Systems and methods for controlling electrosurgical energy delivery

Abstract

A computer-implemented method for controlling delivery of electrosurgical energy to a blood vessel to seal the blood vessel includes collecting data from an electrosurgical system including an instrument and an energy source while the instrument delivers electrosurgical energy from the energy source to the blood vessel, predicting a rupture pressure probability of the blood vessel based on the data using a machine learning algorithm, and determining whether the blood vessel is sufficiently sealed based on the prediction. The data includes electrical parameters related to the delivery of the electrosurgical energy. In the event that the blood vessel is determined not to be sufficiently sealed, determining an output by a second machine learning algorithm, communicating the determined output to a computing device associated with the energy source for use in formulating an energy delivery algorithm, and delivering additional electrosurgical energy from the energy source to the blood vessel according to the energy delivery algorithm using the instrument to seal the blood vessel.

Description

background Technical Field

[0002] This disclosure relates to the prediction of vascular parameters, and more specifically to systems and methods for controlling the delivery of electrosurgical energy from surgical devices based on estimated vascular parameters by incorporating predictions of vascular parameters based on machine learning.

[0003] Related technical background

[0004] Surgical instruments are used to perform various functions on tissue structures. For example, surgical forceps are clamp-like devices that rely on the mechanical action between their jaws to grasp, clamp, and restrain tissue. Energy-based surgical forceps utilize both mechanical clamping action and energy to treat, for example, coagulation, cauterization, and / or sealing of blood vessels.

[0005] Surgical instruments such as energy-based forceps can often effectively manipulate tissue without requiring tissue temperature feedback. For example, energy-based forceps achieve vascular sealing by subjecting blood vessels to a controlled distribution of energy under controlled pressure. Summary of the Invention

[0006] As used herein, the term "far side" refers to the portion described as being farther from the user, while the term "proximal side" refers to the portion described as being closer to the user. Furthermore, to the extent consistent, any aspect described herein may be used in combination with any or all of the other aspects described herein.

[0007] According to an aspect of this disclosure, a computer-implemented method for controlling the delivery of electrosurgical energy to a blood vessel for sealing the vessel is proposed. This computer-implemented method includes collecting data from an electrosurgical system comprising the instrument and the energy source while an instrument delivers electrosurgical energy from an energy source to the blood vessel to seal it. The method further includes transmitting the data to a machine learning algorithm; predicting a rupture pressure probability of the blood vessel using the machine learning algorithm; and determining, based on the prediction, whether the blood vessel is adequately sealed. The data includes electrical parameters related to the delivery of the electrosurgical energy. If it is determined that the blood vessel is not adequately sealed, the method further includes determining an output based on the data using a second machine learning algorithm, transmitting the determined output to a computing device associated with the energy source for formulating an energy delivery algorithm, and using the instrument to deliver additional electrosurgical energy from the energy source to the blood vessel according to the energy delivery algorithm to seal the blood vessel.

[0008] In one aspect of this disclosure, the machine learning algorithm may include a neural network.

[0009] In another aspect of this disclosure, the neural network may include a feedforward network, a convolutional network, and / or a recurrent network.

[0010] In one aspect of this disclosure, the method may further include training the neural network based on one or more patterns in measurement sensor data or identification data.

[0011] In another aspect of this disclosure, the method may further include training the neural network based on training data, including impedance, vascular temperature, vascular mass, vascular surface area, cumulative energy, and / or rupture pressure probability.

[0012] In another aspect of this disclosure, training may include supervised learning, unsupervised learning, and / or reinforcement learning.

[0013] In one aspect of this disclosure, reinforcement learning may include rewards based on rupture pressure probability values ​​and / or impedance penalty values.

[0014] In another aspect of this disclosure, the rupture pressure probability value can be a scaler. The rupture pressure probability value can be determined by: increasing the rupture pressure probability value by a first amount when the blood vessel temperature is within a first temperature range for a first predetermined time period of protein denaturation; increasing the rupture pressure probability value by a second amount when the blood vessel temperature is within a second temperature range for a second predetermined time period to remove a predetermined percentage of water; and increasing the rupture pressure probability value by a third amount when the blood vessel temperature is within a third temperature range for a third predetermined time period to allow for thermosetting gelatin coagulation and jaw cooling.

[0015] In another aspect of this disclosure, the impedance penalty value can be a second scaler. The impedance penalty value can be determined by setting it to -1 if the blood vessel impedance is greater than a first impedance threshold. If the blood vessel impedance is less than the first impedance threshold, the impedance penalty value can be determined by setting it to 0.

[0016] In another aspect of this disclosure, the method may further include stopping the delivery of electrosurgical energy and / or additional electrosurgical energy once it is determined that the blood vessel is adequately sealed.

[0017] According to an aspect of this disclosure, a system for controlling electrosurgical energy is proposed. The system includes an electrosurgical system comprising instruments and an energy source, one or more processors, and a memory coupled to the one or more processors. The memory has instructions stored thereon that, when executed by the one or more processors, cause the system to: collect data from the electrosurgical system, including electrical parameters related to the delivery of the electrosurgical energy, when the instruments deliver electrosurgical energy from the energy source to a blood vessel to seal the vessel; transmit the data to a machine learning algorithm; predict a rupture pressure probability of the blood vessel using the machine learning algorithm; and determine, based on the prediction, whether the blood vessel is adequately sealed. If it is determined that the blood vessel is not adequately sealed, the instructions, when executed, also cause the system to determine an output based on the data using a second machine learning algorithm; transmit the determined output to a computing device associated with the energy source for formulating an energy delivery algorithm; and use the instruments to deliver additional electrosurgical energy from the energy source to the blood vessel according to the energy delivery algorithm to seal the blood vessel.

[0018] In a further aspect of this disclosure, the machine learning algorithm may include a neural network.

[0019] In another aspect of this disclosure, the neural network may include a feedforward network, a convolutional network, or a recurrent network.

