Gastrointestinal endoscope probe intelligent operation regulation and control method and system

Abstract

The invention provides an intelligent operation regulation and control method and system for a gastrointestinal endoscope probe, and the method comprises the steps: monitoring the multi-modal data of intestinal peristalsis through a sensor group integrated at the tail end of the gastrointestinal endoscope probe; performing deep feature extraction on the intestinal tract movement multi-modal data; an intestinal peristalsis prediction model is trained based on historical peristalsis characteristics, and the change condition of future intestinal peristalsis is predicted; and after the user triggers photographing, calculating a formula score in the sliding time window [t, t + T], and selecting the moment with the highest score to trigger the probe to photograph. Through the method and the corresponding system, the intellectualization and controllability of operation are enhanced, and a detection image with higher quality is obtained.

Description

Technical Field

[0001] The present invention provides a method and system for intelligent operation and control of a gastrointestinal endoscope probe, belonging to the technical field of gastrointestinal endoscope probe control. Background Art

[0002] Existing gastroenteroscopic probe operation methods are usually based solely on the intuitive degree of movement in the endoscopic field of view, or by repositioning the probe to minimize image jitter. This method cannot achieve fine-grained control when facing intestinal peristalsis in different patients and different parts of the body. Due to the lack of effective prediction of the timing of intestinal peristalsis, operators may miss the right time to take pictures, resulting in an inability to guarantee image quality; when the intestines are moving rapidly, if the probe is taken too early or too late, motion artifacts or image blur will occur, affecting the diagnostic effect. Although the probe position can be repeatedly fine-tuned and multiple pictures can be taken during the operation, it is time-consuming and inefficient; and the patient's discomfort during the examination will also increase due to frequent probe movements; since it is impossible to take pictures at the optimal time, doctors may need to repeat the operation multiple times to obtain a clear image. Summary of the Invention

[0003] The present invention provides a method and system for intelligent operation and control of a gastrointestinal endoscope probe to solve the above-mentioned problems:

[0004] The present invention proposes a method for intelligent operation and control of a gastrointestinal endoscopic probe, the method comprising:

[0005] Monitor intestinal motility multimodal data through the sensor group integrated at the end of the gastroenteroscope probe;

[0006] Performing deep feature extraction on the intestinal peristalsis multimodal data;

[0007] Based on historical peristaltic characteristics, the intestinal peristalsis prediction model is trained to predict future changes in intestinal peristalsis;

[0008] After the user triggers the photo, the formula score is calculated within the sliding time window [t, t+T], and the moment with the highest score is selected to trigger the probe to take a photo.

[0009] Furthermore, the sensor group integrated at the end of the gastroenteroscope probe monitors multimodal data of intestinal motility, including:

[0010] The pressure distribution data of the intestinal wall on the gastroenteroscope probe is collected through the micro pressure sensor array integrated at the end of the gastroenteroscope probe;

[0011] The acceleration sensor integrated at the end of the gastrointestinal endoscope probe is used to collect the motion acceleration of the probe in the x, y, and z directions;

[0012] The flexible strain sensor integrated at the end of the gastroenteroscope probe captures the deformation information of the contact between the probe and the intestinal tissue;

[0013] The pressure value of the i-th pressure sensor in the micro pressure sensor array at time t is recorded as p i (t), the acceleration values ​​of the acceleration sensor in the three directions at time t are a x (t), a y (t), a z (t), the output signal of the flexible strain sensor is s(t).

[0014] Furthermore, deep feature extraction is performed on the intestinal motility multimodal data, including:

[0015] Pressure signal sequence Perform short-time Fourier transform (STFT) to obtain the distribution characteristics of the pressure signal in the time-frequency two-dimensional plane.

[0016]

[0017] Where w(t) represents the window function, τ represents the time offset, and f represents the frequency;

[0018] By analyzing the STFT spectrum, the peristaltic frequency characteristics in different frequency bands are extracted, and combined with the time dimension information, the peristaltic frequency f and its change trend over time f(t) are obtained;

[0019] Calculate the kurtosis Kurt(a) and skewness Skew(a) of the acceleration signal sequence a(t):

[0020]

[0021] in, is the mean value of the acceleration signal, σ is the standard deviation, and the peak value difference of the acceleration signal is combined to construct the composite creep amplitude index A:

[0022] A=α×(max(a(t))-min(a(t)))+β×Kurt(a)+γ×Skew(a)

[0023] Among them, α, β, and γ are weight coefficients;

[0024] The output signal s(t) of the flexible strain sensor is subjected to morphological analysis to extract the rise time, fall time, peak value and width of the waveform. Combined with the pressure and acceleration data, a complete intestinal peristalsis feature vector F(t) = [f(t), A(t), s rise (t), s fall (t), s peak (t), s wideth (t)].

