An intrauterine tissue deformation analysis and negative pressure suction force self-adaptive regulation and control device based on an optical flow method

Abstract

The application discloses an intrauterine tissue deformation analysis and negative pressure suction force self-adaptive regulation and control device based on an optical flow method, comprising a negative pressure suction device and an image acquisition device; by integrating a miniature camera in a sampling suction tube, real-time acquisition of intrauterine pathological tissue images is realized, and the image signals are transmitted to a processor; the opening degree of a regulating valve is automatically controlled by the processor according to the image analysis result, so that the negative pressure suction force output by a vacuum pump is dynamically adjusted; compared with the manual adjustment of the negative pressure after the image is observed by the naked eyes in the prior art, the automatic closed-loop control of the negative pressure suction force is realized, the reaction lag and subjective difference caused by the manual operation of the negative pressure control in the prior art are solved, the negative pressure is automatically reduced when the processor detects that the tissue deformation is too large, the excessive stretching or breaking of the tissue can be avoided, the integrity of the pathological sample is ensured, and the pathological diagnosis accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of medical device control technology, and in particular to a device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method. Background Technology

[0002] Intrauterine lesion tissue sampling is a common procedure in gynecological clinical diagnosis and treatment, widely used for pathological examination of diseases such as endometrial polyps, submucosal fibroids, endometrial hyperplasia, and endometrial cancer. Currently, intrauterine lesion tissue sampling mainly employs the principle of negative pressure suction, where a sampling suction tube is inserted into the uterine cavity, and negative pressure is used to obtain lesion tissue samples.

[0003] Existing negative pressure aspiration sampling devices typically include a vacuum pump, regulating valve, collection container, and sampling aspiration tube. During operation, the doctor manually sets the negative pressure value or adjusts it to an experienced value before initiating aspiration. However, due to the diverse types of lesions within the uterine cavity (e.g., polyps are tough, fibroids are hard, and endometrial tissue is loose and soft), different tissues exhibit significant differences in their tolerance to negative pressure, and the deformation state of the same tissue dynamically changes over time during aspiration. Existing devices have the following technical problems:

[0004] The negative pressure control of existing devices relies entirely on preset settings by doctors or manual adjustments, and cannot sense the real-time deformation of tissue under negative pressure. When the negative pressure is too high, soft endometrial tissue or polyp tissue may be overstretched or even broken, causing the obtained pathological samples to lose their complete cellular structure, seriously affecting the accuracy of pathological diagnosis; when the negative pressure is too low, it cannot effectively adsorb lesion tissue, resulting in insufficient sample volume or sampling failure, requiring repeated operations, increasing patient suffering and operation time.

[0005] Although some devices have integrated miniature cameras into the suction tube (such as the uterine cavity observation suction surgery system) to achieve visual observation during the sampling process, the image information is only used for doctors' naked eye observation and has not established any signal connection or feedback mechanism with the negative pressure control system. Doctors still have to manually adjust the negative pressure while observing the image, which is slow to react and highly dependent on personal experience, making it difficult to make precise adjustments at the moment when tissue deformation occurs.

[0006] The mechanical properties of different diseased tissues vary significantly. The same negative pressure value may be too high for polyps and insufficient for fibroids. Existing devices cannot automatically identify tissue types and adjust the negative pressure strategy accordingly. Doctors can only rely on visual experience to make subjective judgments, resulting in low standardization of operation and a long learning curve. Summary of the Invention

[0007] In order to solve the above-mentioned technical problems, the present invention provides a device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method.

[0008] The technical solution of this invention is implemented as follows:

[0009] A device for intrauterine tissue deformation analysis and adaptive control of negative pressure suction based on optical flow method includes a negative pressure suction device and an image acquisition device:

[0010] The negative pressure suction device includes a sampling suction tube, a vacuum pump, a regulating valve, and a collection container. The front end of the sampling suction tube is provided with a suction port, and the rear end of the sampling suction tube is provided with a negative pressure connector.

[0011] The vacuum pump includes a negative pressure output port;

[0012] The regulating valve includes an air inlet, an air outlet, and a control terminal, wherein the air inlet is connected to the negative pressure output port of the vacuum pump;

[0013] A collection container has an inlet and an outlet, the inlet being connected to the air outlet of the regulating valve, and the outlet being connected to the negative pressure connector of the sampling suction tube. The collection container is connected in series in the negative pressure passage between the regulating valve and the sampling suction tube.

[0014] The negative pressure suction device also includes a processor for controlling the output negative pressure of the vacuum pump. The processor includes an image input terminal and a negative pressure control signal output terminal. The image input terminal is connected to the image output terminal of the image acquisition device, and the negative pressure control signal output terminal is communicatively connected to the control terminal of the regulating valve.

[0015] The image acquisition device includes a miniature camera integrated into the sampling suction tube and a light source assembly disposed on one side of the miniature camera. The image acquisition device also includes a signal cable, which includes a first end and a second end, the first end of which is communicatively connected to the signal output end of the miniature camera.

[0016] Furthermore, the processor integrates:

[0017] The optical flow calculation module has its input end connected to the image input end and its output end outputting optical flow field data.

[0018] The deformation analysis module has its input end connected to the output end of the optical flow calculation module, and its output end outputs tissue deformation parameters.

[0019] The control module has its input terminal connected to the output terminal of the deformation analysis module, and its output terminal connected to the negative pressure control signal output terminal of the processor.

[0020] Furthermore, the sampling suction tube is provided with an axially extending baffle inside the tube, which divides the inner cavity of the tube into mutually isolated image acquisition channels and negative pressure suction channels;

[0021] The miniature camera and the light source are sealed at the front end of the image acquisition cavity. The signal cable passes through the image acquisition cavity, and the second end of the signal cable extends to the rear end of the sampling suction tube and constitutes the image output end of the hysteroscopy.