[0020] In a further aspect of this disclosure, the instructions, when executed, may also cause the system to train the neural network based on one or more patterns in the measured sensor data and / or identified data.

[0021] In a further aspect of this disclosure, the instructions, when executed, may also enable the system to train the neural network based on training data, including impedance, vascular temperature, vascular mass, vascular surface area, cumulative energy, and / or rupture pressure probability.

[0022] In another aspect of this disclosure, training may include supervised learning, unsupervised learning, and / or reinforcement learning.

[0023] In a further aspect of this disclosure, reinforcement learning may include rewards based on rupture pressure probability values ​​and / or impedance penalty values.

[0024] In a further aspect of this disclosure, the rupture pressure probability value is between 0 and 1. The rupture pressure probability value can be determined by: increasing the rupture pressure probability value by a first amount when the blood vessel temperature is within a first temperature range for a first predetermined time period of protein denaturation; increasing the rupture pressure probability value by a second amount when the blood vessel temperature is within a second temperature range for a second predetermined time period to remove a predetermined percentage of water; and increasing the rupture pressure probability value by a third amount when the blood vessel temperature is within a third temperature range for a third predetermined time period to allow thermosetting gelatin coagulation and jaw cooling.

[0025] In another aspect of this disclosure, the reward may also include an impedance penalty value between -1 and 0. The impedance penalty value can be determined by setting it to -1 when the blood vessel impedance is greater than a first impedance threshold. When the blood vessel impedance is less than the first impedance threshold, the impedance penalty value can be determined by setting it to 0.

[0026] In one aspect of this disclosure, when it is determined that the blood vessel is adequately sealed, the instruction, when executed, can also cause the system to stop delivering electrosurgical energy or additional electrosurgical energy.

[0027] According to aspects of this disclosure, a non-transitory storage medium stores a program that causes a computer to execute a computer-implemented method for controlling the delivery of electrosurgical energy to a blood vessel to seal it. The computer-implemented method includes collecting data from an electrosurgical system comprising an instrument and an energy source as an instrument delivers electrosurgical energy from an energy source to the blood vessel to seal it. The method further includes transmitting the data to a machine learning algorithm that predicts the probability of rupture pressure in the blood vessel and determines, based on the prediction, whether the blood vessel is adequately sealed. If it is determined that the blood vessel is not adequately sealed, the method further includes determining an output based on the data using a second machine learning algorithm, transmitting the determined output to a computing device associated with the energy source for formulating an energy delivery algorithm, and using the instrument to deliver additional electrosurgical energy from the energy source to the blood vessel to seal it. The data includes electrical parameters related to the delivery of electrosurgical energy according to the energy delivery algorithm. Attached Figure Description

[0028] Various aspects and features of this disclosure are described below with reference to the accompanying drawings, in which:

[0029] Figure 1A This is a perspective view of a surgical system provided in this disclosure, which includes energy-based surgical instruments and a generator;

[0030] Figure 1BThis is a perspective view of another surgical system provided in this disclosure, which includes surgical instruments and a generator based on another energy source;

[0031] Figure 2 Based on this disclosure Figure 1A and 1B A block diagram of the system's generator;

[0032] Figure 3 yes Figure 1A A cross-sectional view of the end effector assembly of an energy-based surgical instrument in a surgical system, illustrating the grasping energy and the application of energy to tissue according to the present disclosure;

[0033] Figure 4 It is a logic diagram based on the machine learning algorithm disclosed in this invention;

[0034] Figure 5 It is provided and configured in accordance with this disclosure to be compatible with Figure 1A A block diagram of the controller used in the surgical system;

[0035] Figure 6 This is a flowchart of a method for controlling the delivery of electrosurgical energy to a blood vessel to seal the blood vessel, according to the present disclosure;

[0036] Figure 7 It is a flowchart of reinforcement learning provided in this disclosure;

[0037] Figure 8 It is a diagram of energy-organization interactions provided in this disclosure;

[0038] Figures 9A-9F These are various graphs illustrating energy-organization interactions provided in this disclosure;

[0039] Figure 10 It is a graph of rupture pressure probability versus time for various blood vessel diameters provided in this disclosure;

[0040] Figure 11 It is a power versus time graph based on various blood vessel diameters provided in this disclosure;

[0041] Figure 12 It is a graph of tissue temperature versus time for various blood vessel diameters provided in this disclosure;

[0042] Figure 13 It is a time-varying graph of tissue impedance for various vessel diameters provided in this disclosure; and

[0043] Figure 14 It is a graph of cumulative energy versus time for various blood vessel diameters provided in this disclosure. Detailed Implementation

[0044] Tissue sealing involves heating the tissue to liquefy the collagen and elastin within, causing it to reform into a fused mass and significantly reducing the boundaries between opposing tissue structures. To achieve tissue sealing without causing unnecessary damage to the tissue at the surgical site or collateral damage to adjacent tissues, it is necessary to control the energy applied to the tissue, thereby controlling the tissue temperature during the sealing process. For proper tissue sealing, a balance must be struck between maintaining sufficient heating during the sealing process to denature proteins and evaporate fluids, and avoiding unnecessary / collateral damage.

[0045] Regarding the real-time utilization of vascular rupture pressure probability information to control energy application to tissue for tissue sealing, it is desirable to determine the rupture pressure probability during the tissue sealing process. As described below, this can be achieved by utilizing data already available from the electrosurgical system and running a machine learning algorithm to estimate the rupture pressure probability based on said data. The estimated rupture pressure probability can then be fed back to the controller for controlling energy application to the tissue accordingly. As used herein, the rupture pressure probability is the probability that a sealed vessel will not rupture under a threshold fluid pressure supplied through the vessel. For example, the rupture pressure probability indicating a suitable vessel seal could be a 95% probability of a rupture pressure greater than 360 mmHg, although other suitable probabilities and / or pressures may also be considered.

