the mitral valve or other valves), and another signal is an ECG signal representing a cardiac phase. The phase of interest is the systolic phase, but other cardiac phases, such as the diastolic phase or a phase between both, may be required in other interventions. In other embodiments, the phase refers to the phase of the respiratory cycle in other organs, such as in lung imaging, or others.

[0095] At step S620, one or more medical signals are analyzed to establish a time window for deploying a medical device (particularly a cardiovascular device such as a mitral valve clamp). The medical device is deployed by a deployment device such as a catheter, etc. Figure 4 As described in [the text].

[0096] In mitral valve clip deployment, a clip CL is to be deployed. The clip is attached to the leaflet of the mitral valve during intervention. In a preferred embodiment, information from two signal streams (physiological stream and image stream) is combined in the analysis. However, in other embodiments, only one stream is considered, which is less preferred. In the embodiment, a time window for the embodiment is established if the heart is in its desired cardiac phase according to the physiological ECG signal. This represents the physiological condition. Additionally, images in the image stream are analyzed to find instances of the desired geometric or mechano-geometric configuration between the device to be deployed (e.g., the mitral valve clip) and the target anatomical structure (in this case, the leaflet of the mitral valve). The latter condition can be referred to as the close fit condition as described above in Figure 5. Preferably, the conclusion that a deployment window for the device exists is only reached if both conditions are met.

[0097] If, at step S620, it is determined that a deployment window cannot be established based on the current medical signals, the signal flow returns to step S610 to receive subsequent images, and the method continues to monitor the incoming signal flow. In this case, if a user request to still deploy is received at step S630, the request is rejected, and deployment is blocked.

[0098] However, if a deployment window is determined to exist at step S620, a deployment request recorded earlier at S630 is now permitted, and deployment can be facilitated or implemented at step S640a. Alternatively, deployment is also initiated if a deployment request is received for the first time within that window.

[0099] Enabling or facilitating deployment may include providing users with visual or other sensory indications that a suitable deployment window has been found.

[0100] Alternatively, in a fully automated autonomous environment, control signals are sent to the robot delivery system at step S640b to enable deployment automatically without user interaction.

[0101] In either case, the deployment may include an actuator that controls the delivery device to switch the clamp from an open position to a closed position to achieve the action described above of clamping the two leaflets of the mitral valve together.

[0102] Now for reference Figure 7 This shows three timing diagrams a), b), and c) representing different use cases. The line segment designated "M" indicates a user request. Figure 7The line segments designated "ECG" and "TEE" indicate the deployment window for each channel over time t. Specifically, the "ECG" segment indicates the ECG signal being deployed, while the "TEE" segment indicates the arrival of the deployed image. However, in a preferred embodiment, an overall deployment window is established only when there is temporal overlap between the windows of each channel, i.e., when both of the above conditions are met. The dashed line at "M" indicates a user request for deployment that is blocked.

[0103] The TEE segment is longer than the corresponding ECG signal to indicate different refresh rates, where the ECG signal reacts much faster than the image refresh rate. More specifically, in Figure a), user request M is allowed because the windows of each channel in the ECG and TEE overlap. In Figure b), user request M is denied. While analysis of the ECG channel will allow the request, analysis of the image channel TEE will not. In other words, no deployment window is established. In Figure c), user request M is initially denied (as shown by the dashed section) because there is no corresponding window in the ECG channel that can overlap with the image channel window. Later, such an overlap exists because a new window is established in the ECG channel, where the window in the image channel is still pending, and therefore the deployment request is now allowed. In the multi-channel embodiment, there are therefore two independent monitoring loops, one for each channel, to establish the corresponding conditions. Deployment is allowed if both are met. Line or dot segments with respective indications of the deployment window or deployment time are provided. Figure 7 The timing diagrams in any of a)-c) are schematic representations of embodiments of the above-described graphical display concept.