[0022] The front opening of the negative pressure suction channel forms the suction port, and the negative pressure connector is located at the rear end of the sampling suction tube and communicates with the negative pressure suction channel.

[0023] Furthermore, the negative pressure suction device also includes:

[0024] A pressure sensor is installed on the pipeline between the outlet of the regulating valve and the inlet of the collection container, and the signal output terminal of the pressure sensor is connected to the pressure signal input terminal of the processor.

[0025] A filter is disposed between the outlet of the collection container and the negative pressure connector of the sampling suction tube;

[0026] An overflow prevention device is provided inside the collection container, and the output end of the overflow prevention device is connected to the closing control end of the regulating valve.

[0027] Furthermore, the control module integrates a safety threshold storage unit, which is connected to the comparator of the control module. The comparator compares the tissue deformation parameters output by the deformation analysis module with the preset safety threshold in the safety threshold storage unit, and generates a boost signal, a depressurization signal, or a maintenance signal based on the comparison result. These signals are then output to the control terminal of the regulating valve through the negative pressure control signal output terminal.

[0028] Furthermore, the processor also integrates a motion compensation module, the input of which is connected to the image input, and the output of which is connected to the compensation input of the optical flow calculation module.

[0029] Furthermore, the processor also integrates a tissue type recognition module, the input of which is connected to the image input, and the output of which is connected to the second input of the control module.

[0030] Furthermore, the deformation parameters output by the deformation analysis module include the microstructure strain rate. and deformation gradient, where For optical flow velocity field, These are spatial coordinates.

[0031] Furthermore, the optical flow calculation module is any one of the Lucas-Kanade optical flow calculation unit, the Horn-Schunck optical flow calculation unit, or an optical flow estimation unit based on a deep neural network.

[0032] Furthermore, the second end of the signal cable of the image acquisition device is connected to the image input end of the processor via an HDMI cable, SDI cable, DVI cable, or USB 3.0 cable; the output image frame rate of the miniature camera is not less than 30fps, and the resolution is not less than 1280×720.

[0033] Compared with the prior art, the present invention has the following advantages:

[0034] This invention integrates a miniature camera into a sampling suction tube to acquire real-time images of lesions within the uterine cavity. The image signals are transmitted to a processor, which automatically controls the opening of a regulating valve based on image analysis results. This dynamically adjusts the negative pressure suction output of the vacuum pump. Compared to existing technologies where doctors manually adjust the negative pressure based solely on visual observation of images, this invention achieves automatic closed-loop control of negative pressure suction. This solves the problems of lag and subjective differences caused by manual operation in existing technologies. When the processor detects excessive tissue deformation, it automatically reduces the negative pressure to prevent excessive stretching or breakage of the tissue, ensuring the integrity of the pathological sample and improving the accuracy of pathological diagnosis. When the processor detects insufficient tissue adsorption, it automatically increases the negative pressure to ensure sufficient sampling, reducing repetitive operations and patient discomfort. Furthermore, the combination of image acquisition and processing provides a hardware foundation for subsequent automatic tissue type identification and differentiated negative pressure control, improving the intelligence and standardization of uterine cavity lesion tissue sampling. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the structure of a device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method according to the present invention.

[0036] Figure 2 This is a framework diagram of the processor in this invention;

[0037] Figure 3 This is a flowchart of the negative pressure adaptive control process in this invention.

[0038] 1. Negative pressure suction device; 2. Image acquisition device; 3. Sampling suction tube; 4. Collection container. Detailed Implementation

[0039] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0040] Example 1

[0041] like Figure 1 As shown, this embodiment provides a device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method, including a negative pressure attraction device 1 and an image acquisition device 2;

[0042] The negative pressure suction device 1 includes a sampling suction tube 3, a vacuum pump, a regulating valve, and a collection container 4. The sampling suction tube 3 can be a disposable camera suction tube with a suction port at the front end and a negative pressure connector at the rear end.

[0043] The sampling suction tube 3 has an axially extending baffle inside, which divides the inner cavity of the tube into an isolated image acquisition channel and a negative pressure suction channel.

[0044] The vacuum pump has a negative pressure output port; the regulating valve has an inlet, an outlet and a control terminal. The inlet of the regulating valve is connected to the negative pressure output port of the vacuum pump through a pipeline. The control terminal of the regulating valve is connected to a manual adjustment knob. The collection container 4 has an inlet and an outlet. The inlet of the collection container 4 is connected to the outlet of the regulating valve through a pipeline. The outlet of the collection container 4 is connected to the negative pressure connector of the sampling suction tube 3 through a pipeline.

[0045] Specifically, the image acquisition device 2 includes a miniature camera, a light source assembly, and a signal cable. The miniature camera and the light source assembly are sealed at the front end of the image acquisition cavity of the sampling suction tube 3, with the light source assembly located to one side of the miniature camera. The first end of the signal cable is connected to the signal output end of the miniature camera, and the signal cable passes through the image acquisition cavity. The second end of the signal cable extends to the rear end of the sampling suction tube 3 and forms the image output end, which is connected to a medical monitor via a cable.

[0046] In this embodiment, the sampling suction tube 3 is inserted into the uterine cavity through the vagina, and the medical monitor is turned on. The doctor observes the real-time image on the monitor, manipulates the sampling suction tube 3 to align its suction port with the target lesion tissue, and sets the negative pressure by rotating the manual adjustment knob according to the type of lesion tissue. The negative pressure generated by the vacuum pump is transmitted to the suction port through the regulating valve, drawing the lesion tissue into the negative pressure suction cavity and into the collection container 4. The doctor continuously observes the image on the monitor and manually adjusts the negative pressure according to the tissue deformation. After obtaining a sufficient amount of tissue, the vacuum pump is turned off, the sampling suction tube 3 is withdrawn, and the tissue in the collection container 4 is sent for pathological examination.

