AI-based generator for surgical instruments

Abstract

Using the generator (10) according to the invention, the treatment of biological tissue can be reliably carried out by means of an electrosurgical instrument, in particular by means of an argon plasma probe, without depending on the personal skills of the treatment person. Multiple test treatments of a tissue sample are performed and a training dataset is created based on the test treatments by means of image-supported measurement generation, e.g., using a camera or a medical imaging device, such as CT. From the machine learning-based training dataset, a control dataset is created that controls the devices (10) located in the operating room during subsequent use without the need for camera observation by means of the operating field. Only typical sensor data patterns that have been assigned to a particular organization effect during a training session monitored by a camera are evaluated.

Description

Technical Field

[0001] The present invention relates to an AI generator for operating surgical instruments, wherein the generator includes a control module based on or including artificial intelligence. Background Technology

[0002] Electrosurgical generators are used to supply current to electrosurgical instruments and, when necessary, to supply additional operating media (fluids such as argon, CO2, NaCl solution, etc.). Therefore, the control of the electrosurgical generator is particularly important because it directly controls the surgical outcome by controlling the introduction of energy into the biological tissue. However, the relationship between variables characterizing the current (such as frequency, voltage, current intensity, power, modulation type, and peak factor) and electrical variables of the load formed by the biological tissue (such as impedance) and the achievable surgical outcome is complex. For this reason, different approaches have been attempted in the past to apply the concepts of artificial intelligence to generator control. Regarding this, US 2023 / 0071343 A1 describes a computer-implemented method that uses electrical variables, the on / off state of the instrument, and image data obtained from a camera for training purposes. Furthermore, during subsequent use of a corresponding training module, this module uses the currently detected electrical variables and the additional image data to estimate whether the current settings will lead to successful application.

[0003] According to US 2020 / 0265309 A1, a machine learning-based tissue parameter estimation algorithm is known. Based on electrical parameters, tissue parameters used to control energy output are determined.

[0004] DE 10 2020 105 835 A1 discloses an apparatus for performing endoscopic surgery, wherein an optical recording device is part of the apparatus, the field of view of which is directed onto the tissue to be treated or the tissue to be treated by means of RF electrodes. Before or during treatment of the tissue, tissue type is classified in the RF electrode region based on optical measurement signals from the recording device, and an RF mode matching the determined tissue type is set based on the results of the optical classification.

[0005] Further prior art is formed by DE 10 2021 101 410 A1, US 2023 / 0420032 A1, EP 4 134029 A1 and EP 3 541 313.

[0006] For electrosurgical treatment of tissues, particularly argon plasma coagulation, considerable experience is required to assess the later-visible results—that is, the effect on the tissue—during the surgical procedure. The plasma acting on the tissue is itself luminescent and can therefore produce a blinding effect. Furthermore, the emitted light, due to its spectral composition, can complicate the purely visual assessment of the obtained results. Additionally, the generation of smoke, vapor, or partial coverage of the treatment site by instruments or tissue components can further complicate the assessment of results during treatment. Summary of the Invention

[0007] Therefore, the subject of this invention is to provide a generator capable of operating an instrument that can achieve desired effects on tissue, such as the desired depth of penetration of the effect in the tissue, which is independent of the visual observation of the therapist and largely independent of the therapist's personal ability.

[0008] This problem is solved using the generator according to claim 1: The generator according to the invention includes a control module connected to sensors for electrical variables. The electrical variables are primarily and preferably derived solely from the current and voltage supplied to the electrical instruments connected to the generator. The electrical variables can be supplied by sensors, such as current sensors, voltage sensors, power factor determination sensors, nonlinear determination sensors connected to the load, etc. However, the control module for generator operation only receives the electrical variables and does not receive image data that may come from imaging devices (such as cameras or CT scanners). Instead, the control module can be "blind" during the evaluation of the effects of surgical applications. Therefore, the control module operates based on an AI module that creates a control dataset during training runs. During training, the aforementioned electrical variables are incorporated into training data that is also monitored later during application on a patient. During training, in addition to electrical variables containing information about the depth of penetration of tissue effects (such as coagulation or ablation) caused by the treatment, image data sets can be used in particular. For example, this penetration depth can be provided in digital form as an effect label, profile, or in a color-coded manner, and can be linked to the detected electrical variables as additional training information. The term "image data" as used in this application includes measurement data from a camera and image-like descriptions obtained from measurements taken by non-imaging methods. For example, point-shaped or two-dimensional spectral measurement data from a diffuse reflectance spectral device can be processed into a parametric two-dimensional map, which forms image data in the sense of this specification.