[0104] The deployment of the mitral valve clip described above is based on only one embodiment, and other applications are also envisioned herein, such as computer-controlled catheter steering, for example, in cases of aortic stenosis and transcatheter aortic valve treatment, where the catheter needs to pass through the narrow orifice of a calcified aortic valve. During ventricular systole, blood is pumped through the aortic valve, and the jet of blood pushes the catheter tip away from the orifice, making passage through the valve very difficult during this cardiac phase. In this embodiment, an ECG signal can be used to prevent forward propulsion of the catheter during ventricular systole and to select the appropriate cardiac phase for passage through the aortic valve. Catheter manipulation through the orifice can be supported solely by the physiological channel, or in place of the image channel, or in addition to the image channel. In this embodiment, for example, images (e.g., angiographic images) are analyzed for the correct orifice diameter.

[0105] Now for reference Figure 8 , Figure 8 A flowchart of a method for training a machine learning model is shown. Once trained, the model can then be used to implement the analysis steps S620 or the signal analyzer SA described above.

[0106] More specifically, at step S810, the training data is preferably provided as pairs of training input data and their corresponding targets. Training data can be requested from historical data stored in a medical repository.

[0107] At step S820, a machine learning algorithm is applied to the data to learn the aforementioned latent relationships. This step may include iteratively adjusting pre-assigned parameters of the machine learning model. The machine learning model may be implemented as a neural network, particularly in a deep learning architecture.

[0108] At step S830, once the model has been sufficiently trained, that is, after a predetermined number of iterations or after a certain stopping condition is met, the machine-trained learning model can be released for use.

[0109] Training data can be retrieved from previously acquired historical images and / or electrocardiograms from previous interventions (such as those available in hospital PACS systems or other medical data repositories). In particular, medical image datasets representing previous cardiac interventions involving the deployment of cardiovascular devices as described herein can be used in model training.

[0110] Human experts (such as radiologists or others, depending on the modality) can review such training images to label or annotate them, indicating whether the instances shown in the images would be desirable or undesirable for deployment. Thus, experts can assign labels, and the training images, thus labeled, form training data. These labels form the targets used in a machine learning model. This model can include a neural network (deep learning) with two or more hidden layers.

[0111] In addition to or in lieu of images, experts can tag historical electrocardiograms. Experts can also use estimates of the length of the deployment window as a corresponding target.

[0112] The model's parameters can be tuned based on the training data during the learning phase. In this training phase, the initially filled parameters are adjusted using an optimization scheme (such as forward / backward propagation or other gradient-based techniques). For example, adjustment occurs over one or more iterations, allowing all training data to accumulate to a properly tuned neural network. Once properly trained, the network can then be used in a signal analyzer. Images previously unseen during the use of the system for a given intervention are then fed into the network and forward-propagated to generate a prediction of whether to deploy given the image.

[0113] The machine learning model can then analyze the new images and / or electrocardiograms to predict whether deployment is desirable. To ensure near real-time delivery of the deployment window description, it is preferable to train or deploy the machine learning model on a dedicated high-performance computing unit that includes data processors configured for parallel computing. Examples include multi-core processors, but may also include dedicated special-purpose chips such as GPUs or others.

[0114] Now for reference Figure 9 , Figure 9 An example of a machine learning model is shown, which involves a deep neural network with multiple hidden layers. More specifically, Figure 9 This is a schematic diagram of a convolutional neural network (CNN) as envisioned in some, but not necessarily all, embodiments. In particular, this NN embodiment does not exclude alternative techniques such as support vector machines, linear or nonlinear regression or classification algorithms, decision trees, etc., all of which are equally envisioned herein.

[0115] The fully configured NN obtained after training (described more fully below) can be considered as an approximate representation of a latent mapping between two spaces: i) the space of images and / or ECG readings and ii) a binary decision space {0, 1}, where one element represents “deployment” and the other “no deployment”. Thus, the learning task is essentially a classification task, where a given image is classified as an instance of whether it represents a deployment. Alternatively, the latent mapping can be understood as a mapping between i) the space of images and / or ECG readings and ii) the time intervals of predictor patterns. Images represent historical samples of the corresponding interventions that are desired to be supported, such as (mitral) valve repairs via the aforementioned clamp deployment or catheter turning application, or any other intervention of interest. The trained machine learning model attempts to approximate this mapping. The approximation can be achieved during the learning or training process, where parameters (which themselves form a high-dimensional space) are adjusted based on the training data in the optimization scheme.