[0047] Example 2

[0048] like Figures 1-3 As shown, this embodiment provides a device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method, including a negative pressure attraction device 1 and an image acquisition device 2;

[0049] The negative pressure suction device 1 includes a sampling suction tube 3, a vacuum pump, a regulating valve, a collection container 4, and a processor;

[0050] The sampling suction tube 3 can be a disposable camera suction tube with a suction port at the front end and a negative pressure connector at the rear end. The tube body of the sampling suction tube 3 has an axially extending baffle that divides the tube body into an isolated image acquisition channel and a negative pressure suction channel. The vacuum pump has a negative pressure output port, and the regulating valve has an air inlet, an air outlet, and a control end. The air inlet of the regulating valve is connected to the negative pressure output port of the vacuum pump through a pipeline.

[0051] The collection container 4 has an inlet and an outlet. The inlet of the collection container 4 is connected to the outlet of the regulating valve through a pipeline, and the outlet of the collection container 4 is connected to the negative pressure connector of the sampling suction tube 3 through a pipeline.

[0052] The processor has an image input terminal and a negative pressure control signal output terminal. The image input terminal is connected to the image output terminal of the image acquisition device 2, and the negative pressure control signal output terminal is communicatively connected to the control terminal of the regulating valve.

[0053] The processor integrates an optical flow calculation module, a deformation analysis module, and a control module. The input end of the optical flow calculation module is connected to the image input end, and its output end outputs optical flow field data.

[0054] The input of the deformation analysis module is connected to the output of the optical flow calculation module, and its output outputs tissue deformation parameters.

[0055] The input of the control module is connected to the output of the deformation analysis module, and its output is connected to the negative pressure control signal output of the processor. The control module integrates a safety threshold storage unit, which pre-stores the upper limit and lower limit of the safety deformation parameters corresponding to different tissue types.

[0056] The safety threshold storage unit is connected to the comparator of the control module. The comparator compares the tissue deformation parameters output by the deformation analysis module with the preset safety threshold in the safety threshold storage unit, and generates a boost signal, a depressurization signal or a maintenance signal based on the comparison result. The signal is then output to the control terminal of the regulating valve through the negative pressure control signal output terminal.

[0057] The image acquisition device 2 includes a miniature camera, a light source assembly, and a signal cable. The miniature camera and the light source assembly are sealed at the front end of the image acquisition cavity of the sampling suction tube 3. The light source assembly is located on one side of the miniature camera. The first end of the signal cable is connected to the signal output end of the miniature camera. The signal cable passes through the image acquisition cavity. The second end of the signal cable extends to the rear end of the sampling suction tube 3 and forms an image output end. This image output end is connected to the image input end of the processor.

[0058] When using the device in this embodiment, the sampling suction tube 3 is inserted into the uterine cavity through the vagina. The device is then activated, and the miniature camera collects a continuous sequence of images of the tissue area within the uterine cavity in real time. The image signals are sent to the processor via the image input terminal.

[0059] Specifically, the optical flow calculation module receives a continuous sequence of images and calculates the optical flow field between adjacent image frames;

[0060] Optical flow calculation is based on the assumption of constant brightness, assuming image points The brightness at time t is After time Later movement to Then the constant brightness constraint is ;

[0061] Taylor expansion yields the optical flow constraint equation. ,in , For optical flow velocity components, , , , , , The partial derivatives of image brightness with respect to space and time;

[0062] The optical flow calculation module can use the Lucas-Kanade algorithm to obtain a sparse motion vector field through the least squares method, or use an optical flow estimation network based on a deep neural network to output a dense optical flow field. Since the tissue near the suction port will move towards the suction port under negative pressure, the corresponding region in the optical flow field generates a motion vector pointing towards the suction port.

[0063] Specifically, the sparse optical flow calculation method based on the Lucas-Kanade algorithm utilizes the spatial consistency assumption, assuming that all pixels within a local window W centered on the feature point (of size (2w+1)×(2w+1) pixels) have the same optical flow vector. For each pixel within the window Optical flow constraint equations can be derived for all of them:

[0064] All in the window Transform the equations into matrix form: ,in:

[0065] , The system of equations is overdetermined (the number of equations is greater than the number of unknowns). It is solved using the least squares method, i.e., by finding... make The minimum, its normal solution is:

[0066] ,in, , The summation applies to all pixels within the window;

[0067] To ensure Reversible, the window must contain sufficient texture information (i.e., the gradient is not singular). Before calculating optical flow, this invention first uses the Shi-Tomasi algorithm to extract corner points in the image as feature points. The Shi-Tomasi corner response function is... ,in , For matrix The feature value. Only when the minimum feature value is greater than a preset threshold (e.g., 0.01) is the pixel selected as a feature point, ensuring the stability of the optical flow solution;

[0068] For each selected feature point, a window W (e.g., w=7, window size 15×15 pixels) is taken around it, and the optical flow vector is calculated according to the above formula. ;

[0069] To handle large motions, this invention employs an image pyramid method: the original image is downsampled layer by layer to construct a pyramid, optical flow is calculated starting from the top layer (lowest resolution), and the result is used as the initial value for the next layer for iterative optimization, thereby enabling the tracking of large displacement motions at the pixel level and above.

[0070] When there are many suspended matter, reflective or low-texture areas in the intrauterine fluid environment, the Lucas-Kanade algorithm may fail due to insufficient feature points or noise interference. Therefore, in one embodiment, an optical flow estimation network based on a deep neural network can be used to directly output a dense optical flow field end-to-end.

[0071] Taking the RAFT (Recurrent All-Pairs Field Transforms) network as an example, its specific calculation process is as follows:

[0072] Images of two adjacent frames and The high-dimensional feature maps are obtained by inputting the shared weights into the feature extraction network (composed of multiple convolutional layers). and Where H and W are the height and width of the feature map (usually 1 / 8 of the original image), and D is the feature dimension (e.g., 256).