[0025] Furthermore, the intestinal motility prediction model is trained based on historical motility characteristics to predict future changes in intestinal motility, including:

[0026] The LSTM network is used to perform time series modeling on the historical peristaltic feature sequence F(t-n+1), F(t-n+2), ..., F(t) to learn the long-term dependency of intestinal peristalsis features.

[0027] A deep Q-network was introduced as a reinforcement learning module, with prediction accuracy and inspection efficiency as reward functions, to dynamically optimize the LSTM prediction results and train the intestinal motility prediction model.

[0028] The intestinal peristalsis prediction model inputs the peristalsis feature vectors of the past n moments and outputs the peristalsis speed v(t+1) and peristalsis intensity level I(t+1) at the next moment.

[0029] Furthermore, after the user triggers a photo, the formula score is calculated within the sliding time window [t, t+T], and the moment with the highest score is selected to trigger the probe to take a photo, including:

[0030] When the user triggers a photo, the probe calculates a score using the advance time calculation model within the sliding time window [t, t+T], and selects the moment with the highest score to trigger the probe to take a photo.

[0031] Based on the predicted intestinal peristalsis velocity v(t+1), the pressure distribution data of the intestinal wall on the gastrointestinal endoscope probe, and the intestinal peristalsis intensity level, a dynamic imaging lead time calculation model is constructed:

[0032]

[0033] Among them, T represents the best time to take a photo, p i (t) represents the pressure value of the i-th sensor in the micro pressure sensor array at time t, M represents the number of sensors in the micro pressure sensor array, and v t+1 It represents the predicted intestinal movement speed at time t+1, I t+1 Indicates the predicted creep intensity level at the next moment, I s represents the preset security strength threshold, ω1, ω2 and ω3 represent weight coefficients, λ and γ represent attenuation factors, and control the sensitivity of each indicator. n Indicates the preset pressure when the intestine is completely relaxed, P m Indicates the pressure when the intestine contracts strongly, v n represents the speed constant, and I represents the security strength constant.

[0034] The present invention provides an intelligent operation and control system for a gastrointestinal endoscope probe, the system comprising:

[0035] A data acquisition module is used to monitor multimodal data of intestinal motility through a sensor group integrated at the end of a gastroenteroscopic probe;

[0036] A feature extraction module, used for performing deep feature extraction on the intestinal peristalsis multimodal data;

[0037] The training prediction model module is used to train the intestinal motility prediction model based on historical motility characteristics to predict future changes in intestinal motility;

[0038] The photo-taking module is triggered at the optimal time. After the user triggers the photo-taking, the formula score is calculated within the sliding time window [t, t+T], and the moment with the highest score is selected to trigger the probe to take a photo.

[0039] Furthermore, the data acquisition module includes:

[0040] A pressure data monitoring module is used to collect pressure distribution data of the intestinal wall on the gastrointestinal endoscopic probe through the micro pressure sensor array integrated at the end of the gastrointestinal endoscopic probe;

[0041] The acceleration data monitoring module is used to collect the acceleration of the probe in the x, y, and z directions through the acceleration sensor integrated at the end of the gastrointestinal endoscope probe;

[0042] The deformation monitoring module is used to capture deformation information of the contact between the probe and intestinal tissue through the flexible strain sensor integrated at the end of the gastroenteroscope probe;

[0043] The storage module is used to record the pressure value of the i-th pressure sensor in the micro pressure sensor array at time t as p i (t), the acceleration values ​​of the acceleration sensor in the three directions at time t are a x (t), a y (t), a z (t), the output signal of the flexible strain sensor is s(t).

[0044] Furthermore, the feature extraction module includes:

[0045] Extract pressure distribution feature module for pressure signal sequence Perform short-time Fourier transform (STFT) to obtain the distribution characteristics of the pressure signal in the time-frequency two-dimensional plane.