[0009] The sensors considered in the training dataset need not be the same as those connected to the control module during operation. Preferably, the measurements in the training dataset are provided by other sensors of a similar type that detect the same type of electrical values. Therefore, it is possible to create a training dataset with a specific configuration independent of the sensor technology used during operation.

[0010] Image data can be a single image recorded when the desired level of effect is achieved, and can be stored. Alternatively, image data can include multiple images acquired over time, or can include video sequences. For training, image data from an external camera recording treatment and the treatment process can be used. Alternatively or additionally, image data may not originate from a camera, but from a medical imaging device, such as a computed tomography (CT) scanner, magnetic resonance imaging (MRT) scanner, ultrasound scanner, optical coherence tomography (OCT) scanner, or diffuse reflectance spectroscopy (DRS) scanner. The image data used in this embodiment may include additional information about the depth of penetration.

[0011] Furthermore, tissue effects obtained on biological tissues during training were categorized and assigned corresponding effect labels. In this sense, the effect labels refer to each target variable associated with treatment success, including qualitative parameters such as the degree of coagulation or ablation, and quantitative parameters such as the depth of penetration of the achieved tissue effect. This can be performed manually or automatically.

[0012] In another specific embodiment, image data serves as a source of information regarding the depth of penetration of the tissue effect created by the treatment, such as coagulation or ablation. The depth effect can be provided in the form of numerical values, depth profiles, or color-coded illustrations within two-dimensional or three-dimensional images. Measurement of the penetration depth can be based on changes in tissue properties associated with the coagulation or ablation effect. These changes can be optical, acoustic, electrical, or chemical in nature and can be detected using known medical measurement methods. The penetration depth determined in this way is taken into account when assigning labels and is part of a corresponding effect label linked to the detected electrical variable.

[0013] Tissue effects can be achieved using argon plasma coagulation, during which energy is transferred via ionized argon gas without direct electrode contact. Alternatively or additionally, contact coagulation can be used, during which electrodes are in direct contact with the tissue surface during energy output. Both methods are suitable for producing a penetration depth that can be detected as part of an effect tag and can be combined with controls according to the invention.

[0014] The AI ​​module is configured to determine a control dataset from a training dataset using machine learning. This control dataset indicates which electrical variables or patterns of electrical variables must be present to achieve a tissue effect corresponding to a desired effect label, such as the desired penetration depth of the tissue effect. In a first embodiment, the electrical variable pattern thus includes predefined values ​​or ranges of values ​​for each monitored electrical variable. In an extended embodiment, the electrical variable pattern includes individual electrical variable values ​​obtained over time intervals or time courses for each monitored electrical variable. In another, more detailed embodiment, the pattern includes electrical variables monitored at at least two different power output states of the generator. Different power outputs can be achieved through different voltages, currents, current limits, modulation forms, etc. Different forms of modulation can be, for example, continuous wave (CW), amplitude modulation, or pulse-pause sampling with a constant or variable pulse-pause ratio. These different power outputs can define different operating modes that create different surgical effects. Furthermore, the pattern can include time courses of one or more electrical variables before and after switching between different power outputs or modes of the generator. Utilizing these different methods, different requirements for the accuracy of treatment outcomes can be met.

[0015] The two modes described above are preferably distinguished, at least in terms of the electrical power output to the instrument. While a high power output is possible in the first operating mode, the second operating mode is characterized by lower power. Preferably, the power in the first operating mode is designed to cause inactivation and coagulation of biological tissue and is associated with rapidly achieving the desired tissue effect, preferably with a predefined penetration depth. The power is preferably higher than 10 W. The power in the second operating mode is preferably designed not to produce a surgical effect on the tissue, particularly no inactivation effect. The power can be limited to 10 W or lower.

[0016] Preferably, the control module with the AI ​​module is configured to switch from a first operating mode to a second operating mode when a pattern of the electrical variable is determined according to any of the above embodiments, and to assign a desired effect label to that mode. Optionally, it can be configured to further monitor the pattern of the electrical variable after switching to the second operating mode. This particularly allows for consideration of the nonlinear and time-dependent electrical properties of biological tissues. For example, among other things, tissue impedance may even change during the second operating mode with low power after the effects of high power are utilized. For example, it may decrease due to rewetting of previously dried tissue areas.

[0017] Furthermore, if the pattern of the electrical variable detected in the second operating mode corresponds to an effect label that is too low, such as too low penetration depth and / or too low tanning degree in patient tissue, the control module can be configured to switch back from the second (weak) mode to the first (strong) mode. Conversely, if the pattern detected in the second operating mode corresponds to an effect label that is too high, such as too deep penetration depth and / or too high tanning degree in patient tissue, it can also be configured to switch back from the second (weak) mode to the third mode. This situation is, in itself, unexpected and usually not intentional, because tissue changes caused by electrical effects are largely irreversible. The third mode can be a shutdown operating mode that prevents further acquisition of the inherently desired tissue effect. Tissue effects are changes in tissue, such as coagulation.