[0073] calculate Each position in The feature similarity (dot product) of all positions in the matrix is ​​used to obtain the four-dimensional correlation tensor. This tensor describes the probability of each pixel in the first frame matching all pixels in the second frame. To reduce computational cost, the search range can be limited using a local window, or a multi-scale correlation pyramid can be employed.

[0074] Initialize the optical flow field =0, in each iteration, use the current optical flow field pair The features are resampled (warped) to obtain the same as... Alignment features are then identified. Next, the corresponding correlation features are retrieved from the correlation tensor based on the current optical flow position. These features, along with the current optical flow field and image context features, are input into an update module based on a Convolutional Gated Recurrent Unit (ConvGRU). This module outputs the optical flow residual. Used to update the optical flow field: The optical flow field typically converges after 10-20 iterations.

[0075] The resulting low-resolution optical flow field (1 / 8 resolution) is then upsampled to the original image resolution to obtain the dense optical flow vector for each pixel. ;

[0076] In this embodiment, lightweight networks such as LiteFlowNet or PWC-Net can be used, and inference engines such as TensorRT or ONNX Runtime can be used to accelerate the network on embedded GPUs or FPGAs, ensuring that the processing time per frame is less than 33ms (corresponding to 30fps). The network weights are pre-trained on synthetic fluid motion datasets and real hysteroscopic video datasets to adapt them to the image features of the intrauterine fluid environment.

[0077] In the final optical flow field, the tissue region near the suction port will generate a motion vector pointing towards the suction port. The direction of the optical flow vector can be used to determine the movement trend of the tissue: a vector pointing towards the suction port indicates that the tissue is being adsorbed, and a vector moving away from the suction port indicates that the tissue is being expelled or undergoing elastic retraction.

[0078] The deformation rate of a tissue can be quantified based on the magnitude of the optical flow vector (i.e., the velocity of motion): the greater the velocity, the more drastic the deformation of the tissue per unit time, and there is a risk of breakage when it exceeds the safe range;

[0079] By analyzing the spatial distribution of the optical flow vector field, the deformation mode of the tissue can also be determined: if the optical flow vector converges in concentric circles around the suction port, it indicates that the tissue is uniformly stretched; if the vector direction is disordered or there are obvious discontinuous regions, it indicates that there may be tearing or uneven deformation inside the tissue. This information will be transmitted to the deformation analysis module to calculate the tissue strain rate and deformation gradient.

[0080] The motion vector field output by the optical flow calculation module is fed into the deformation analysis module. The deformation analysis module calculates the tissue deformation parameters based on the motion vector field, including the tissue strain rate and deformation gradient. In the two-dimensional case, the components of the strain rate tensor are... ;

[0081] Organizational equivalent variability is defined as ;

[0082] The deformation gradient is obtained by calculating the Jacobian matrix of the optical flow vector field, i.e. The eigenvalues ​​of this matrix reflect the degree of stretching or compression of the tissue in the principal direction, where ∂ represents the sign of the partial derivative. This represents the partial derivative of u with respect to x.

[0083] The control module will calculate the tissue equivalent variability rate in real time. Compared with the thresholds pre-stored in the safety threshold storage unit, a preset safety upper limit is set for endometrial tissue. The lower limit of operation is 0.4 / s. It is 0.1 / s;

[0084] In this embodiment, the control module integrates a safety threshold storage unit and a comparator. The safety threshold storage unit pre-stores the upper limit of the safety deformation parameter corresponding to different tissue types. and lower limit of operation The real-time equivalent strain rate output by the deformation analysis module. The threshold in the safety threshold storage unit is sent to the first input terminal of the comparator, and the threshold is sent to the second input terminal of the comparator.

[0085] The comparator performs the following numerical comparisons within each image frame period: If > If the tissue is determined to be in an overstretched state, the control module generates a decompression signal and sends it to the control terminal of the regulating valve via the negative pressure control signal output terminal. The regulating valve then reduces its opening, lowering the negative pressure output by the vacuum pump. < If the negative pressure is deemed insufficient to effectively adsorb tissue, the control module generates a pressure boosting signal, and the regulating valve increases its opening to increase the negative pressure; if ≤ ≤ The control module generates a sustaining signal, and the regulating valve maintains its current opening.

[0086] Taking the generation of the boost signal as an example, when the comparator determines that boost is needed, the control module calculates the required adjustment step size based on the deviation between the current negative pressure value and the target negative pressure value. This embodiment uses a proportional-integral-derivative (PID) control algorithm to adjust the deviation... Input to PID controller, output control quantity ,in , , These are the proportional, integral, and derivative coefficients, respectively. The control quantity output by the PID controller corresponds to the target opening value of the regulating valve;

[0087] The boost signal, depressurization signal, or maintenance signal generated by the processor is transmitted to the control terminal of the regulating valve in the form of an electrical signal through the negative pressure control signal output terminal. This embodiment provides the following transmission methods:

[0088] Analog voltage signal transmission: The processor integrates a digital-to-analog converter (DAC) to convert digital control signals into 0-10V or 0-5V analog voltage signals. The control terminal of the regulating valve receives this voltage signal, and its internal proportional electromagnet generates a corresponding electromagnetic force based on the voltage magnitude, thereby driving the valve core to move to the target opening degree. The voltage value and valve opening degree have a linear relationship: 0V corresponds to the valve being closed (0% opening), 10V corresponds to the valve being fully open (100% opening), and intermediate voltage values ​​correspond to corresponding proportional opening degrees.

[0089] Analog current signal transmission: The processor converts the control quantity into a 4-20mA current signal through a voltage-controlled current source circuit. 4mA corresponds to valve closure, and 20mA corresponds to valve full opening. Compared with voltage signals, current signals have stronger anti-interference capabilities and are suitable for use in operating room environments with electromagnetic interference.