[0046] The systems and methods of this disclosure, detailed below, can be incorporated into any type of surgical system for treating tissue, such as the electrosurgical system detailed below. For illustrative purposes and in no way limit the scope of the appended claims, this disclosure describes systems and methods for estimating the probability of rupture pressure applied to tissue by controlling electrosurgical energy within the context of an electrosurgical system.

[0047] Now for reference Figure 1A and 1B , Figure 1A An electrosurgical system is depicted, comprising an axis-based surgical forceps 100 and an electrosurgical generator 160 used therewith. Figure 1B An electrosurgical system comprising a hemostatic surgical forceps 200 and an electrosurgical generator 160 used therewith is described. For the purposes of this document, forceps 100, forceps 200, or any other suitable surgical instrument and / or system may be used according to this disclosure. It will be apparent that different electrical and mechanical connections and considerations apply to each particular type of instrument and system; however, aspects and features of this disclosure remain generally consistent regardless of the configuration of the instrument or system used with it.

[0048] Now go to Figure 1AThe forceps 100 defines a longitudinal axis “AA” and includes a housing 120, a handle assembly 130, a rotary assembly 170, a trigger assembly 180, and an end effector assembly 10. The forceps 100 further includes a shaft 112 having a distal end 114 configured to mechanically engage with the end effector assembly 10 and a proximal end 116 configured to mechanically engage with the housing 120. The forceps 100 may also include a surgical cable extending therefrom and configured to connect the forceps 100 to an electrosurgical generator 160, such that at least one of the conductive tissue handling surfaces 13, 14 of the jaw members 11, 12 of the end effector assembly 10 can be energized, for example, when an activation switch 190 is activated, to handle tissue clamped therein.

[0049] Continue to refer to Figure 1A The handle assembly 130 includes a fixed handle 150 and a movable handle 140. The fixed handle 150 is integrally associated with the housing 120, and the handle 140 is movable relative to the fixed handle 150. The rotating assembly 170 is rotatable in either direction about the longitudinal axis "AA" to cause the end effector assembly 10 to rotate about the longitudinal axis "AA". The housing 120 houses the internal working parts of the pliers 100.

[0050] An end effector assembly 10 is shown attached to the distal end 114 of a shaft 112 and includes a pair of opposing jaw members 11 and 12. Each of the jaw members 11 and 12 includes conductive tissue-processing surfaces 13 and 14, respectively, configured to grip tissue therebetween and conduct electrosurgical energy therebetween to process, for example, sealed tissue. The end effector assembly 10 is designed as a unilateral assembly, i.e., jaw member 12 is fixed relative to the shaft 112, and jaw member 11 is movable relative to the shaft 112 and the fixed jaw member 12. However, the end effector assembly 10 can alternatively be configured as a bidirectional assembly, i.e., in which both jaw members 11 and 12 are movable relative to each other and relative to the shaft 112. In some embodiments, a blade assembly (not shown) is disposed within a shaft 112, and a blade channel (not shown) is defined within one or two jaw members 11, 12 to allow a blade (not shown) to reciprocate through it, for example, to cut tissue disposed between jaw members 11, 12 when a trigger 182 of the trigger assembly 180 is activated.

[0051] Continue to refer to Figure 1A The movable handle 140 of the handle assembly 130 is ultimately connected to a drive assembly (not shown), which together mechanically cooperate to move the jaw members 11 and 12 between spaced and proximal positions to grip tissue between the tissue processing surfaces 13 and 14 of the jaw members 11 and 12. Figure 1AAs shown, the movable handle 140 is initially spaced apart from the fixed handle 150, and correspondingly, the jaw members 11 and 12 are in a spaced-apart position. The movable handle 140 can be pressed down from this initial position to a pressed-down position corresponding to the proximity position of the jaw members 11 and 12.

[0052] Now for reference Figure 1B The diagram shows a pair of pliers 200, which includes two elongated shafts 212a and 212b, each having proximal ends 216a and 216b and distal ends 214a and 214b, respectively. The pliers 200 is configured for use with an end effector assembly 20 similar to the end effector assembly 10 of the pliers 100 (see [link to diagram]). Figure 1A More specifically, the end effector assembly 20 is attached to the distal ends 214a and 214b of shafts 212a and 212b, respectively, and includes a pair of jaw members 21 and 22 that are movable relative to each other. Each shaft 212a and 212b includes handles 217a and 217b disposed at its proximal ends 216a and 216b. Each handle 217a and 217b defines finger holes 218a and 218b therethrough for receiving a user's finger. It is understood that the finger holes 218a and 218b facilitate the movement of shafts 212a and 212b from an open position relative to each other to a closed position, in which jaw members 21 and 22 are arranged spaced apart relative to each other, and in the closed position, jaw members 21 and 22 cooperate to clamp tissue between them.

[0053] A ratchet 230 may be included for selectively locking the jaw members 21 and 22 of the pliers 200 in various different positions relative to each other. It is foreseeable that the ratchet 230 may include scales or other visual markings that allow the user to easily and quickly determine and control the amount of desired closing force between the jaw members 21 and 22.

[0054] Continue to refer to Figure 1B One of the shafts may be adapted to receive a surgical cable configured to connect the forceps 200 to the electrosurgical generator 160. As will be described in more detail below, the electrosurgical generator 160 provides power to the end effector assembly 20 such that at least one of the conductive tissue processing surfaces 23, 24 of the jaw members 21, 22 of the end effector assembly 20 can be energized to process, for example, tissue sealed therein.

[0055] Similar to pliers 100 ( Figure 1A The pliers 200 may also include a knife assembly (not shown) disposed within either of the shafts 212a, 212b and a knife passage (not shown) defined within one or two jaw members 21, 22 to allow the blade (not shown) to reciprocate therefrom.