[0116] More specifically, the machine learning component can be implemented as a neural network (“NN”), particularly a convolutional neural network (“CNN”). Continuing to refer to Figure 11, this illustrates in more detail the CNN architecture as envisioned in the embodiments herein.

[0117] CNNs can operate in two modes: "training mode / phase" and "use mode / phase." In training mode, an initial model of the CNN is trained based on a set of training data to produce a trained CNN model. In use mode, new, untrained data is fed into the pre-trained CNN model to operate during normal use. Training mode can be a one-time operation, or this can continue in repeated training phases to improve performance. Everything described so far regarding the two modes applies to any kind of machine learning algorithm and is not limited to CNNs, or, for this problem, not limited to neural networks.

[0118] A CNN comprises a set of interconnected nodes organized in layers. The CNN includes an output layer (OL) and an input layer (IL). The input layer (IL) is sized to accept training input images representing the mutual geometry of medical devices at the organ of interest (such as a mitral valve clamp and mitral valve). In other words, the input layer (IL) is an m x n matrix, where n and m are the number of pixels in m rows and n columns. The output layer (OL) for the proposed binary classification task can be chosen as a 2D vector (x, y), where an entry (e.g., x) represents "deployment" and an entry (e.g., y) represents "not deployed".

[0119] The CNN preferably has a deep learning architecture, i.e., there are at least one, preferably two or more, hidden layers between the OL and IL layers. The hidden layers may include one or more convolutional layers CL1, CL2 (“CL”) and / or one or more pooling layers PL1, PL2 (“PL”) and / or one or more fully connected layers FL1, FL2 (“FL”). CL is not fully connected, and / or the connections from CL to the next layer may vary, but are typically fixed in FL. The CNN may be stored in memory MEM.

[0120] Nodes are associated with numerical values ​​(“weights”) that represent how a node responds to inputs from earlier nodes in the previous layer.

[0121] The set of all weights defines the configuration of the CNN. Weights are the parameters of the model. During the learning phase, the initial configuration is adjusted based on the training data using a learning algorithm such as forward-backward (“FB”)-propagation or other optimization schemes such as gradient descent. Gradients are obtained with respect to the parameters of the objective function.

[0122] The training mode is preferably supervised, i.e., based on labeled training data. Labeled training data consists of pairs of training data entries. For each pair, one entry (image and / or electrocardiogram) is the training input data, and the other entry (labeled with "0" or "1") is a target that is known prior to be correctly associated with its training input data entry. This association defines the label and is preferably provided by a human expert.

[0123] In training mode, preferably, training input data from multiple such pairs are applied to the input layer to propagate through the CNN until an output is reached at the operational level (OL). Initially, the output is typically different from the target. During optimization, the initial configuration is readjusted to achieve a good match between the input training data of all pairs and their corresponding targets. The matching is measured by a similarity metric, which can be formulated according to an objective function or a cost function. The aim is to tune the parameters to produce a low-cost, i.e., good match.

[0124] More specifically, in the NN model, input training data entries are applied to the input layer (IL) and passed through a cascade of convolutional layers CL1, CL2 and possibly one or more pooling layers PL1, PL2, and finally passed to one or more fully connected layers. The convolutional modules are responsible for feature-based learning (e.g., identifying patient characteristics and features in the background data), while the fully connected layers are responsible for more abstract learning, such as the impact of features on treatment. The output layer OL contains output data representing estimates for the corresponding objective.

[0125] according to Figure 9 The exact grouping and order of the layers is only one exemplary embodiment, and other groupings and orders of layers are contemplated in different embodiments. Furthermore, the number of layers for each type (i.e., any one of CL, FL, PL) can be related to... Figure 8 The arrangements shown are different. The depth of a CNN can also be... Figure 9 The depths shown are different. All of the above also applies to other neural networks envisioned in this paper, such as fully connected classical perceptrons / typed neural networks and recurrent neural networks. Unlike the above, unsupervised learning or reinforcement learning schemes can also be envisioned in different embodiments. Figure 9 The variation of the “normal” neural network architecture in NNs also envisions a residual network architecture (“ResNet”), in which the local output from one layer is not fed to the next layer, but one or more such layers are skipped and combined with the output of the downstream layer. Skipping one or more such intermediate layers in this way allows against the vanishing gradient effect. See, for example, “Deep Residual Learning for Image Recognition” by Kaiming He et al. (available online at arXiv:1512.03385v1[cs.CV]10Dec 2015).