[0090] PWM (Pulse Width Modulation) signal transmission: The processor generates a square wave signal with a fixed frequency (e.g., 1kHz) and a variable duty cycle. The duty cycle (the proportion of the high-level time to the entire cycle) is linearly related to the valve opening: 0% duty cycle corresponds to the valve being closed, and 100% duty cycle corresponds to the valve being fully open. The PWM signal is filtered and amplified by the internal drive circuit of the regulating valve, and converted into an analog current to control the electromagnet. PWM has the advantages of low power consumption and strong resistance to digital noise.

[0091] Digital communication transmission: The processor sends digital commands containing the target opening value to the control valve via digital communication interfaces such as RS-232, RS-485, CAN bus, or I²C bus. The control valve integrates a microcontroller and communication interface, receives the command, parses it, and executes it. This method is suitable for complex systems that need to control multiple valves simultaneously or require feedback of valve status information.

[0092] Taking PWM signal transmission as an example, when the control module determines that pressure needs to be increased, the processor increases the current PWM duty cycle from 20% to 25%. After the regulating valve control terminal receives the PWM signal with the increased duty cycle, its internal drive circuit converts the PWM signal into a proportional drive current. This current passes through the proportional electromagnet coil, generating an electromagnetic force proportional to the current intensity. The electromagnetic force overcomes the spring force and pushes the valve core to move, increasing the valve opening from 20% to 25%. After the opening increases, the negative pressure flow through the valve increases, and the negative pressure transmitted to the suction port increases accordingly. The pressure sensor monitors the negative pressure value in real time and feeds it back to the processor. The processor further fine-tunes the PWM duty cycle based on the feedback value until the actual negative pressure reaches the target value.

[0093] The above process is continuously looped with the processing cycle of each frame of image, forming a closed-loop control until sampling is completed.

[0094] For example, taking a patient with endometrial polyps as an example, the doctor inserts a disposable camera suction tube into the uterine cavity through the vagina. On the monitor, a polyp with a diameter of about 1.5 cm is observed on the lower wall of the uterine cavity. After the device is started, the vacuum pump initially outputs a negative pressure of 200 mmHg, and the processor begins to acquire image sequences in real time.

[0095] The optical flow calculation module performs calculations on two consecutive frames of images, extracting approximately 300 feature points on the surface of the polyp, with each feature point obtaining a motion vector; in the initial stage, the polyp tissue is stationary, and the optical flow velocity components of each feature point are close to zero;

[0096] When negative pressure is applied to the polyp, the characteristic points in the central region of the polyp begin to move towards the suction port, and the optical flow velocity field... and The absolute value of the component gradually increases;

[0097] The deformation analysis module calculates the equivalent strain rate of the polyp tissue based on the motion vector field. For example, in the first frame image, at a certain feature point... , , , Then the equivalent variability is calculated as follows: The calculated initial equivalent strain rate is approximately 0.049 s². -1 This value is 0.15s below the lower limit of operation for polyp tissue. -1 The control module determines that the negative pressure is insufficient, outputs a boost signal, increases the opening of the regulating valve from 20% to 25%, and increases the negative pressure output by the vacuum pump from 200 mmHg to 250 mmHg.

[0098] After three pressurization adjustments (each increasing by 10 mmHg), when the negative pressure reached 280 mmHg, the equivalent strain rate increased to 0.18 s². -1 Within the safe window of polyp tissue (0.15 s) -1 ~0.6 s -1 Inside, the control module outputs a sustaining signal, and the regulating valve maintains its current opening. At this time, the polyp can be seen being slowly drawn into the suction port on the display screen, with the tissue morphology remaining intact and without excessive stretching or breakage.

[0099] Once the polyp was completely absorbed, the characteristic point motion vector suddenly disappeared, and the equivalent strain rate plummeted to 0.02 s. -1 The control module detects this change, determines that the sampling is complete, automatically adjusts the opening of the regulating valve to zero, stops the vacuum pump from outputting negative pressure, and alerts the doctor through a buzzer. The entire sampling process lasts about 12 seconds, during which the processor makes about 360 negative pressure adjustment decisions (calculated at 30fps), without requiring any manual intervention from the doctor.

[0100] Complete polyp tissue was visible in collection container 4, with a smooth surface and no tears. The pathological section showed clear cell structure, and the diagnosis was endometrial polyp. Compared with traditional manual negative pressure sampling, this device avoids the risk of polyp breakage due to excessive negative pressure, which would make it impossible to determine the pathological grade. It also avoids the need for repeated aspiration due to insufficient negative pressure.

[0101] To further improve control accuracy and adaptability, the processor in this embodiment can also integrate a motion compensation module and a tissue type recognition module. The motion compensation module extracts background feature points in a continuous image sequence, calculates the global optical flow vector generated by the movement of the endoscope itself, and removes the global optical flow vector from the total optical flow field, retaining only the optical flow vector of local tissue deformation caused by negative pressure attraction.

[0102] For example, in actual surgical procedures, when the doctor holds the sampling suction tube 3, slight shaking or active movement is inevitable. These global movements will introduce motion vectors that are unrelated to tissue deformation into the optical flow field, interfering with the accurate analysis of local tissue deformation caused by negative pressure suction.

[0103] The motion compensation module's input is connected to the processor's image input, and its output is connected to the optical flow calculation module's compensation input. The motion compensation module performs the following processing flow within each image frame period:

[0104] The motion compensation module first performs ORB (Oriented FAST and Rotated BRIEF) feature point detection on two consecutive frames of images, and extracts pixels with significant gradient changes in the images as candidate feature points;

[0105] The ORB algorithm combines FAST corner detection and BRIEF feature description, and has the advantages of fast calculation speed, good rotation invariance and robustness to image noise.

[0106] Since structures such as the uterine cavity wall remain relatively stationary during the procedure, feature points in these areas are suitable as background feature points.

[0107] The motion compensation module analyzes the motion consistency of feature points across consecutive frames and clusters feature points with similar motion trajectories into a background feature point set.