[0056] Now for reference Figure 2 A block diagram of exemplary components of an electrosurgical generator 160 according to aspects of this disclosure is shown. In the illustrated embodiment, the generator 160 includes a controller 500, a power supply 164, a radio frequency (RF) energy output stage 162, a sensor module 166, one or more analog-to-digital converters (ADCs) 168, and one or more connector ports 169 that accommodate various types of electrosurgical instruments, such as forceps 10 and forceps 200 (respectively). Figure 1A and 1B Generator 160 may include a user interface (not shown) that allows a user to select various parameters for generator 160, such as operating mode and power settings. In various embodiments, the user can specify the power setting between zero and a power limit, such as five watts, thirty watts, seventy watts, or ninety-five watts.

[0057] The electrosurgical generator 160 can be any suitable type of generator to accommodate various types of electrosurgical instruments (e.g., monopolar and bipolar electrosurgical instruments). The electrosurgical generator 160 can also be configured to operate in various modes, such as ablation, cutting, coagulation, and sealing. The electrosurgical generator 160 may include a switching mechanism (e.g., a relay) to switch the supply of RF energy between connectors 169 to which various electrosurgical instruments can be connected. For example, when an electrosurgical instrument such as forceps 100 ( Figure 1A ) or pliers 200 ( Figure 1B When connected to the electrosurgical generator 160, the switching mechanism switches the supply of RF energy to the appropriate plug 169. In an embodiment, the electrosurgical generator 160 may be configured to supply RF energy to multiple instruments simultaneously.

[0058] In various embodiments, generator 160 may include sensor module 166, which includes multiple sensors, such as RF current sensors and RF voltage sensors. The components of generator 160, namely RF output stage 162 and the RF current and voltage sensors of sensor module 166, may be disposed on a printed circuit board (PCB). The RF current sensor of sensor module 166 may be coupled to an active terminal and provide a measurement of the RF current provided by RF output stage 162. In an embodiment, the RF current sensor of sensor module 166 may be coupled to a return terminal. The RF voltage sensor of sensor module 166 is coupled to both the active and return terminals and provides a measurement of the RF voltage provided by RF output stage 162. In an embodiment, the RF current and voltage sensors of sensor module 166 may be coupled to active and return leads that interconnect the active and return terminals and to RF output stage 162, respectively.

[0059] The RF current and voltage sensors of sensor module 166 sense and provide the sensed RF voltage and current signals to controller 500 of generator 160, which can then adjust the output of power supply and / or RF output stage 162 in response to the sensed RF voltage and current signals. Controller 500 can additionally or alternatively determine power, impedance, and / or other characteristics based on the sensed RF voltage and current, and adjust the output of power supply and / or RF output stage 162 accordingly. Controller 500 will be described in more detail below (see...). Figure 5 ).

[0060] The sensed voltage and current from sensor module 166 are fed to ADC 168. ADC 168 samples the sensed voltage and current to obtain digital samples of the voltage and current of RF output stage 162. The digital samples are processed by controller 500 and used to generate control signals to control the DC / AC inverter and preamplifier of RF output stage 162. ADC 168 transmits the digital samples to controller 500 for further processing.

[0061] Now for reference Figure 3 , showed Figure 1A A cross-sectional view of the end effector assembly 10 of the pliers 100. Jaw members 11 and 12 clamp tissue 302 between the conductive surfaces 13 and 14 of the jaw members 11 and 12. Generator 160 ( Figure 2 Electrosurgical energy is supplied to conductive surfaces 13 and 14 at different potentials, such that the electrosurgical energy is conducted between them and through the grasped tissue 302 to heat and thereby treat (e.g., seal) the tissue 302. As described above, by controlling the energy from generator 160... Figure 2 The application of energy to surfaces 13 and 14, and the pressure applied to the tissue via surfaces 13 and 14, allows for the control of heating of tissue 302 to achieve tissue sealing. Also as described above, this disclosure provides a system and method for estimating the probability of rupture pressure for controlling the application of energy to tissue 302 (e.g., a blood vessel) using available data. This is advantageous because, although the probability of rupture pressure cannot be directly measured using conventional sensors, it provides an estimated probability of rupture pressure in real time, indicating the effectiveness of the tissue seal.

[0062] refer to Figure 4 The present disclosure illustrates a logic diagram of a machine learning algorithm 408. The training of the machine learning algorithm 408 may be based on sensor measurements 402, generator control parameters 404, and / or tissue parameters 406 as inputs to the machine learning algorithm 408 (e.g., a neural network). The machine learning algorithm 408 outputs an estimate 410 of the rupture pressure probability.

[0063] In various embodiments, generator control parameters 404 associated with a specific sensor measurement 402 are used as input to machine learning algorithm 408 during training. In various embodiments, generator control parameters 404 may include, for example, power, current, voltage, time, the slope of power, current, and / or voltage, or other generator 160 parameters. In various embodiments, tissue parameters may include impedance, tissue temperature, or accumulated energy from one or more of the following non-limiting list. In various embodiments, controller 500 may, for example, communicate with a remote server that stores adjusted control parameters, text data, and / or the output of machine learning algorithm 408.

[0064] In various embodiments, the output of the machine learning algorithm 408, such as a neural network, can be used as training data for reinforcement learning. It is conceivable that training can be performed on a separate system, such as a GPU server, simulation, etc., and then the trained network can be deployed in the surgical system. In various embodiments, the controller 500 outputs an estimated probability of blood vessel rupture pressure from the machine learning algorithm 408.

[0065] refer to Figure 5 The diagram illustrates a controller 500 according to the present disclosure. The controller 500 includes a processor 520 connected to a computer-readable storage medium or memory 530, which may be a volatile type of memory such as RAM, or a non-volatile type of memory such as flash memory, disk media, etc. In various embodiments, the processor 520 may be another type of processor, such as, but not limited to, a digital signal processor, microprocessor, ASIC, graphics processing unit (GPU), field-programmable gate array (FPGA), or central processing unit (CPU). In various embodiments, in contrast to the processor, network inference may also be performed in a system where weights can be implemented as mediated, chemically, or otherwise inference computations.