[0126] As envisioned in this paper, the labeled training data may need to be reformatted into a structured form. As mentioned, the labeled training data can be arranged as vectors, matrices, or tensors (arrays with dimensions greater than 2). This reformatting can be accomplished by a data preprocessor module (not shown), such as a script or filter, which runs through patient records from a hospital information system or other image repository to pull out images of interventions of interest (e.g., mitral valve repair).

[0127] The training dataset is applied to the initially configured CNN model, and then the training data is processed according to a learning algorithm such as the FB-propagation algorithm described above. At the end of the training phase, the pre-trained CNN can then be used to help the user predict the correct deployment window upon intervention.

[0128] The aforementioned ML form can also be used to identify desired cardiac phases and / or avoided cardiac phases in a given ECG sequence. In this case, instead of images or in addition to images, the training data pairs include historical ECG curve portions labeled by medical experts using codes representing the corresponding cardiac phases. ECG curve portions can be encoded as vectors (v1,…v…). i ,..,v T ), where each entry represents the voltage value at time i, and the associated label or target is the encoding of the correct cardiac phase.

[0129] Some or all of the above steps can be implemented in hardware, software, or a combination thereof. Hardware implementations may include appropriately programmed FPGAs (Field-Programmable Gate Arrays) or hard-wired IC chips. For good responsiveness and high throughput, multi-core processors such as GPUs or TPUs can be used to implement the training and use of machine learning models as described above, particularly for neural networks (NNs).

[0130] The components of the image processing system SYS can be implemented as software modules or routines within a single software suite and run on a general-purpose computing unit (PU), such as a workstation associated with imagers IA, IA2, or a server computer associated with a group of imagers. Alternatively, the components of the image processing system IPS can be arranged in a distributed architecture and connected to a suitable communication network. Alternatively or additionally, some or all of the components of the system SYS can be arranged in hardware such as a properly programmed FPGA (Field Programmable Gate Array) or as hard-wired IC chips.

[0131] One or more features disclosed herein may be configured or implemented as / have circuitry encoded in a computer-readable medium, and / or combinations thereof. Circuitry may include discrete and / or integrated circuits, application-specific integrated circuits (ASICs), system-on-a-chip (SoCs), and combinations thereof, machines, computer systems, processors and memories, and computer programs.

[0132] In another exemplary embodiment of the invention, a computer program or computer program unit is provided, characterized in that it is adapted to run on a suitable system the method steps of one of the methods described in the foregoing embodiments.

[0133] Therefore, a computer program unit can be stored on a computer unit, which may also be part of an embodiment of the present invention. This computing unit can be adapted to perform or cause the execution of the steps of the methods described above. Furthermore, it can be adapted to operate components of the apparatus described above. The computing unit is adaptable to automatic operation and / or to execute user commands. The computer program can be loaded into the working memory of a data processor. Therefore, the data processor can be equipped to perform the methods of the present invention.

[0134] This exemplary embodiment of the invention covers both computer programs that initially use the invention and computer programs that, through updates, transform existing programs into programs that use the invention.

[0135] Furthermore, the computer program unit may be able to provide all the necessary steps to perform the process of the exemplary embodiments of the methods described above.

[0136] According to yet another exemplary embodiment of the invention, a computer-readable medium, such as a CD-ROM, is provided, wherein the computer-readable medium has computer program units stored thereon, the computer program units being described in the preceding sections.

[0137] Computer programs may be stored and / or distributed on suitable media (particularly, but not necessarily, non-transient media), such as optical storage media or solid-state media provided together with or as part of other hardware, but the computer programs may also be distributed in other forms, such as via the Internet or other wired or wireless communication systems.

[0138] However, computer programs can also be provided on networks such as the World Wide Web and can be downloaded from such networks to the working memory of a data processor. According to yet another exemplary embodiment of the invention, a medium is provided for making computer program units available for download, said computer program units being arranged to perform a method according to one of the embodiments described above.