[0046]

[0047] Where w(t) represents the window function, τ represents the time offset, and f represents the frequency;

[0048] The module for obtaining change trend data is used to extract the peristaltic frequency characteristics in different frequency bands by analyzing the STFT spectrum, and to obtain the peristaltic frequency f and its change trend f(t) over time by combining the time dimension information;

[0049] Construct a composite creep amplitude indicator module to calculate the kurtosis Kurt(a) and skewness Skew(a) of the acceleration signal sequence a(t):

[0050]

[0051] in, is the mean value of the acceleration signal, σ is the standard deviation, and the peak value difference of the acceleration signal is combined to construct the composite creep amplitude index A:

[0052]

[0053] Among them, α, β, and γ are weight coefficients;

[0054] The intestinal peristalsis feature vector module is constructed to perform morphological analysis on the output signal s(t) of the flexible strain sensor, extract the rise time, fall time, peak value and width of the waveform, and combine the pressure and acceleration data to construct a complete intestinal peristalsis feature vector F(t) = [f(t), A(t), s rise (t), s fall (t), s peak (t), s wideth (t)].

[0055] Furthermore, the training prediction model module includes:

[0056] The time series modeling module is used to perform time series modeling on the historical peristaltic feature sequence F(t-n+1), F(t-n+2), ..., F(t) using an LSTM network to learn the long-term dependencies of intestinal peristalsis features;

[0057] The optimization module introduces a deep Q network as a reinforcement learning module, uses prediction accuracy and inspection efficiency as reward functions, dynamically optimizes the LSTM prediction results, and trains the intestinal motility prediction model;

[0058] The prediction module, the intestinal peristalsis prediction model inputs the peristalsis feature vectors of the past n moments, and outputs the peristalsis speed v(t+1) and peristalsis intensity level I(t+1) at the next moment.

[0059] Furthermore, triggering the photo-taking module at the optimal time includes:

[0060] The scoring calculation module is used for user-triggered photography. The probe calculates the score within the sliding time window [t, t+T] using the advance time calculation model. The moment with the highest score is selected to trigger the probe to take a photo.

[0061] The module for storing the dynamic photo-taking advance time calculation model is used to construct the dynamic photo-taking advance time calculation model based on the predicted intestinal peristalsis speed v(t+1), the pressure distribution data of the intestinal wall on the gastrointestinal endoscope probe, and the intestinal peristalsis intensity level:

[0062]

[0063] Among them, T represents the best time to take a photo, p i (t) represents the pressure value of the i-th sensor in the micro pressure sensor array at time t, M represents the number of sensors in the micro pressure sensor array, and v t+1 It represents the predicted intestinal movement speed at time t+1, I t+1 Indicates the predicted creep intensity level at the next moment, I s represents the preset security strength threshold, ω1, ω2 and ω3 represent weight coefficients, λ and γ represent attenuation factors, and control the sensitivity of each indicator. n Indicates the preset pressure when the intestine is completely relaxed, P m Indicates the pressure when the intestine contracts strongly, v n represents the speed constant, and I represents the security strength constant.

[0064] The beneficial effects of the present invention are: reducing motion artifacts and improving image quality. Taking pictures when intestinal peristalsis is minimal or relatively stable can significantly reduce motion artifacts (motion blur), thereby improving image clarity and recognizability; improving diagnostic accuracy and efficiency. By predicting intestinal peristalsis and selecting intelligent photo-taking time points, doctors can obtain higher-quality image information when reading films or observing in real time, thereby improving the detection rate and diagnostic accuracy of early lesions, while also reducing the operation of repeated photos and improving the efficiency of surgery or examinations; enhancing the intelligence and controllability of operations. This method uses deep learning to analyze multimodal data, and provides more intelligent decision-making assistance functions while ensuring user convenience, thereby reducing dependence on physician experience; reducing patient discomfort and risks, obtaining images that meet diagnostic needs in a shorter time, and reducing the number of probe operations during the examination process, thereby reducing irritation to the intestinal mucosa, improving patient examination comfort and reducing examination risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 This is a schematic diagram of the intelligent operation and control method of a gastrointestinal endoscopic probe described in the present invention. DETAILED DESCRIPTION

[0066] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein may be combined with each other.