[0018] The control module may include distance measurement functionality. It can block data generation, i.e., the electrical variables of the sensors, which are difficult to interpret or may become uninterpretable. If the distance between the instrument and the tissue is too long, signal distortion can occur due to the (too) high influence of the electrical nonlinearity of the arc or plasma provided between the instrument (especially its electrodes) and the tissue, which can affect or prevent reliable signal interpretation. For example, distance determination can be performed in a pattern recognition manner, in which specific patterns based on electrical variables of excessively long discharge distances (plasma) are identified, for example, to request a surgeon to reset the instrument.

[0019] The present invention also relates to a method for training a control module and its AI module for an electrical generator, and to subsequent operations of the generator as described above. The method includes a training process in which a training dataset is generated. The latter is preferably obtained using predefined, unchanging settings for the generator. Using this setting, an instrument supplied by the generator influences tissue, and the resulting electrical variables, particularly their temporal progression and values, and the resulting tissue changes can be detected, which can be recorded by means of a camera in the form of images, image sequences, or video sequences. Additionally or alternatively, the depth of tissue changes can also be detected in the form of measurement data, which may include, for example, image sequences, video sequences, spectral progressions, etc., using imaging equipment such as CT, MRT, ultrasound, OCT, or DRS. The processed tissue samples are examined automatically or manually and classified in terms of the obtained tissue effect and / or the penetration depth of the obtained tissue effect. Accordingly, effect labels are assigned to the images, image sequences, and / or video sequences, and the patterns of the acquired electrical variables. Based on the acquired patterns of electrical variables, images, image sequences, and / or video sequences, along with all assigned effect labels (training dataset), a control dataset is generated that represents only the relationship between electrical variables and effect labels. Therefore, by using the control dataset, without the aid of camera images or other images as results, controlling the electrical effects on biological tissues allows the treatment outcome to correspond to the desired effect label.

[0020] By doing so, a generator is provided that includes possible selections of desired effect intensity in a specific operating mode, just as a conventional generator would. However, the effect intensity is no longer a value on an abstract scale, such as from 1 to 10, assigned to the output power, but rather a preset treatment outcome. For example, the effect intensity can be preset to multiple levels, such as four levels (low coagulation, weak coagulation, strong coagulation, very strong coagulation). If the operator selects weak coagulation, the generator operates in a first operating mode based on a control dataset until weak coagulation is achieved, and then switches to a second operating mode, in which no further coagulation is achieved. If the operator selects strong coagulation, the generator also operates in the first operating mode initially until strong coagulation is indeed achieved, at which point it switches to the second operating mode. In this way, the treatment outcome is largely independent of the operator's personal skill. The operator can always achieve the desired surgical outcome on the tissue in front of him / her, and it is truly independent of subjective optical assessments of tissue changes during treatment.

[0021] In one specific embodiment, the present invention includes a method for coagulating and / or ablating patient tissue, particularly mucosal tissue. This method is particularly useful for large-area surface treatments because it is advantageous for conditions such as obesity therapy, Barrett's esophagus, endometriosis, or similar diseases where a tissue effect over the largest possible area is desired. For example, this can be used to perform ablation and / or coagulation of the gastric mucosa. For example, the method includes the following steps: - Inserting an endoscopic-guided electrosurgical instrument into the tissue area of ​​the patient's body to be treated, such as the patient's stomach; - Select the desired tissue effect, including at least the desired depth of penetration of the tissue effect and / or the desired degree of tissue inactivation in the area of ​​tissue to be treated; - Detect electrical variables during instrumental treatment and determine the current tissue effect, particularly the current penetration depth and / or the current degree of tissue inactivation, based on these electrical variables and a control dataset that has been pre-generated using machine learning, wherein the control dataset has been obtained from a training dataset that includes electrical variables and target variables assigned to it; - The instrument's power output is controlled according to the specific tissue effect to achieve the desired tissue effect, and then the instrument's power output is automatically reduced or terminated to avoid over-powering.

[0022] Provided it is technically useful and effective, the indicated method steps may also be performed in a different order or at least in partial parallel.

[0023] The methods described above, in particular, can minimize damage to the depth of the treated tissue, allowing for the treatment of, for example, only superficial tissue layers, such as the superficial mucosa. Depending on the treatment, the superficial tissue layer may contain Ghrelin cells, which are crucial for obesity treatment, or it may contain lesions of endometriosis.

[0024] The desired tissue penetration depth can be set before tissue ablation and / or coagulation. In particular, this method does not require additional surgical treatments, such as the introduction of materials between or within the mucosa to protect the submucosa.