[0066] In various embodiments, memory 530 may be random access memory, read-only memory, disk storage, solid-state storage, optical disk storage, and / or other types of memory. In various embodiments, memory 530 may be decoupled from controller 500 and may communicate with processor 520 via a communication bus on a circuit board and / or via a communication cable, such as a serial ATA cable or other types of cable. Memory 530 contains computer-readable instructions executable by processor 520 to operate controller 500. In various embodiments, controller 500 may include a network interface 540 for communicating with other computers or servers. In embodiments, storage device 510 may be used to store data. In various embodiments, controller 500 may include one or more FPGAs 550. FPGA 550 may be used to execute various machine learning algorithms, such as those provided in this disclosure, as detailed below.

[0067] Memory 530 stores appropriate instructions that are executed by processor 520 for use via ADC 168 (see...) Figure 2 The controller 500 receives sensing data, such as sensing data from sensor module 166, accesses storage device 510 of the controller 500, determines one or more tissue parameters, such as rupture pressure probability, based on the sensing data and information stored in storage device 510, and provides feedback based on the determined tissue parameters. Although shown as part of generator 160, it is also conceivable that the controller 500 is remote from generator 160, for example, on a remote server, and can be accessed by generator 160 via a wired or wireless connection. In embodiments where the controller 500 is remote, it is conceivable that the controller 500 can be accessed by and connected to multiple generators 160.

[0068] The storage device 510 of the controller 500 stores one or more machine learning algorithms and / or models, which are configured to be based on an ADC 168 (see...). Figure 2 One or more tissue parameters, such as the rupture pressure probability, can be estimated from sensor circuitry, such as sensing data received from sensor module 166. Machine learning algorithms can be trained on and learn from experimental data and / or data from previous processes initially input into one or more machine learning applications, enabling the machine learning applications to estimate tissue parameters based on such data. Such data may include tissue impedance data, power data, time, and / or any other suitable data.

[0069] General reference Figure 2-5 Machine learning algorithms are advantageous for determining vascular parameters, at least because they do not require complex sensor components and predefined classification rules and / or algorithms. Instead, machine learning algorithms utilize initial input data, such as previous process data, current process data, and / or experimental data, to determine statistical characteristics and / or correlations, enabling them to determine vascular parameters of unknown vessels by analyzing the data derived therefrom. Therefore, when one or more machine learning algorithms are trained as described above, for example using end effector component 10, they can be used to determine the parameters of a sealed vessel. More specifically, the processor 520 of controller 500 is configured to, in response to receiving sensed data from sensor circuitry (e.g., from sensor module 166) via ADC 168, input the sensed data into a machine learning algorithm stored in storage device 510 to determine one or more vascular parameters of a vessel sealed using end effector component 10. Although described with respect to an electrosurgical system, the aspects and features of controller 500, and the machine learning algorithms configured therewith, are equally applicable to use with other suitable surgical systems, such as ablation systems and / or ultrasound systems.

[0070] Once the vascular parameters are determined by the controller 500, depending on the vascular parameters, settings, user input, etc., the controller 500 may, for example, output alarms and / or warnings to the user interface, implement, switch or modify a specific energy delivery algorithm based on its power supply 164 and RF output stage 162 to provide energy to the end effector assembly 10, and / or disable further energy delivery to the end effector assembly 10.

[0071] The terms "artificial intelligence," "data model," or "machine learning" may include, but are not limited to, neural networks, deep neural networks, recurrent neural networks (RNNs), generative adversarial networks (GANs), Bayesian regression, Naive Bayes, Monte Carlo methods, nearest neighbor methods, least squares methods, mean methods, and support vector regression methods, as well as other data science and artificial science techniques. An exemplary use is pattern recognition and predictions related to vascular parameters, which will be described in more detail below.

[0072] The term "application" can include computer programs designed to perform specific functions, tasks, or activities to benefit a user. An application can refer to software, for example, as a standalone program or running locally or remotely in a web browser, or other software that a person skilled in the art would understand as an application. An application can run on the controller 500 or on a user device, including, for example, on a mobile device, IoT device, or server system.

[0073] Now for reference Figure 6 A flowchart of a computer-implemented method 600 for controlling the delivery of electrosurgical energy to a blood vessel to seal the vessel is shown. Those skilled in the art will understand that one or more operations of method 600 may be performed, repeated, and / or omitted in different orders without departing from the scope of this disclosure. In various embodiments, the illustrated method 600 can be controlled by a controller 500 (… Figure 5 This can be performed on a remote device or on another server or system. In various embodiments, some or all of the operations in the illustrated method 600 may be performed using an electrosurgical system such as instrument 100 or 200 and generator 160 (see [link]). Figure 1A and 1B The operation is performed as described above. Other variations are considered within the scope of this disclosure. Regarding the controller, such as the controller 500 of the generator 160... Figure 2 and 5 The operation of method 600 is described, but it should be understood that the operation shown also applies to other systems and their components.

[0074] First, in step 602, the controller 500 can collect data from the electrosurgical system, which includes the instrument and the energy source, while the instrument delivers electrosurgical energy from the energy source to the blood vessel to seal it. This data may include at least one parameter associated with the delivery of electrosurgical energy. For example, the data may include impedance, power, voltage, current, blood vessel temperature, blood vessel mass, blood vessel surface area, accumulated energy, or rupture pressure probability.