[0067] The following description sets forth numerous specific details to facilitate a thorough understanding of the present invention. The embodiments described are merely a portion of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0068] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0069] One embodiment of the present invention provides a method for intelligent operation and control of a gastrointestinal endoscopic probe, the method comprising:

[0070] Monitor intestinal motility multimodal data through the sensor group integrated at the end of the gastroenteroscope probe;

[0071] Performing deep feature extraction on the intestinal peristalsis multimodal data;

[0072] Based on historical peristaltic characteristics, the intestinal peristalsis prediction model is trained to predict future changes in intestinal peristalsis;

[0073] After the user triggers the photo, the formula score is calculated within the sliding time window [t, t+T], and the moment with the highest score is selected to trigger the probe to take a photo.

[0074] The working principles and effects of the above technical solution are as follows: multimodal sensor data acquisition and real-time monitoring: various sensors (such as pressure sensors, accelerometers, and electromyographic / physiological signal sensors) integrated at the end of the gastrointestinal endoscopic probe are used to acquire multimodal data generated by intestinal motility in real time; deep feature extraction: the collected multimodal data is input into deep learning or other complex feature extraction algorithms to analyze and extract intestinal motility features and obtain high-dimensional feature vectors used to characterize intestinal motility behavior; an intestinal motility prediction model based on historical data is trained by combining historical motility data and corresponding deep features. This model can predict changes in intestinal motility over a future period (e.g., a sliding time window [t, t+T]) based on current real-time feature data; and intelligent selection of automatic photo capture time: after the user triggers the photo capture, the system uses the prediction model to calculate the motility status score (e.g., a score based on image clarity, motion jitter, and other indicators) for each possible moment within the sliding time window [t, t+T] and automatically selects the moment with the highest score to trigger the probe to capture the photo, thereby ensuring that the captured image is more stable, clear, and has the greatest diagnostic value. Reduce motion artifacts and improve image quality. Taking pictures when intestinal peristalsis is minimal or relatively stable can significantly reduce motion artifacts (motion blur), thereby improving image clarity and recognizability; improve diagnostic accuracy and efficiency. By predicting intestinal peristalsis and selecting intelligent photo-taking time points, doctors can obtain higher-quality image information when reading films or observing in real time, thereby improving the detection rate and diagnostic accuracy of early lesions, while also reducing repeated photo-taking operations and improving surgical or examination efficiency; enhance the intelligence and controllability of operations. This method uses deep learning to analyze multimodal data, providing more intelligent decision-making support functions while ensuring user convenience, and reducing dependence on physician experience; reduce patient discomfort and risks, obtain images that meet diagnostic needs in a shorter time, and reduce the number of probe operations during the examination process, thereby reducing irritation to the intestinal mucosa, improving patient examination comfort and reducing examination risks.

[0075] In one embodiment of the present invention, a sensor group integrated at the end of a gastroenteroscopic probe is used to monitor multimodal data of intestinal motility, including:

[0076] The pressure distribution data of the intestinal wall on the gastroenteroscope probe is collected through the micro pressure sensor array integrated at the end of the gastroenteroscope probe;

[0077] The acceleration sensor integrated at the end of the gastrointestinal endoscope probe is used to collect the motion acceleration of the probe in the x, y, and z directions;

[0078] The flexible strain sensor integrated at the end of the gastroenteroscope probe captures the deformation information of the contact between the probe and the intestinal tissue;

[0079] The pressure value of the i-th pressure sensor in the micro pressure sensor array at time t is recorded as p i (t), the acceleration values ​​of the acceleration sensor in the three directions at time t are a x (t), a t (t), a z (t), the output signal of the flexible strain sensor is s(t).

[0080] In one embodiment of the present invention, deep feature extraction is performed on the intestinal motility multimodal data, including:

[0081] Pressure signal sequence Perform short-time Fourier transform (STFT) to obtain the distribution characteristics of the pressure signal in the time-frequency two-dimensional plane.

[0082]

[0083] Where w(t) represents the window function, τ represents the time offset, and f represents the frequency;

[0084] By analyzing the STFT spectrum, the peristaltic frequency characteristics in different frequency bands are extracted, and combined with the time dimension information, the peristaltic frequency f and its change trend over time f(t) are obtained;

[0085] Calculate the kurtosis Kurt(a) and skewness Skew(a) of the acceleration signal sequence a(t):

[0086]

[0087] in, is the mean value of the acceleration signal, σ is the standard deviation, and the peak value difference of the acceleration signal is combined to construct the composite creep amplitude index A:

[0088] A=α×(max(a(t))-min(a(t)))+β×Kurt(a)+γ×Skew(a)

[0089] Among them, α, β, and γ are weight coefficients;

[0090] The output signal s(t) of the flexible strain sensor is subjected to morphological analysis to extract the rise time, fall time, peak value and width of the waveform. Combined with the pressure and acceleration data, a complete intestinal peristalsis feature vector F(t) = [f(t), A(t), s rise (t), s fall (t), s peak (t), s wideth (t)].