[0025] In a preferred embodiment, the penetration depth corresponds, for example, to approximately two-thirds, preferably half or a quarter, of the submucosal thickness. Additionally or alternatively, scarring can be created in the patient's stomach wall, which reduces the elasticity of the stomach wall and thus reduces the stomach's storage capacity, allowing the desired degree of scarring to be set and achieved. Using a control dataset trained on mucosal tissue allows for particularly precise setting of energy output for specific characteristics of the gastric mucosa. Attached Figure Description

[0026] Further details and advantageous embodiments of the invention will become apparent from the following description and accompanying drawings. In the drawings: Figure 1 The schematic diagram illustrates a generator according to the invention, instruments connected to the generator, and a biological object during a surgical procedure. Figure 2 The diagram illustrates an example of a generator used to produce a training dataset during the influence of biological objects using a camera and instruments connected to the generator. Figure 2a Another example of a generator is shown in the diagram, which is used to generate a training dataset using a camera and instruments connected to the generator during the influence of biological objects. Figure 2b Another example of a generator is shown in the diagram, which is used to generate a training dataset using a CT scanner and instruments connected to the generator during the influence of a biological object. Figure 3 The diagram illustrates the process of obtaining and refining from the training dataset. Figure 3a The diagram illustrates the use of a CT scanner instead of a camera to acquire and extract training datasets. Figure 4 The different modes of operation of electrical variables acquired during the treatment of biological tissues are shown in line graph form. Figure 5 A line diagram is shown to illustrate the switching of the generator between different operating modes. Detailed Implementation

[0027] exist Figure 1 The diagram illustrates a generator 10 according to the invention, to which an instrument 11, particularly an argon plasma instrument, is connected. The instrument 11 is connected to a generator output section 13 via a cable 12, to which a neutral electrode 14 is also connected for guiding the current output back to the instrument 11. This configuration is used in monopolar instruments. In the case of bipolar instruments, both poles of the output from the generator output section 13 are connected to the instrument 11. According to one example, the control module 19 can control the degree of effect that can be achieved on tissue, meaning, for example, that the degree of browning or inactivation of the treated tissue and / or the preset penetration depth of the tissue effect can be controlled. Therefore, this control is based on a control dataset created using training data. For example, the training dataset can be generated on a defined tissue type, such as mucosal tissue.

[0028] If instrument 11 is specifically an argon plasma instrument, it is additionally supplied with gas, particularly argon, from generator 10 or other supply device; however, in Figure 1For clarity, the gas is not depicted in the text. Therefore, parameters of the gas flow may also be included in the training dataset and thus in the control dataset.

[0029] Instrument 11 includes electrode 15, which can be arranged in the gas delivery channel, particularly the argon delivery channel, and in... Figure 1 The image is depicted in dashed lines. A plasma discharge 16 is initiated at electrode 15 and supplied by generator 10, within which current flows from electrode 15 to biological tissue 17 connected to neutral electrode 14. This current exerts a thermal effect on the biological tissue.

[0030] For the supply of instrument 11, this means that in order to provide electrical power at output 13, generator 10 includes source 18, particularly an RF source for high-frequency voltage that can be controlled by control module 19. In particular, control module 19 can therefore control selected electrical variables, such as voltage modulation, magnitude (amplitude), current intensity, and the like.

[0031] For example, source 18 is formed by an oscillator that oscillates at a high frequency and is configured to output several 1000V signals at output section 13. peak The voltage (peak-to-peak voltage) and a power of several watts, preferably >10 W, for example 100 W. Electrical characteristic variables, such as voltage or current, the phase angle between voltage and current, and the non-harmonic distortion of the current, are detected. Sensor block 20 operates to provide the measured variables to control module 19, as indicated by arrow 21. Arrow 21 marks the direction of information flow and therefore includes only one arrow tip. However, it is also possible that this embodiment is configured such that control module 19 specifically requests sensor data and transmits a data query to sensor block 20 for this purpose.

[0032] Control module 19 controls oscillator 18, as indicated by arrow 22. Furthermore, control module 19 can receive information from oscillator 18 that is independent of the surgical outcome achieved on tissue 17. For example, such information could be about the oscillation frequency of oscillator 18 or its modulation type (e.g., continuous wave (CW) or pulsed, such as on / off sampling).

[0033] Control module 19 includes a control dataset created from a training dataset using machine learning. This control dataset is configured to establish a relationship between effect intensity or level, including tissue effect, and desired penetration depth, and electrical variables provided by oscillator 18 and / or sensor block 20. An input device 24 is used to preset the corresponding effect intensity or tissue effect, including penetration depth. This input device is either part of generator 10 or configured to be detached from generator 10, for example, via a mobile device such as a tablet or mobile phone. Specifically, input device 24 may also be part of instrument 11.