[0139] It should be noted that embodiments of the invention have been described with reference to different subjects. In particular, some embodiments are described with reference to claims of the method type, while other embodiments are described with reference to claims of the device type. However, unless otherwise stated, those skilled in the art will understand from the above and below description that any combination of features relating to different subjects, except for any combination of features belonging to one type of subject, is also considered to be disclosed in this application. However, it is possible to combine all features to provide more synergistic effects than a simple sum of features.

[0140] Although the invention has been illustrated and described in detail in the accompanying drawings and the foregoing description, these illustrations and descriptions should be considered illustrative or exemplary, not restrictive. The invention is not limited to the disclosed embodiments. Those skilled in the art, through studying the drawings, the disclosure, and the dependent claims, will be able to understand and implement other variations of the disclosed embodiments in practicing the claimed invention.

[0141] In the claims, the word "comprising" does not exclude other elements or steps, and the words "a" or "an" do not exclude multiple. A single processor or other unit may perform the functions of several items recited in the claims. Although specific measures are recited in dissimilar dependent claims, this does not indicate that combinations of these measures cannot be advantageously used. No reference numerals in the claims should be construed as limiting the scope.

Claims

1. A system (SYS) for supporting medical processes, comprising: An interface (IN) configured to receive at least one medical input signal describing the state of a target anatomical structure, wherein the medical input signal includes: i) Medical images of the target anatomical structure acquired by an imaging device (IA), and ii) Physiological signals of the target anatomical structure acquired by a measurement device (PR); and A signal analyzer (SA) is configured to analyze the medical images and the physiological signals to determine a time window for deploying a cardiovascular device (CL) to be deployed by a deployment device (CAT) relative to the target anatomical structure, wherein the images are analyzed to find the moment when the required geometric or mechano-geometric configuration between the device to be deployed and the target anatomical structure is found.

2. The system of claim 1, further comprising a controller (CON) configured to issue control signals based on a determined time window to enable operation of the deployed device.

3. The system according to claim 2, wherein, The control signal is operable to cause the deployment device (CAT) to transition between at least two states, including a released state in which the cardiovascular device (CL) can be deployed and a held state in which the cardiovascular device (CL) cannot be deployed.

4. The system according to any one of claims 1-3, further comprising an output interface (OUT) for providing a sensory indication of whether the deployment time has been so determined.

5. The system according to claim 1, wherein, The medical image is any one or more of an MRI image, an X-ray image, or an ultrasound image, and / or the physiological signal is an ECG signal.

6. The system according to any one of the preceding claims, wherein, The cardiovascular device (CL) includes any one or more of the following: i) a mitral valve clamp, ii) a heart valve replacement, and / or wherein the deployment device (CAT) includes a catheter and / or a robotic device.

7. The system according to any one of the preceding claims, wherein, The signal analyzer (SA) includes a pre-trained machine learning component (MLC).

8. The system according to any one of the preceding claims, wherein, The signal analyzer (SA) is configured to analyze the medical images to find instances of the desired geometric or mechano-geometry configuration between the cardiovascular device to be deployed and the target anatomical structure.

9. An apparatus (AR) comprising: The system (SYS) according to any one of claims 1-8, and at least one device (IA, PR) for measuring the medical input signal; And deployment devices (CATs) for deploying intravascular devices.

10. A method for supporting the computer-implemented medical process, comprising: Receive (S610) at least one medical input signal describing the state of the target anatomical structure, the medical input signal including: i) a medical image of the target anatomical structure acquired by an imaging device (IA), and ii) physiological signals of the target anatomical structure acquired by a measuring device (PR); and The medical images and physiological signals are analyzed (S620) to determine a time window for deploying a cardiovascular device (CL) to be deployed by a deployment device (CAT) relative to the target anatomical structure, wherein the images are analyzed to find the moment when the required geometric or mechano-geometric configuration between the device to be deployed and the target anatomical structure is found.

11. A computer program unit, which, when run by at least one processing unit (PU), is adapted to cause the processing unit (PU) to perform the method according to claim 10.

12. A computer-readable medium having stored thereon a program unit according to claim 11.

Images