[0091] In one embodiment of the present invention, a prediction model for intestinal motility is trained based on historical motility characteristics to predict future changes in intestinal motility, including:

[0092] The LSTM network is used to perform time series modeling on the historical peristaltic feature sequences F(t-n+1), F(t-n+2), ..., F(t) to learn the long-term dependencies of intestinal peristalsis features.

[0093] A deep Q-network was introduced as a reinforcement learning module, with prediction accuracy and inspection efficiency as reward functions, to dynamically optimize the LSTM prediction results and train the intestinal motility prediction model.

[0094] The intestinal peristalsis prediction model inputs the peristalsis feature vectors of the past n moments and outputs the peristalsis speed v(t+1) and peristalsis intensity level I(t+1) at the next moment.

[0095] In one embodiment of the present invention, after a user triggers a photo, a formula score is calculated within a sliding time window [t, t+T], and the moment with the highest score is selected to trigger the probe to take a photo, including:

[0096] When the user triggers a photo, the probe calculates a score using the advance time calculation model within the sliding time window [t, t+T], and selects the moment with the highest score to trigger the probe to take a photo.

[0097] Based on the predicted intestinal peristalsis velocity v(t+1), the pressure distribution data of the intestinal wall on the gastrointestinal endoscope probe, and the intestinal peristalsis intensity level, a dynamic imaging lead time calculation model is constructed:

[0098]

[0099] Among them, T represents the best time to take a photo, p i (t) represents the pressure value of the i-th sensor in the micro pressure sensor array at time t, M represents the number of sensors in the micro pressure sensor array, and v t+1 It represents the predicted intestinal movement speed at time t+1, I t+1 Indicates the predicted creep intensity level at the next moment, I s represents the preset security strength threshold, ω1, ω2 and ω3 represent weight coefficients, λ and γ represent attenuation factors, and control the sensitivity of each indicator. n Indicates the preset pressure when the intestine is completely relaxed, P m Indicates the pressure when the intestine contracts strongly, v n represents the speed constant, and I represents the security strength constant.

[0100] One embodiment of the present invention provides an intelligent operation and control system for a gastrointestinal endoscope probe, the system comprising:

[0101] A data acquisition module is used to monitor multimodal data of intestinal motility through a sensor group integrated at the end of a gastroenteroscopic probe;

[0102] A feature extraction module, used for performing deep feature extraction on the intestinal peristalsis multimodal data;

[0103] The training prediction model module is used to train the intestinal motility prediction model based on historical motility characteristics to predict future changes in intestinal motility;

[0104] The photo-taking module is triggered at the optimal time. After the user triggers the photo-taking, the formula score is calculated within the sliding time window [t, t+T], and the moment with the highest score is selected to trigger the probe to take a photo.

[0105] In one embodiment of the present invention, the data acquisition module includes:

[0106] A pressure data monitoring module is used to collect pressure distribution data of the intestinal wall on the gastrointestinal endoscopic probe through the micro pressure sensor array integrated at the end of the gastrointestinal endoscopic probe;

[0107] The acceleration data monitoring module is used to collect the acceleration of the probe in the x, y, and z directions through the acceleration sensor integrated at the end of the gastrointestinal endoscope probe;

[0108] The deformation monitoring module is used to capture deformation information of the contact between the probe and intestinal tissue through the flexible strain sensor integrated at the end of the gastroenteroscope probe;

[0109] The storage module is used to record the pressure value of the i-th pressure sensor in the micro pressure sensor array at time t as p i (t), the acceleration values ​​of the acceleration sensor in the three directions at time t are a x (t), a y (t), a z (t), the output signal of the flexible strain sensor is s(t).

[0110] In one embodiment of the present invention, the feature extraction module includes:

[0111] Extract pressure distribution feature module for pressure signal sequence Perform short-time Fourier transform (STFT) to obtain the distribution characteristics of the pressure signal in the time-frequency two-dimensional plane.