[0034] Clearly, the control of generator 10 is based solely on the desired penetration depth and effect intensity preset via input device 24, and on electrical variables from oscillator 18 and / or sensor block 20. A camera for checking the effect achieved on biological tissue 17 is neither provided nor necessary. Unlike conventional devices, the preset effect intensity or desired penetration depth via input device characterizes the actual effect to be achieved on biological tissue, characterized, for example, by the degree of browning of the tissue or a specific depth of penetration into the tissue. Once the desired effect is achieved, the generator switches to a second operating mode in which no further tissue influence is performed. This is done without requiring camera images of the treated tissue. For this reason, even if the plasma jet is unnecessarily directed onto the tissue area for an extended period, the desired effect is achieved but not exceeded.

[0035] Figure 2The generation of training dataset 27 is explained, upon which a control dataset is subsequently created by AI module 23. A separate training module 19' is provided for this purpose. In addition to the variables generated by sensor block 20, image data v from camera 25 is supplied to training module 19', where the camera's field of view specifically includes the location of tissue 17 affected by discharge 16. Training module 19' is used to generate training dataset 27. Furthermore, a training input device 24' is provided in training module 19', through which the person responsible for generating training dataset 27 can input the desired effect. For example, if multiple browning levels of tissue are distinguished, such as 10 browning levels, they can have a range from no browning to the onset of carbonization. During training, this browning degree is assigned to patterns of electrical variables generated using sensor block 20 or received directly from oscillator 18 as effect labels in training dataset 27. Furthermore, different tissue types can be set using training input device 24'. Training module 19' additionally includes block 26', which is configured to create and / or store control signals for oscillator 18 and power output from generator 10 to its associated instrument 11. The control signals can be assigned to different operating modes used during the generation of training dataset 27. In a first operating mode, oscillator 18 outputs high power to instrument 11, producing a surgical effect on tissue 17, such as visible coagulation. In a second, weaker operating mode, the power output from oscillator 18 to instrument 11 is reduced to a limit where discharge 16 no longer achieves a visible effect on the surface of tissue 17.

[0036] exist Figure 2 In this device, generator 10 includes input device 24 and block 26. Through input device 24, the person responsible for generating training dataset 27 can input the effect achieved by the discharge 16 affecting tissue 17 for testing purposes. Block 26 is configured to control oscillator 18, and thus control the power output of generator 10 to instrument 11.

[0037] Figure 2a This illustrates another example of generator 10′ used to generate the training dataset. For Figure 2a The example shown in the figure relates to the reference numerals. Figure 2 The relevant interpretations provided apply accordingly. Figure 2a The examples shown in the text are similar to Figure 2 The main difference in the example shown is that the training module 19' is not provided separately from the generator 10'.

[0038] Figure 2b To illustrate another example, a computed tomography (CT) device 25′ is used as the imaging device instead of camera 25. (Regarding the reference numerals...) Figure 2 and Figure 2aThe relevant explanations also apply to... Figure 2b The example shown is illustrated. The CT device 25' determines image data v' during or after tissue influence as measurement data, from which penetration depth is further derived. Penetration depth can include continuous values ​​or different discrete penetration levels, such as ten levels. The determined depth values ​​are linked in the form of effect labels to electrical variables detected by sensors S1 to S3 and incorporated into training dataset 27. In the illustrated example, training module 19' is as follows... Figure 2a It is integrated into generator 10 as in the example, however, it can also be like... Figure 2 These are configured as separate units, as in the example. Based on this training data, the control module 19 can then be controlled so that the predefined effect depth d is precisely reached but not exceeded.

[0039] Figure 3 The generation of the training dataset is described, which includes video sequences or single images from camera 25 and data from sensor block 20. This data can be the outputs g1, g2, g3 of one or more sensors, such as sensors S1, S2, S3, which characterize, for example, the amount of current, the amount of phase angle between current and voltage, the peak factor, and / or the nonlinearity of the current. Additional variables may also be determined. The above variables are merely examples. In the first embodiment, the variables detected by sensors S1 to S3 can only be detected at the end of the effect. These variables thus form a static pattern without time parameters. However, it is also possible to detect the time progression. This applies to both camera 25 and sensor block 20, allowing the time progression of the corresponding variables to be detected. Figure 3 The upper part thus explains that data from sensor block 20, camera 25, and input device 24' are combined to generate training dataset 27. Although the data from camera 25 and sensor block 20 characterize the current state of optical and electrical variables, the effect labels L1, ..., L10 input via input device 24' characterize the achieved results.