[0108] The motion compensation module utilizes a set of background feature points and employs an affine transformation model to describe the global motion of the endoscope. The affine transformation model contains six parameters and can describe motions such as translation, rotation, scaling, and shearing of the image. Its mathematical form is:

[0109] ,in These are the coordinates of the feature points in the previous frame. These are the coordinates of the corresponding feature point in the current frame. , , , , for rotation and scaling parameters, , The translation parameters are used; the motion compensation module fits the six parameters of the above affine transformation model using the least squares method to obtain the transformation matrix describing the global motion of the endoscope.

[0110] Based on the estimated global motion model, the motion compensation module calculates the global optical flow vector generated by the movement of the endoscope itself for each pixel, and then subtracts the global optical flow vector pixel by pixel from the total optical flow field initially calculated by the optical flow calculation module to obtain the remaining optical flow field.

[0111] The remaining optical flow field contains only local tissue deformation motion vectors caused by negative pressure attraction, and a small amount of residual motion that cannot be fitted by the global motion model;

[0112] To further improve accuracy, the motion compensation module also employs an iterative elimination strategy: after initially eliminating global motion, the statistical distribution of the remaining optical flow field is re-analyzed, and outliers whose motion vectors significantly deviate from the mainstream direction are eliminated as outliers. Then, the global motion model is re-estimated using the remaining feature points, and the above process is repeated until the model converges. The final remaining optical flow field is then fed into the deformation analysis module to calculate the organizational deformation parameters.

[0113] Through the above processing, the motion compensation module effectively eliminates the interference of the shaking and movement caused by the doctor holding the sampling suction tube 3 on the tissue deformation analysis, ensuring that the motion vector output by the optical flow calculation module truly reflects the tissue deformation caused by negative pressure suction.

[0114] The tissue type identification module automatically identifies the target tissue type being attracted by analyzing the texture features and optical flow response characteristics in a continuous image sequence. This includes endometrial tissue, polyp tissue, and fibroid tissue. Based on the identification results, the module retrieves the corresponding safety threshold curve from the safety threshold storage unit to achieve differentiated negative pressure control for different tissues.

[0115] Different pathological tissues have different mechanical properties: endometrial tissue is loose and soft, and has a low tolerance to negative pressure; polyp tissue is tougher and can withstand slightly larger negative pressure; fibroid tissue is hard and requires even greater negative pressure for effective adsorption. To achieve differentiated negative pressure control for different tissues, the processor integrates a tissue type recognition module.

[0116] The tissue type recognition module's input is connected to the processor's image input, and its output is connected to the control module's second input. The tissue type recognition module executes the following processing flow within each image frame period:

[0117] The tissue type recognition module first preprocesses the input uterine cavity image, including: using Gaussian filtering to smooth the image to reduce noise; and using histogram equalization to enhance image contrast and make tissue texture features clearer.

[0118] The region of interest (a circular area centered on the suction port, with a diameter approximately 3 to 5 times the diameter of the suction port) is segmented from the entire image to reduce interference from non-target areas;

[0119] The tissue type identification module extracts various texture features from the region of interest, mainly including:

[0120] The gray-level co-occurrence matrix describes the joint gray-level distribution between pixel pairs with a certain spatial relationship in an image, and is a classic method for analyzing the structure of texture.

[0121] The tissue type identification module calculates the gray-level co-occurrence matrix in four directions (0°, 45°, 90°, 135°) within the region of interest, and extracts the following features from each matrix: energy (reflecting the uniformity of the texture), contrast (reflecting the clarity of the texture and the depth of the grooves), correlation (reflecting the directionality of the texture), and entropy (reflecting the randomness and complexity of the texture).

[0122] Different lesion tissues show significant differences in the above-mentioned texture characteristics: endometrial tissue has a fine and uniform texture and a high energy value; polyp tissue has a mulberry-like or villous surface and a high contrast value; fibroid tissue has a dense texture, strong directionality, and a high correlation value.

[0123] In the initial stage of negative pressure attraction (before the negative pressure significantly changes the tissue morphology), different tissues exhibit different optical flow responses to the same negative pressure: endometrial tissue is softer and has a higher initial optical flow velocity; fibroid tissue is harder and has a lower initial optical flow velocity.

[0124] The tissue type identification module extracts indicators such as average optical flow velocity, velocity variance, and optical flow direction consistency of the region of interest in the first 3 to 5 frames of images after negative pressure is applied, as auxiliary identification features.

[0125] The tissue type identification module also extracts morphological features within the region of interest, including: region area (the number of pixels occupied by the lesion tissue in the image), shape factor (the ratio of the square of the region's perimeter to its area, reflecting the regularity of the tissue shape), edge gradient intensity, etc.

[0126] Polyps typically present as papillary or mulberry-like protrusions with irregular edges; fibroids typically present as round or oval bulges with clear borders; endometrial tissue presents as a flat, wavy surface.

[0127] The tissue type identification module inputs the extracted multidimensional features (texture features, optical flow response features, and morphological features) into a pre-trained classifier to make classification decisions;

[0128] The classifier can employ a Support Vector Machine (SVM), which is pre-trained on a large dataset of labeled intrauterine lesion images. It divides the feature space into different tissue type regions by finding the optimal classification hyperplane. During training, the classifier learns the distribution patterns of different tissue types in the feature space and adjusts the model parameters to maximize classification accuracy.

[0129] During the runtime phase, the classifier outputs tissue type labels (endometrial tissue, polyp tissue, or fibroid tissue) based on the features of the currently input image, and also outputs a confidence score for the classification result. When the confidence score is lower than a preset threshold (e.g., 0.7), the control module uses a default safety threshold curve (taking the most conservative parameters, i.e., the safety threshold for endometrial tissue) to ensure safety is prioritized.

[0130] The organization type label output by the organization type identification module is sent to the second input terminal of the control module;

[0131] The control module retrieves the corresponding safety threshold curve from the safety threshold storage unit based on the identification result: if it is identified as endometrial tissue, it retrieves... 0.4 / s The rate is 0.1 / s; if polyp tissue is identified, call... 0.6 / s The rate is 0.15 / s; if it is identified as fibroid tissue, call... 0.8 / s The sampling rate is 0.2 / s. Through the above adaptive threshold call, the device can adopt differentiated negative pressure control strategies for different tissues to achieve the best sampling effect while ensuring safety.