[0112]

[0113] Where w(t) represents the window function, τ represents the time offset, and f represents the frequency;

[0114] The module for obtaining change trend data is used to extract the peristaltic frequency characteristics in different frequency bands by analyzing the STFT spectrum, and to obtain the peristaltic frequency f and its change trend f(t) over time by combining the time dimension information;

[0115] Construct a composite creep amplitude indicator module to calculate the kurtosis Kurt(a) and skewness Skew(a) of the acceleration signal sequence a(t):

[0116]

[0117] in, is the mean value of the acceleration signal, σ is the standard deviation, and the peak value difference of the acceleration signal is combined to construct the composite creep amplitude index A:

[0118] A=α×(max(a(t))-min(a(t)))+β×Kurt(a)+γ×Skew(a)

[0119] Among them, α, β, and γ are weight coefficients;

[0120] The intestinal peristalsis feature vector module is constructed to perform morphological analysis on the output signal s(t) of the flexible strain sensor, extract the rise time, fall time, peak value and width of the waveform, and combine the pressure and acceleration data to construct a complete intestinal peristalsis feature vector F(t) = [f(t), A(t), s rise (t), s fall (t), s peak (t), s wideth (t)].

[0121] In one embodiment of the present invention, the training prediction model module includes:

[0122] Time series modeling module, used to use LSTM network to model the historical creep feature sequence F(t-n+1), F(t-n+

[0123] 2) ,…,F(t) performs time series modeling to learn the long-term dependencies of intestinal motility features;

[0124] The optimization module introduces a deep Q network as a reinforcement learning module, uses prediction accuracy and inspection efficiency as reward functions, dynamically optimizes the LSTM prediction results, and trains the intestinal motility prediction model;

[0125] The prediction module, the intestinal peristalsis prediction model inputs the peristalsis feature vectors of the past n moments, and outputs the peristalsis speed v(t+1) and peristalsis intensity level I(t+1) at the next moment.

[0126] In one embodiment of the present invention, triggering the photo-taking module at the optimal time includes:

[0127] The scoring calculation module is used for user-triggered photography. The probe calculates the score within the sliding time window [t, t+T] using the advance time calculation model. The moment with the highest score is selected to trigger the probe to take a photo.

[0128] The module for storing the dynamic photo-taking advance time calculation model is used to construct the dynamic photo-taking advance time calculation model based on the predicted intestinal peristalsis speed v(t+1), the pressure distribution data of the intestinal wall on the gastrointestinal endoscope probe, and the intestinal peristalsis intensity level:

[0129]

[0130] Among them, T represents the best time to take a photo, p i (t) represents the pressure value of the i-th sensor in the micro pressure sensor array at time t, M represents the number of sensors in the micro pressure sensor array, and v t+1 It represents the predicted intestinal movement speed at time t+1, I t+1 Indicates the predicted creep intensity level at the next moment, I s represents the preset security strength threshold, ω1, ω2 and ω3 represent weight coefficients, λ and γ represent attenuation factors, and control the sensitivity of each indicator. n Indicates the preset pressure when the intestine is completely relaxed, P m Indicates the pressure when the intestine contracts strongly, v n represents the speed constant, and I represents the security strength constant.

[0131] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A method for intelligent operation and control of a gastrointestinal endoscopic probe, characterized in that: The method comprises: Monitor intestinal motility multimodal data through the sensor group integrated at the end of the gastroenteroscope probe; Performing deep feature extraction on the intestinal peristalsis multimodal data; Based on historical peristaltic characteristics, the intestinal peristalsis prediction model is trained to predict future changes in intestinal peristalsis; After the user triggers the photo, the formula score is calculated within the sliding time window [t, t+T], and the moment with the highest score is selected to trigger the probe to take a photo.

2. The method for intelligent operation and control of a gastrointestinal endoscopic probe according to claim 1, characterized in that: The sensor group integrated at the end of the gastroenteroscope probe monitors multimodal intestinal motility data, including: The pressure distribution data of the intestinal wall on the gastroenteroscope probe is collected through the micro pressure sensor array integrated at the end of the gastroenteroscope probe; The acceleration sensor integrated at the end of the gastrointestinal endoscope probe is used to collect the motion acceleration of the probe in the x, y, and z directions; The flexible strain sensor integrated at the end of the gastroenteroscope probe captures the deformation information of the contact between the probe and the intestinal tissue; The pressure value of the i-th pressure sensor in the micro pressure sensor array at time t is recorded as p i (t), the acceleration values ​​of the acceleration sensor in the three directions at time t are a x (t), a y (t), a z (t), the output signal of the flexible strain sensor is s(t).