[0040] To train the generation of dataset 27, multiple sample treatment processes were performed, during which time, based on... Figure 2 The equipment performs electrical effects on suitable samples. Machine learning is used to determine from this... Figure 3The control dataset 26 is described below. It includes the desired effect (e.g., for moderate browning) as an input variable via input device 24. From the information in the training dataset, it readily has the corresponding values ​​that sensors S1 to S3 need to have to achieve the desired effect. The training dataset 27 may include effect labels characterizing visible tissue changes, as well as quantitative parameters, such as the depth of penetration of the tissue effect achieved. The control dataset 26 now monitors sensors S1 to S4 in terms of achieving predefined values. Therefore, the control module 19 operates as follows: The oscillator 18 is capable of operating at high power in at least a first operating mode to achieve a surgical effect on tissue 17, and at a second, lower power to achieve the same effect. When utilizing... Figure 2 During the generation of the training dataset by the device, the oscillator 18 first operates in a first working mode and then is turned off or switched to a second working mode, after which effect labels are defined based on the achieved treatment outcomes. Effect labels may include qualitative features, such as the degree of visible tissue changes (e.g., browning), and / or quantitative features, such as the depth of tissue effect penetration achieved. When using... Figure 1 During actual patient treatment with the device, control module 19 now operates in a first working mode using a control dataset at the start of treatment. The actual, however, transformations (e.g., browning) achieved by tissue 17 that are not readily identifiable to the surgeon are... Figure 4 The above is illustrated by curve 29. Clearly, browning increases over time until it reaches its maximum. The corresponding variables of sensors S1, S2, S3, and additional sensors can have different time progressions. For example, the value g1 of sensor S1 can represent a current capable of a decreasing process. For example, the value g2 of sensor S2 can be, for example, a humidity value, where tissue humidity can decrease over time and with the progress of treatment. The third value g3 of sensor S3 can be any other electrical variable or a variable characterizing tissue 17 calculated from electrical variables.

[0041] Figure 3a Another training scenario is illustrated, in which image recording is performed using a CT device 25'. The CT device 25' provides image data from which the penetration depth d of the achieved tissue effect is determined. The depth indication, as part of the effect label, is assigned to electrical variables g1 to g3 simultaneously detected by sensors S1 to S3 and recorded in the training dataset 27. This embodiment can be combined with training module 19' integrated in generator 10 and training module embodiments separated from generator 10.

[0042] exist Figure 3 In this embodiment, the therapist has pre-defined the effect, which signifies the degree of browning. Figure 4 The tolerance range is defined in box 30. Figure 3a In this embodiment, the therapist has preset a corresponding effect depth that signifies the (maximum) penetration depth of the tissue effect. In the control dataset 26, the values ​​characterized by box 30 are assigned to the characteristic process and therefore to the electrical variable patterns monitored by sensors S1, S2, and S3, which in Figure 4 This is explained in the text. The mode can be a point t in time for sensors S1, S2, and S3. m The combination of variables g1, g2, and g3 at (time measurement points). However, this model can also include time-process segments or the entire time process of variables from the three sensors S1, S2, and S3. Figure 5 The description states that once the mode is detected, the control module 19 will reduce the energy output to the instrument 11 from a high value (HIGH) to a low value (LOW).

[0043] Using the generator 10 according to the invention, treatment of biological tissues can be performed reliably, without relying on the personal skills of the therapist, using electrosurgical instruments, particularly argon plasma probes. Multiple test treatments of tissue samples are performed using camera-supported measurements, and a training dataset is created based on these test treatments. A control dataset is generated from the training dataset using machine learning, wherein the control dataset controls the device 10 in the operating room in subsequent uses without the aid of camera observation in the operating room. Only typical sensor data patterns assigned to specific tissue effects during camera-monitored training sessions are evaluated. This allows for the determination of achieving preset values ​​based on the degree of effect and / or penetration depth d.

[0044] Furthermore, thanks to this invention, it is possible to generate a thermal effect on tissue very quickly (as fast as possible) and in a controlled manner (without excessive heat). To this end, the control module 19 operates in two states (working modes) during the application phase: a high-power state and a low-power state. The high-power state (first working mode) is used to achieve the desired thermal effect on the tissue. The low-power state (second working mode) is not intended to achieve a thermal effect. When the plasma—which is used to transfer power to the tissue without contact—is ignited and current flows in the first working mode, electrical data (here: peak voltage Up, peak current I) used as input data for the control module 19 are detected. p Effective (RMS) voltage U r ms Effective (root mean square) current I r ms(Power factor, frequency, resistance, spark generation). Based on the predictions of control module 19, the oscillator automatically switches between a first operating mode and a second operating mode to immediately limit tissue damage and achieve the desired level of inactivation. If the tissue damage predicted by control module 19 is below a preset threshold (desired effect), system 18 remains in the first operating mode. If the AI ​​prediction exceeds the predefined effect, system 18 switches to the second operating mode to prevent additional thermal damage. The second operating mode is characterized by an output power that is low enough to avoid causing additional tissue damage (here: 10 W), yet high enough to allow for effective AI prediction of the electrical data achieved in this mode. Therefore, the output power is so high that a stable plasma can be maintained in the second operating mode to select the electrical data used as the basis for deciding whether to switch back to the first operating mode.