[0075] For example, an electrosurgical system may include a generator and electrosurgical instruments, such as those mentioned above. Figure 1A and 1B Detailed description. When surgeons operate an electrosurgical system during surgery, they can use the system to apply electrosurgical (RF) energy to blood vessels to seal them. For more details, see also [link to relevant documentation]. Figure 2 Tissue 302 (e.g., blood vessels) can be gripped by jaw components 11, 12 (or Figure 1B Between the conductive tissue processing surfaces 13 and 14 of the clamping members 21 and 22, electrosurgical (RF) energy can be conducted between the tissue processing surfaces 13 and 14 and heated through the tissue 302 to seal the blood vessel 302. During such sealing, the sensor circuitry of the generator 160, such as sensor module 166, can sense parameters of the blood vessel and / or energy, such as impedance and power, and / or provide data from which impedance and / or power can be derived, such as time, voltage, and / or current data. Pressure is also expected to be sensed or determined. This can occur as a snapshot or over a time interval and can be determined at the start of the blood vessel sealing, for example, within 250 ms after the start of blood vessel closure, to avoid permanent damage to the tissue if it is determined that the applied energy should be interrupted. The sensed data may include, for example, the time of power application, the power applied to the tissue, and / or the impedance of the blood vessel. Sensor module 166 can measure data from the electrosurgical system, such as the voltage and / or current of the electrosurgical energy delivered to the blood vessel. In various embodiments, voltage and current can be used to derive power and / or impedance. The sensing data obtained by the sensor circuit can be relayed to the controller 500 (in this embodiment, via the ADC 168) for further processing, as detailed below.

[0076] In step 604, the controller 500 feeds data as input to a machine learning algorithm, such as a neural network. In embodiments, training the neural network can be accomplished by recognizing patterns in the shape of the impedance curve and / or in the power-to-time curve. In various embodiments, the neural network can be used to train data, such as jaw members 11, 12 (…). Figure 1A ) or jaw components 21, 22 ( Figure 1BThe impedance, power, time, tissue electrical properties, tissue thermal properties, electrical properties of the electrosurgical device, or thermal properties of the electrosurgical device can be used as input data. In various embodiments, the output of the neural network can be used as training data for supervised learning. It is conceivable that training can be performed on a separate system, such as a GPU server, a simulation, etc., and then the trained network can be deployed in the electrosurgical system 1100.

[0077] In step 606, controller 500 uses a machine learning algorithm to estimate the rupture pressure probability of the sealed blood vessel. For example, once the blood vessel is sealed, the rupture pressure probability is a measure of the likelihood that the vessel will not rupture when a fluid pressure below a threshold pressure is applied. For example, determining a target rupture pressure probability for a properly sealed blood vessel could be a 95% probability that the rupture pressure is greater than 360 mmHg, although other suitable probabilities and / or pressures (double, triple, etc. systolic blood pressure) could also be considered. In various embodiments, the rupture pressure probability can be estimated using a machine learning algorithm. The rupture pressure probability indicates the likelihood that the sealed blood vessel will fail, and is therefore a measure of whether the seal is adequate. That is, by estimating the rupture pressure probability, controller 500 can determine whether the blood vessel is adequately sealed, or if not, whether energy needs to be continued until the rupture pressure probability is reached.

[0078] In various embodiments, the neural network may include a feedforward network, a convolutional network, or a recurrent network. In various embodiments, the neural network may be trained using one or more patterns from measurement sensor data or recognition data. In various embodiments, training the machine learning algorithm may be performed by a computing device external to generator 160, and the resulting algorithm may be transmitted to controller 500 of generator 160.

[0079] In various embodiments, training data may be used to train the neural network, including, for example, impedance, tissue temperature, tissue mass, tissue surface area, accumulated energy, or rupture pressure probability. In various embodiments, training may include at least one of supervised learning, unsupervised learning, or reinforcement learning.

[0080] Reinforcement learning is a field of machine learning that relates to the concept of how a software agent should act in an environment to maximize cumulative rewards. For example, the following illustrative example can help understand the basic approach behind reinforcement learning: First, the agent observes and constructs its own representation of the environment (state). Next, the agent takes some action and observes the response received in the environment (updated state). If the agent receives an undesirable response from an action taken in a given environment, the agent receives a negative reward and is less likely to take that particular action again in a given state. If the agent receives a desirable response from an action taken in a given environment, the agent receives a positive reward and is more likely to take that particular action again in a given state. This process is repeated until the agent finds a policy (the action to take in different situations) that maximizes the total cumulative reward.

[0081] In various embodiments, the neural network may include, for example, a two-layer feedforward network. In various embodiments, the neural network may include a feedforward network, a convolutional network, and / or a recurrent network.

[0082] In step 608, controller 500 determines whether the blood vessel is adequately sealed based on the estimated rupture pressure probability. In step 610, if controller 500 determines that the blood vessel is not adequately sealed, a machine learning algorithm outputs an indication, such as an estimate of the rupture pressure probability itself, a relative measure of the difference between the estimated rupture pressure probability and the target rupture pressure probability, or otherwise.

[0083] In step 612, the controller 500 transmits the instructions output from the machine learning algorithm to a computing device associated with, for example, the energy source of the controller 500, for developing (e.g., switching, confirming, modifying, generating, etc.) a vascular sealing algorithm.

[0084] In step 614, the instrument delivers additional electrosurgery from the energy source to the blood vessel to seal it according to the vascular sealing algorithm. When it is determined that the blood vessel is adequately sealed (e.g., after one or more iterations of the above method), the machine learning algorithm outputs its indication, which is transmitted by the controller 500 to the computing device to, for example, stop the energy supply and end the vascular sealing process.

[0085] refer to Figure 7The diagram illustrates a block diagram of reinforcement training according to this disclosure. As mentioned above, reinforcement learning is a field of machine learning related to the concept of how a software agent should act in an environment to maximize cumulative rewards. First, when sealing a blood vessel, power can be applied to the blood vessel 706 via an electrosurgical instrument. Next, agent 702 can observe and construct its own representation of the blood vessel 706 based on blood vessel parameters (state 708), such as cumulative energy, impedance, or temperature. Next, agent 702 will take certain actions 704 (e.g., apply energy, stop applying energy, change the application of energy) and observe the received response (e.g., estimated rupture pressure probability) in a given environment. Controller 500 can determine rewards and / or penalties 710 for training the neural network 712 in reinforcement learning 700.