3. The method for intelligent operation and control of a gastrointestinal endoscope probe according to claim 1, characterized in that: Performing deep feature extraction on the intestinal motility multimodal data includes: Pressure signal sequence Perform short-time Fourier transform (STFT) to obtain the distribution characteristics of the pressure signal in the time-frequency two-dimensional plane. Where w(t) represents the window function, τ represents the time offset, and f represents the frequency; By analyzing the STFT spectrum, the peristaltic frequency characteristics in different frequency bands are extracted, and combined with the time dimension information, the peristaltic frequency f and its change trend over time f(t) are obtained; Calculate the kurtosis Kurt(a) and skewness Skew(a) of the acceleration signal sequence a(t): in, is the mean value of the acceleration signal, σ is the standard deviation, and the peak value difference of the acceleration signal is combined to construct the composite creep amplitude index A: A=α×(max(a(t))-min(a(t)))+β×Kurt(a)+γ×Skew(a) Among them, α, β, and γ are weight coefficients; The output signal s(t) of the flexible strain sensor is subjected to morphological analysis to extract the rise time, fall time, peak value and width of the waveform. Combined with the pressure and acceleration data, a complete intestinal peristalsis feature vector F(t) = [f(t), A(t), s rise (t),s fall (t),s peak (t), s wideth (t)].

4. The method for intelligent operation and control of a gastrointestinal endoscope probe according to claim 1, characterized in that: The intestinal motility prediction model is trained based on historical motility characteristics to predict future changes in intestinal motility, including: The LSTM network is used to perform time series modeling on the historical peristaltic feature sequences F(t-n+1), F(t-n+2), ..., F(t) to learn the long-term dependencies of intestinal peristalsis features. A deep Q-network was introduced as a reinforcement learning module, with prediction accuracy and inspection efficiency as reward functions, to dynamically optimize the LSTM prediction results and train the intestinal motility prediction model. The intestinal peristalsis prediction model inputs the peristalsis feature vectors of the past n moments and outputs the peristalsis speed v(t+1) and peristalsis intensity level I(t+1) at the next moment.

5. The method for intelligent operation and control of a gastrointestinal endoscope probe according to claim 1, characterized in that: After the user triggers a photo, the formula score is calculated within the sliding time window [t, t+T]. The moment with the highest score is selected to trigger the probe to take a photo, including: When the user triggers a photo, the probe calculates a score using the advance time calculation model within the sliding time window [t, t+T], and selects the moment with the highest score to trigger the probe to take a photo. Based on the predicted intestinal peristalsis velocity v(t+1), the pressure distribution data of the intestinal wall on the gastrointestinal endoscope probe, and the intestinal peristalsis intensity level, a dynamic imaging lead time calculation model is constructed: Among them, T represents the best time to take a photo, p i (t) represents the pressure value of the i-th sensor in the micro pressure sensor array at time t, M represents the number of sensors in the micro pressure sensor array, and v t+1 It represents the predicted intestinal movement speed at time t+1, I t+1 Indicates the predicted creep intensity level at the next moment, I s represents the preset security strength threshold, ω1, ω2 and ω3 represent weight coefficients, λ and γ represent attenuation factors, and control the sensitivity of each indicator. n Indicates the preset pressure when the intestine is completely relaxed, P m Indicates the pressure when the intestine contracts strongly, v n represents the speed constant, and I represents the security strength constant.

6. A gastrointestinal endoscope probe intelligent operation and control system, characterized in that: The system comprises: A data acquisition module is used to monitor multimodal data of intestinal motility through a sensor group integrated at the end of a gastroenteroscopic probe; A feature extraction module, used for performing deep feature extraction on the intestinal peristalsis multimodal data; The training prediction model module is used to train the intestinal motility prediction model based on historical motility characteristics to predict future changes in intestinal motility; The photo-taking module is triggered at the optimal time. After the user triggers the photo-taking, the formula score is calculated within the sliding time window [t, t+T], and the moment with the highest score is selected to trigger the probe to take a photo.