[0132] The negative pressure suction device 1 may also include a pressure sensor, a filter and an overflow prevention device. The pressure sensor is installed on the pipeline between the outlet of the regulating valve and the inlet of the collection container 4. The signal output terminal of the pressure sensor is connected to the pressure signal input terminal of the processor to form a pressure-deformation dual closed-loop control.

[0133] In this embodiment, a pressure sensor is installed on the pipeline between the outlet of the regulating valve and the inlet of the collection container 4. The pressure sensor is a medical-grade silicon piezoresistive pressure sensor with a measurement range of -760 mmHg to 0 mmHg (absolute pressure), an accuracy of ±1% of full scale, and a response time of less than 5 ms. The sensor integrates a temperature compensation circuit, which can maintain stable measurement accuracy within the operating temperature range of 0℃ to 50℃. The signal output terminal of the pressure sensor is connected to the pressure signal input terminal of the processor to transmit the real-time negative pressure value to the processor in the form of an analog voltage signal (0-5V) or a digital signal (I²C bus).

[0134] After receiving the feedback signal from the pressure sensor, the processor fuses it with the optical flow analysis results for judgment. Specifically, the processor's control module simultaneously receives the tissue equivalent strain rate output by the deformation analysis module. and the actual negative pressure value fed back by the pressure sensor The control module executes the following fusion logic:

[0135] (1) When > and When the pressure is within the normal range, it is determined that the excessive stretching of the tissue is caused by excessive negative pressure, and the control module generates a pressure reduction signal;

[0136] (2) When > but When the pressure is below the target negative pressure, it is determined that the excessive stretching of the tissue may be caused by mechanical abnormalities of the tissue itself (such as the tissue being too fragile). The control module maintains the current negative pressure and issues an alarm to alert the doctor.

[0137] (3) When < and When the pressure is lower than the target negative pressure, it is determined that the insufficient negative pressure is indeed due to insufficient output of the vacuum pump, and the control module generates a pressure boosting signal.

[0138] (4) When < but When the target negative pressure has been reached or exceeded, it is determined that the insufficient tissue adsorption may be due to tissue displacement or poor adhesion of the suction port. The control module will issue a prompt suggesting that the doctor adjust the position of the sampling suction tube 3.

[0139] (5) When Inside the safe window but If abnormal fluctuations occur (such as a sudden drop), it is determined that there may be a pipeline leak or blockage, and the control module will issue an alarm.

[0140] The filter is installed between the outlet of the collection container 4 and the negative pressure connector of the sampling suction tube 3;

[0141] An overflow prevention device is installed inside the collection container 4. The output end of the overflow prevention device is connected to the closing control end of the regulating valve. When the liquid level in the collection container 4 reaches the upper limit, the overflow prevention device automatically outputs a closing signal to cut off the negative pressure passage.

[0142] The filter is installed on the pipeline between the outlet of the collection container 4 and the negative pressure connector of the sampling suction tube 3. The specific connection method is as follows: the inlet of the filter is connected to the outlet of the collection container 4 through the pipeline, and the outlet of the filter is connected to the negative pressure connector of the sampling suction tube 3 through the pipeline.

[0143] The filter uses a hydrophobic microporous membrane as the filter medium. The membrane material is polytetrafluoroethylene (PTFE) or polyvinylidene fluoride (PVDF), with a pore size of 0.22 μm to 0.45 μm. The hydrophobic membrane allows gas to pass freely but effectively blocks bacteria, viruses, and other microorganisms in liquids and aerosols. Its main functions include: preventing liquids or aerosols in collection container 4 from being sucked into the vacuum pump under negative pressure, thus protecting the vacuum pump from contamination and corrosion; filtering dust, tissue debris, and other fine particles in the sucked-in gas, preventing these impurities from entering the regulating valve and vacuum pump, affecting the valve's sealing performance and the pump's operating efficiency; and acting as a biosafety barrier to prevent pathogenic microorganisms in collection container 4 from spreading into the environment through the negative pressure pipeline, protecting operators from infection.

[0144] The filter housing is injection molded from transparent medical-grade polycarbonate (PC) material, making it easy to observe the condition of the filter membrane. The filter is designed to be replaceable, and when the filter membrane is contaminated or clogged, it can be easily removed from the tubing and replaced with a new filter. After each surgery, the filter is disposed of as a disposable consumable along with the sampling suction tube 3 to prevent cross-infection.

[0145] An anti-overflow device is installed inside the collection container 4 to automatically cut off the negative pressure passage when the liquid level in the collection container 4 reaches the upper limit, preventing liquid from being sucked into the downstream pipeline and vacuum pump.

[0146] The core component of the overflow prevention device is a float-type liquid level shut-off valve. Specifically, a vertical guide rod is installed inside the outlet of the collection container 4 (i.e., inside the collection container 4, at the inlet of the outlet pipe). A hollow float is fitted onto the guide rod. The float is made of a low-density material (such as polypropylene or polyethylene), with a density less than water, allowing it to float in the liquid. An elastic sealing gasket is provided on the upper end of the float, directly opposite the outlet port of the collection container 4. The inner edge of the outlet port of the collection container 4 has a sealing surface that mates with the sealing gasket.

[0147] The overflow prevention device works based on liquid buoyancy. When the liquid level in collection container 4 is low, the float, under the influence of gravity, is positioned at the lower end of the guide rod, remaining separated from the outlet port. This ensures the outlet pipe remains unobstructed, allowing negative pressure to be transmitted normally. As the surgery progresses, the aspirated tissue and fluid continuously enter collection container 4, causing the liquid level to gradually rise.