[0045] To further reduce energy introduction and the resulting tissue damage in the second operating mode, the voltage can be pulsed. In this case, pulse means that the voltage (and current) switches back and forth between on and off periods in the second operating mode. During the on period, a defined low power is applied. During the off period, no power is output to the tissue. By doing so, the total energy introduction during the second operating mode is greatly reduced, which further reduces tissue inactivation in this state. The sum of the on and off periods can be set to 10 ms; the on period is set from 2.4 to 10 ms. The remaining time is the off period in which no current is output. If 10 ms is selected for the on period, the off period is 0 ms, and no power reduction is performed, so that the low power state is always in the on period. The control between the high and low power states (first and second operating modes) should avoid (current) excess and should achieve a regenerative and therefore uniform tissue effect. Once the preset tissue effect is achieved, the control module 19 switches to a state in which less power is available. In this low power state, no significant tissue effect is produced.

[0046] The pure delay from the point in time when the measured electrical data is received, through the prediction of the control module 19, to the possible switching of the state is approximately 15 ms. An additional delay in the millisecond range is added, for example, due to the transient response of the RF source 18. Control between the high and low power states should avoid (current) excess and should achieve a regenerative and therefore uniform organization effect. Once the preset organization effect is achieved, the control module 19 switches to the state where less power is available. In this low-power state, no significant organization effect is generated.

[0047] To train the AI, electrical data is characterized as the degree of inactivation of the tissue surface. For this purpose, images of the treated area are recorded and grouped (labeled) based on surface coloration. The corresponding electrical data (i.e., the patterns formed by the latter) are used to train different groups or labels (which can then be selected as effect levels on the user interface). The color of the treated tissue is chosen because this is information available to the physician during surgical applications. Due to the equipment used (a system consisting of endoscopes, video processors, monitors, etc.), the color of the tissue may vary systematically and is not comparable over a specific treatment duration. However, defined settings are used during image recording to ensure that the color of the images is not tampered with by the equipment used (e.g., cameras) and to ensure that the results are comparable.

[0048] For the use of control module 19, no images are required, but only electrical data is used for the prediction of tissue inactivation.

[0049] AI-controlled argon plasma coagulation is merely an example of implementing AI control functionality within a module where a previously trained (unalterable) AI algorithm controls power output. The training of the tag-based control module 19—which compares visual data of real tissue with corresponding electrical data—can be used for different operating modes / indications, such as shrinkage (visual) for thermal fusion quality, degree of inactivation during electrosurgical cutting of tissue (visual coagulation zone, collateral damage), penetration depth (e.g., during ESD), expansion of the RF ablation zone (visual ablation zone), and electrical data during electrosurgical cutting.

[0050] Furthermore, this method is not limited to electrical data; it can be extended to additional data, such as for the control of cushion formation in hydraulic technology, temperature regulation during RFA, or ice ball formation in cryogenic technology.

[0051] After switching to the second operating mode with low power, the electrical variables of sensors S1, S2, and S3 can have different values ​​and can again change over time, such as... Figure 4 This is illustrated by the branching of the dashed curve. For example, the current of sensor S1 may initially decrease due to the reduction in power, but may then slightly increase again due to the rewetting of the tissue over time. Similarly, the variables monitored by other sensors S2 and S3 may change again over time. The patterns created in this way can also be included in the training dataset and thus used by control module 19 to verify the achieved effect according to the desired effect label.

[0052] Using the generator 10 according to the invention, treatment of biological tissues can be reliably performed with the aid of electrosurgical instruments, particularly with the aid of argon plasma probes, without depending on the individual skill of the treating personnel. In the case of image-supported measurement generation, such as using a camera or medical imaging device, such as a CT scanner, multiple test treatments of tissue samples are performed, and a training dataset is created based on said test treatments. A control dataset 10 is created from the training dataset based on machine learning to control the device located in the operating room during subsequent use, without the aid of camera observation in the operating room. Only typical sensor data patterns that have been assigned to specific tissue effects during the training sessions monitored by the camera are evaluated.