[0086] The reward may include a value based on the rupture pressure probability, and the penalty may include a value below a threshold based on impedance. More specifically, the reward may include a rupture pressure probability value between 0 and 1. In various embodiments, when a blood vessel temperature is maintained at a first temperature range (e.g., 70°C–80°C) for a first predetermined time period (e.g., 100 ms) for protein denaturation, the rupture pressure probability value may include an increase of, for example, about 0.33.

[0087] In various embodiments, when the blood vessel temperature is maintained at a second temperature range (e.g., between 140°C and 152°C) for a second predetermined time period to remove a certain percentage of water (e.g., 25%), the rupture pressure probability value can increase, for example, by about 0.33. In various embodiments, when the blood vessel temperature is maintained at a third temperature range (e.g., between 40°C and 50°C) for a third predetermined time period (e.g., 100 ms) to allow the thermosetting gelatin to solidify and the jaw members to cool, the rupture pressure probability value can increase by about 0.33.

[0088] In various embodiments, the impedance penalty value can be between -1 and 0. In various embodiments, the impedance penalty value can be set to -1 when the impedance of the blood vessel is greater than a first impedance threshold (e.g., 4000 ohms). In various embodiments, the impedance penalty value can be set to 0 when the impedance of the blood vessel is less than the first impedance threshold (e.g., 4000 ohms).

[0089] In various embodiments, the neural network 712 may determine action 704 based on reward and penalty 710. For example, the action may include increasing or decreasing the power applied to the blood vessel to seal it. In various embodiments, the controller 500 may determine that the blood vessel is adequately sealed, and action 704 may include stopping the delivery of electrosurgical energy or additional electrosurgical energy. Consideration of penalties may include a blood vessel temperature exceeding a predetermined threshold, a voltage exceeding a predetermined threshold, and / or the detection of an electric arc.

[0090] refer to Figure 8 The diagram illustrates the energy-tissue interaction according to this disclosure. In various embodiments, the temperature of the tissue can increase when energy is applied to the tissue (such as a blood vessel) for sealing.

[0091] refer to Figure 9A -F, showing various graphs illustrating energy-tissue interactions according to this disclosure. In various embodiments, power can be applied to the blood vessels over time, such as... Figure 9B As shown. In various embodiments, Figure 9A The action can be the power applied to tissue (such as a blood vessel) to seal the vessel. In various embodiments, such as Figure 9C As shown, the temperature of a blood vessel can increase or decrease depending on the applied power over time. The probability of pressure rupture can increase as more power is applied to the blood vessel over time. Figure 9D As shown. After sealing, the impedance (ohms) of the blood vessel can increase, such as... Figure 9E As shown. For example, as moisture is removed during the sealing process, the impedance can increase from a low impedance to approximately 400 ohms. In various embodiments, such as Figure 9F As shown, the cumulative energy (J) can increase as energy is applied to the blood vessels over time.

[0092] refer to Figure 10-14 Using a machine learning algorithm according to this disclosure, graphs of various parameters for various blood vessel diameters during vascular sealing are shown. Figure 10 The graphs show the relationship between rupture pressure probability and time for various blood vessel diameters (e.g., 1 mm, 3 mm, 5 mm, 7 mm, and 9 mm). For example, as energy is applied to the blood vessel over time, the probability that the vessel will not rupture below a threshold increases, and / or the threshold itself may increase. Figure 11 The diagram illustrates the relationship between action and time for various blood vessel diameters. For example, this action can include the power applied to the blood vessel over time. Blood vessels of different diameters may require different amounts of power to achieve a proper seal. Figure 12 The graph shows the relationship between blood vessel temperature and time for various vessel diameters. For example, for vessels of different diameters, the blood vessel temperature can vary differently as power is applied over time during the sealing process. Figure 13 The graph shows the relationship between vascular impedance and time for various vessel diameters. For example, for vessels of different diameters, the vascular impedance can vary differently as power is applied over time during the sealing process. Figure 14 The graph shows the relationship between cumulative energy and time for various blood vessel diameters. Cumulative energy refers to the energy applied to the blood vessel over time (i.e., ∫power dt).

[0093] From the foregoing and with reference to the accompanying drawings, those skilled in the art will understand that modifications may be made to this disclosure without departing from its scope. While several embodiments of this disclosure have been shown in the figures, it is not intended to limit this disclosure, as it is intended to be as broad as permitted in the art and this specification should be read in the same manner. Therefore, the foregoing description should not be construed as restrictive, but merely as illustrative of specific embodiments. Other modifications will be contemplated by those skilled in the art within the scope and spirit of the appended claims.

Claims

1. A device for controlling the delivery of electrosurgical energy to a blood vessel to seal the blood vessel, the device comprising: A means for collecting data from an electrosurgical system including the instrument and the energy source while the instrument delivers electrosurgical energy from an energy source to a blood vessel to seal the blood vessel, the data including at least one electrical parameter related to the delivery of the electrosurgical energy; A means for transmitting the data to a first neural network; A device for predicting the probability of blood vessel rupture pressure using the first neural network based on the following: A representation of blood vessels is constructed based on vascular parameters including accumulated energy, impedance, and temperature; Perform an action that includes at least one of applying energy, stopping the application of energy, or changing the application of energy; Receive additional data collected after the action is executed as a response; Based on the response, the probability value of the rupture pressure is predicted; Determine a reward or penalty, where the reward is based on the probability value of rupture pressure and the penalty is based on the resistance of the blood vessel; as well as The first neural network is trained based on rewards or penalties; and Device for determining whether the blood vessel is adequately sealed based on the prediction; In cases where it is determined that the blood vessel is not adequately sealed, the device includes: A means for determining an output based on the data using a second neural network; A means for transmitting a determined output to a computing device associated with the energy source for developing an energy delivery algorithm; and A device for using the instrument to deliver additional electrosurgical energy from the energy source to the blood vessel according to the energy delivery algorithm to seal the blood vessel.

2. The device of claim 1, wherein at least one of the first or second neural networks comprises at least one of a feedforward network, a convolutional network, or a recurrent network.