[0148] When the liquid level reaches the preset upper limit (usually 80% to 90% of the volume of the collection container 4), the float floats upward along the guide rod under the action of liquid buoyancy. When the float rises to the top, the elastic sealing gasket on its upper end face fits tightly with the outlet port of the collection container 4, forming a seal under the action of negative pressure, physically blocking the outlet passage, thereby cutting off the transmission of negative pressure and preventing the liquid from being further sucked into the downstream pipeline.

[0149] Meanwhile, the overflow prevention device also integrates a signal output function. A miniature reed switch is set at the top of the guide rod, and a permanent magnet is embedded inside the float. When the float rises to the top, the permanent magnet approaches the reed switch, triggering the switch to close.

[0150] The output of the reed switch is connected to the liquid level alarm input of the processor via a sealed lead, sending a full liquid level signal to the processor. Upon receiving this signal, the processor performs the following operations: automatically outputs a shut-off signal to the control terminal of the regulating valve, adjusting the valve opening to zero, actively cutting off the negative pressure path, forming a dual protection system with mechanical shut-off; it issues an audible and visual alarm via a buzzer or display screen, prompting the doctor that collection container 4 is full and needs to be replaced or emptied; in automatic control mode, the sampling operation can be temporarily paused, automatically resuming after collection container 4 is replaced.

[0151] This embodiment uses a processor to achieve real-time quantitative analysis of tissue deformation and adaptive adjustment of negative pressure, which can ensure sufficient sampling while protecting the integrity of tissue samples, thereby improving the safety of intrauterine lesion tissue sampling and the accuracy of pathological diagnosis.

[0152] The specific embodiments of the invention have been described in detail above, but these are merely examples. The invention is not limited to the specific embodiments described above. Those skilled in the art should understand that the embodiments and descriptions in the specification are only illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method, characterized in that, Includes a negative pressure suction device and an image acquisition device: The negative pressure suction device includes a sampling suction tube, a vacuum pump, a regulating valve, and a collection container. The front end of the sampling suction tube is provided with a suction port, and the rear end of the sampling suction tube is provided with a negative pressure connector. The vacuum pump includes a negative pressure output port; The regulating valve includes an air inlet, an air outlet, and a control terminal, wherein the air inlet is connected to the negative pressure output port of the vacuum pump; A collection container has an inlet and an outlet, the inlet being connected to the air outlet of the regulating valve, and the outlet being connected to the negative pressure connector of the sampling suction tube. The collection container is connected in series in the negative pressure passage between the regulating valve and the sampling suction tube. The negative pressure suction device also includes a processor for controlling the output negative pressure of the vacuum pump. The processor includes an image input terminal and a negative pressure control signal output terminal. The image input terminal is connected to the image output terminal of the image acquisition device, and the negative pressure control signal output terminal is communicatively connected to the control terminal of the regulating valve. The image acquisition device includes a miniature camera integrated into the sampling suction tube and a light source assembly disposed on one side of the miniature camera. The image acquisition device also includes a signal cable, which includes a first end and a second end, the first end of which is communicatively connected to the signal output end of the miniature camera.

2. The device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method according to claim 1, characterized in that, The processor integrates: The optical flow calculation module has its input end connected to the image input end and its output end outputting optical flow field data. The deformation analysis module has its input end connected to the output end of the optical flow calculation module, and its output end outputs tissue deformation parameters. The control module has its input terminal connected to the output terminal of the deformation analysis module, and its output terminal connected to the negative pressure control signal output terminal of the processor.

3. The device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method according to claim 1, characterized in that, The sampling suction tube has an axially extending baffle inside, which divides the inner cavity of the tube into an isolated image acquisition channel and a negative pressure suction channel. The miniature camera and the light source are sealed at the front end of the image acquisition cavity. The signal cable passes through the image acquisition cavity, and the second end of the signal cable extends to the rear end of the sampling suction tube and constitutes the image output end of the hysteroscopy. The front opening of the negative pressure suction channel forms the suction port, and the negative pressure connector is located at the rear end of the sampling suction tube and communicates with the negative pressure suction channel.

4. The device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method according to claim 2, characterized in that, The negative pressure suction device also includes: A pressure sensor is installed on the pipeline between the outlet of the regulating valve and the inlet of the collection container, and the signal output terminal of the pressure sensor is connected to the pressure signal input terminal of the processor. A filter is disposed between the outlet of the collection container and the negative pressure connector of the sampling suction tube; An overflow prevention device is provided inside the collection container, and the output end of the overflow prevention device is connected to the closing control end of the regulating valve.

5. The device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method according to claim 2, characterized in that, The control module integrates a safety threshold storage unit, which is connected to the comparator of the control module. The comparator compares the tissue deformation parameters output by the deformation analysis module with the preset safety threshold in the safety threshold storage unit, and generates a boost signal, a depressurization signal or a maintenance signal based on the comparison result. The boost signal is then output to the control terminal of the regulating valve through the negative pressure control signal output terminal.

6. The device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method according to claim 2, characterized in that, The processor also integrates a motion compensation module, the input of which is connected to the image input, and the output of which is connected to the compensation input of the optical flow calculation module.

7. The device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method according to claim 2, characterized in that, The processor also integrates a tissue type recognition module. The input of the tissue type recognition module is connected to the image input, and the output of the tissue type recognition module is connected to the second input of the control module.

8. The device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method according to claim 2, characterized in that, The deformation parameters output by the deformation analysis module include the tissue strain rate and deformation gradient.

9. The device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method according to claim 2, characterized in that, The optical flow calculation module is any one of the Lucas-Kanade optical flow calculation unit, the Horn-Schunck optical flow calculation unit, or an optical flow estimation unit based on a deep neural network.

10. The device for intrauterine tissue deformation analysis and adaptive control of negative pressure attraction based on optical flow method according to claim 1, characterized in that, The second end of the signal cable of the image acquisition device is connected to the image input terminal of the processor via a cable.

Images