7. The intelligent operation and control system for a gastrointestinal endoscope probe according to claim 6, characterized in that: The data acquisition module includes: A pressure data monitoring module is used to collect pressure distribution data of the intestinal wall on the gastrointestinal endoscopic probe through the micro pressure sensor array integrated at the end of the gastrointestinal endoscopic probe; The acceleration data monitoring module is used to collect the acceleration of the probe in the x, y, and z directions through the acceleration sensor integrated at the end of the gastrointestinal endoscope probe; The deformation monitoring module is used to capture the deformation information of the contact between the probe and the intestinal tissue through the flexible strain sensor integrated at the end of the gastroenteroscope probe; The storage module is used to record the pressure value of the i-th pressure sensor in the micro pressure sensor array at time t as p i (t), the acceleration values ​​of the acceleration sensor in the three directions at time t are a x (t), a y (t), a z (t), the output signal of the flexible strain sensor is s(t).

8. The intelligent operation and control system for a gastrointestinal endoscope probe according to claim 6, characterized in that: The feature extraction module includes: Extract pressure distribution feature module for pressure signal sequence Perform short-time Fourier transform (STFT) to obtain the distribution characteristics of the pressure signal in the time-frequency two-dimensional plane. Where w(t) represents the window function, τ represents the time offset, and f represents the frequency; The module for obtaining change trend data is used to extract the peristaltic frequency characteristics in different frequency bands by analyzing the STFT spectrum, and to obtain the peristaltic frequency f and its change trend f(t) over time by combining the time dimension information; Construct a composite creep amplitude indicator module to calculate the kurtosis Kurt(a) and skewness Skew(a) of the acceleration signal sequence a(t): in, is the mean value of the acceleration signal, σ is the standard deviation, and the peak value difference of the acceleration signal is combined to construct the composite creep amplitude index A: A=α×(max(a(t))-min(a(t)))+β×Kurt(a)+γ×Skew(a) Among them, α, β, and γ are weight coefficients; The intestinal peristalsis feature vector module is constructed to perform morphological analysis on the output signal s(t) of the flexible strain sensor, extract the rise time, fall time, peak value and width of the waveform, and combine the pressure and acceleration data to construct a complete intestinal peristalsis feature vector F(t) = [f(t), A(t), s rise (t), s fall (t), s peak (t), s wideth (t)].

9. The intelligent operation and control system for a gastrointestinal endoscope probe according to claim 6, characterized in that: The training prediction model module includes: The time series modeling module is used to perform time series modeling on the historical peristaltic feature sequence F(t-n+1), F(t-n+2), ..., F(t) using an LSTM network to learn the long-term dependencies of intestinal peristalsis features; The optimization module introduces a deep Q network as a reinforcement learning module, uses prediction accuracy and inspection efficiency as reward functions, dynamically optimizes the LSTM prediction results, and trains the intestinal motility prediction model; The prediction module, the intestinal peristalsis prediction model inputs the peristalsis feature vectors of the past n moments, and outputs the peristalsis speed v(t+1) and peristalsis intensity level I(t+1) at the next moment.

10. The intelligent operation and control system for a gastrointestinal endoscope probe according to claim 6, characterized in that: The triggering of the photo taking module at the optimal time includes: The scoring calculation module is used for user-triggered photography. The probe calculates the score within the sliding time window [t, t+T] using the advance time calculation model. The moment with the highest score is selected to trigger the probe to take a photo. The module for storing the dynamic photo-taking advance time calculation model is used to construct the dynamic photo-taking advance time calculation model based on the predicted intestinal peristalsis speed v(t+1), the pressure distribution data of the intestinal wall on the gastrointestinal endoscope probe, and the intestinal peristalsis intensity level: Among them, T represents the best time to take a photo, p i (t) represents the pressure value of the i-th sensor in the micro pressure sensor array at time t, M represents the number of sensors in the micro pressure sensor array, and v t+1 It represents the predicted intestinal movement speed at time t+1, I t+1 Indicates the predicted creep intensity level at the next moment, I s represents the preset security strength threshold, ω1, ω2 and ω3 represent weight coefficients, λ and γ represent attenuation factors, and control the sensitivity of each indicator. n Indicates the preset pressure when the intestine is completely relaxed, P m Indicates the pressure when the intestine contracts strongly, v n represents the speed constant, and I represents the security strength constant.

Images