3. The apparatus of claim 1, wherein the apparatus further comprises means for training the second neural network based on one or more patterns in the measurement sensor data or identification data.

4. The device of claim 1, wherein the device further comprises means for training at least one of the first or second neural networks based on training data, the training data including at least one of: vascular impedance, vascular temperature, vascular mass, vascular surface area, cumulative energy, or rupture pressure probability.

5. The device of claim 4, wherein the training comprises at least one of supervised learning, unsupervised learning, or reinforcement learning.

6. The device of claim 1, wherein the burst pressure probability value is a scaler and is determined by: When the blood vessel temperature is within a first temperature range for a first predetermined time period for protein denaturation, the rupture pressure probability value is increased by a first amount. When the blood vessel temperature is within a second temperature range for a second predetermined time period to remove a predetermined percentage of water, the rupture pressure probability value is increased by a second amount; and When the blood vessel temperature is within a third temperature range for a third predetermined time period to allow thermosetting gelatin to coagulate and the jaws to cool, the rupture pressure probability value is increased by a third amount.

7. The device of claim 6, wherein the blood vessel-based impedance penalty is a second scaler and is determined by: If the blood vessel impedance is greater than the first impedance threshold, the penalty value is set to -1; and If the impedance of the blood vessel is less than the first impedance threshold, the penalty value is set to 0.

8. The device of claim 1, further comprising means for stopping the delivery of the electrosurgical energy and / or the additional electrosurgical energy, provided that the blood vessel is adequately sealed.

9. A system for controlling electrosurgical energy, the system comprising: Electrosurgical systems, including instruments and power sources; One or more processors; and At least one memory coupled to the one or more processors, the at least one memory having instructions stored thereon, the instructions causing the system to: While the instrument delivers electrosurgical energy from the energy source to the blood vessel to seal the blood vessel, data is collected from the electrosurgical system, the data including at least one electrical parameter related to the delivery of the electrosurgical energy; The data is transmitted to the first neural network; The first neural network predicts the probability of blood vessel rupture pressure based on the following: A representation of blood vessels is constructed based on vascular parameters including accumulated energy, impedance, and temperature; Perform an action that includes at least one of applying energy, stopping the application of energy, or changing the application of energy; Receive additional data collected after the action is executed as a response; Based on the response, the probability value of the rupture pressure is predicted; Determine a reward or penalty, where the reward is based on the probability value of rupture pressure and the penalty is based on the resistance of the blood vessel; as well as The first neural network is trained based on rewards or penalties; and Based on the prediction, determine whether the blood vessel is adequately sealed; If it is determined that the blood vessel is not adequately sealed: The output is determined based on the data using a second neural network; The determined output is transmitted to the computing device associated with the energy source for the purpose of developing an energy delivery algorithm; and Using the device, additional electrosurgical energy is delivered from the energy source to the blood vessel according to the energy delivery algorithm to seal the blood vessel.

10. The system of claim 9, wherein at least one of the first neural network or the second neural network comprises at least one of a feedforward network, a convolutional network, or a recurrent network.

11. The system of claim 10, wherein the instructions, when executed, further cause the system to train at least one of the first or second neural networks based on one or more patterns in the measurement sensor data or identification data.

12. The system of claim 11, wherein the instructions, when executed, further cause the system to train the first or second neural network based on training data, the training data including at least one of the following: vascular impedance, vascular temperature, vascular mass, vascular surface area, accumulated energy, or rupture pressure probability.

13. The system of claim 12, wherein the training includes at least one of supervised learning, unsupervised learning, or reinforcement learning.

14. The system of claim 9, wherein the rupture pressure probability value is between 0 and 1, and is determined by: When the blood vessel temperature is within a first temperature range for a first predetermined time period for protein denaturation, the rupture pressure probability value is increased by a first amount. When the blood vessel temperature is within a second temperature range for a second predetermined time period to remove a predetermined percentage of water, the rupture pressure probability value is increased by a second amount; and While the blood vessel temperature is within a third temperature range for a third predetermined time period to allow for thermosetting gelatin coagulation and clamp cooling, the rupture pressure probability value is increased by a third amount; and The penalty value is between -1 and 0, and is determined by the following: If the blood vessel impedance is greater than a first impedance threshold, the penalty value is set to -1; and If the impedance of the blood vessel is less than the first impedance threshold, the penalty value is set to 0.

15. The system of claim 9, wherein, upon determining that the blood vessel is adequately sealed, the instruction, when executed, further causes the system to stop delivering the electrosurgical energy or the additional electrosurgical energy.

16. A non-transitory storage medium storing a program for enabling a computer to execute a computer-executed method for controlling the delivery of electrosurgical energy to a blood vessel to seal the blood vessel, the computer-implemented method comprising: While the instrument delivers electrosurgical energy from an energy source to a blood vessel to seal the vessel, data is collected from an electrosurgical system including the instrument and the energy source, the data including at least one electrical parameter related to the delivery of the electrosurgical energy; The data is transmitted to the first neural network; The first neural network predicts the probability value of blood vessel rupture pressure based on the following: A representation of blood vessels is constructed based on vascular parameters including accumulated energy, impedance, and temperature; Perform an action that includes at least one of applying energy, stopping the application of energy, or changing the application of energy; Receive additional data collected after the action is executed as a response; Based on the response, the probability value of the rupture pressure is predicted; Determine a reward or penalty, where the reward is based on the probability value of rupture pressure and the penalty is based on the resistance of the blood vessel; as well as The first neural network is trained based on rewards or penalties; and Based on the prediction, determine whether the blood vessel is adequately sealed; If it is determined that the blood vessel is not adequately sealed: The output is determined based on the data using a second neural network; The determined output is transmitted to the computing device associated with the energy source for the purpose of developing an energy delivery algorithm; and Using the device, additional electrosurgical energy is delivered from the energy source to the blood vessel according to the energy delivery algorithm to seal the blood vessel.

Images