[0053] Figure label: 10, 10′ generator, 11 Instruments 12 lines 13 Output section of generator 10 14 Neutral electrode 15 electrodes 16 Discharge 17 Biological tissues 18 Sources 19. Control module with AI 20 sensor blocks Arrows 21 and 22 23 AI Module 24 Input devices 24′ Input device 25 cameras 25′ Computed Tomography Equipment 26 pieces 27 Training Dataset 29 curves 30 Figure 4 The tolerance range box in S1-S3 sensors g1-g3 electrical variables Input section of e1–e3 generator L1, ..., L10 effect labels V Image Data v' Image data from CT equipment M electrical variables g1–g3 mode t time t m Time measurement points

Claims

1. A generator, the generator having a control module (19) with inputs (el, e2, e3) to which only sensors (SI, S2, S3) for electrical variables (gl, g2, g3) are connected, the generator having a power source (18) connected to the control module (19) and controllable by the latter, and the power source being connected to a medical instrument (11) in order to supply the latter with electrical power, wherein the control module (19) comprises a control data set (26) based on a training data set (27) comprising image data (v) and electrical variables (gl, g2, g3) detected by means of sensors (SI, S2, S3).

2. The generator of claim 1, wherein, the control module (19) being configured to operate the source (18) in a first operating mode (HIGH) or in a second operating mode (LOW), wherein the source (18) provides high power in the first operating mode (HIGH) and low power in the second operating mode (LOW).

3. The generator of claim 2, wherein, In the first operating mode (HIGH), the electrical power of the generator (10) is designed for achieving the inactivation and coagulation of biological tissue (17), and in the second operating mode (LOW), the electrical power of the generator (10) is designed for avoiding the inactivation and coagulation of biological tissue (17), but is high enough to ensure stable plasma ignition and effective data determination.

4. Generator according to any of the preceding claims, characterized in that The control data set (26) is created by machine learning based on the training data set (27) with image data (v).

5. Generator according to any of the preceding claims, characterized in that, The image data (v) is a single image, an image sequence or video data, and the electrical variables (gl, g2, g3) are a plurality of measurement points determined as individual measurement values at time intervals, or a time course of electrical variables (gl, g2, g3).

6. Generator according to claim 4 or 5, characterized in that The control data set (26) comprises effect labels (LI,..., L10) obtained from a manual evaluation of the results of tissue test treatments.

7. Generator according to claim 5 or 6, characterized in that The control data set (26) comprises effect labels (LI,..., L10) characterizing different degrees of tissue inactivation and / or different penetration depths of tissue effects.

8. The generator of claim 7, wherein, The control module (19) is configured to switch from the first operating mode (HIGH) to the second operating mode (LOW) upon recognition of a pattern of electrical variables (gl, g2, g3) assigned to a desired effect label (LI,..., L10).

9. The generator of claim 8, wherein, The control module is configured to continue monitoring the pattern (M) of electrical variables (gl, g2, g3) after switching to the second operating mode (LOW).

10. The generator of claim 9, wherein, The control module (19) is configured to switch from the second operating mode (LOW) to the first operating mode (HIGH) if the pattern (M) recorded in the second operating mode (LOW) corresponds to an excessively low effect label (LI,..., L10).

11. Generator according to claim 9 or 10, characterized in that The control module (19) is configured to switch from the second operating mode (LOW) to a third mode (OFF) if the pattern recorded in the second operating mode (LOW) corresponds to an excessively high effect label (L1,..., L10).

12. A generator according to any one of the preceding claims, characterised in that, The control module (19) comprises a distance measurement function.

13. A generator according to any one of the preceding claims, characterized in that The control module (19) is configured to determine the distance between the instrument (11) and the biological object (17) on the basis of electrical variables (g1, g2, g3) detected by sensors (S1, S2, S3).

14. The generator of claim 13, wherein, The control module (19) is configured to determine the distance on the basis of a non-linearity of a load acting at the output (13) of the generator (10).

15. A method for generating a control data set (26) of a control module (19) of an electrosurgical generator (10) and a method for the subsequent operation of such a generator (10), wherein: During a training procedure, a training data set (27) is generated in such a way that the generator (10') is used to influence biological tissue (17) in predefined settings and thus to determine provided or resulting electrical variables (g1, g2, g3), and that resulting tissue changes are recorded by means of an imaging device (25, 25'), that effect labels (L1,..., L10) are assigned to the tissue changes, that a control data set (26) is determined from the training data set (27) using machine learning, wherein the control data set (26) represents a relationship between electrical variables (g1, g2, g3) and effect labels (L1,..., L10), that during an application procedure, the generator is controlled on the basis of the control data set (26) for achieving a desired effect corresponding to a selected effect label (L1,..., L10). ​

Images