An endoscopic biopsy guidance system and control method

Abstract

The application discloses an endoscope biopsy guiding system and a control method. The system comprises an endoscope front-end probe assembly, an image processing host and an operation control terminal. The microscopic image acquisition unit in the endoscope front-end probe assembly is pre-aligned with the physical sampling path of the biopsy tool. The image processing host is configured to acquire a microscopic image, process and generate an evaluation signal representing the pathological value of the target tissue, judge whether the evaluation signal meets the preset trigger condition, and send an enable signal to the operation control terminal to drive the operation limiting unit to release the operation limitation of the sampling execution component when the condition is met. The method comprises corresponding image acquisition, evaluation signal generation, condition judgment and unlocking control steps. By constructing a "verify first and sample later" physical interlocking mechanism, the application can ensure that the sampling operation is only performed in the high pathological value area, effectively avoiding the misoperation caused by blind sampling, and significantly improving the accuracy of biopsy sampling and the positive rate of diagnosis.

Description

Technical Field

[0001] This invention relates to the fields of medical devices, medical image processing, and embedded intelligent control. Specifically, it relates to an endoscopic biopsy guidance system and control method that dynamically controls biopsy sampling operations through real-time in-situ microscopic image feature extraction and temporal evaluation in respiratory intervention, digestive endoscopy, or natural orifice surgery scenarios, using a physical interlocking method. Background Technology

[0002] Endoscopic biopsy is a crucial method for obtaining tissue samples from lesions within the body for pathological diagnosis, such as transbronchial lung biopsy (TBLB), transbronchial lymph node needle aspiration (TBNA), or sampling of submucosal lesions in the digestive tract. Existing navigation technologies (such as electromagnetic navigation, virtual endoscopy, or ultrasound guidance) primarily rely on preoperative computed tomography (CT) or intraoperative macroscopic ultrasound (EBUS) to achieve anatomical localization at the millimeter to centimeter level.

[0003] However, existing biopsy guidance technologies have significant microscopic blind spots in clinical practice. First, there is a serious disconnect between macroscopic anatomical localization and the microscopic properties of tissue. Current navigation accuracy is limited to spatial positioning and cannot distinguish the microscopic heterogeneity within the lesion. A lesion often contains areas of highly diagnostically valuable active tumor cells, areas of necrotic tissue with no diagnostic value, and areas of inflammation. When the biopsy tool (such as a biopsy needle) extends to the end of the endoscopic working channel and contacts the target tissue, the system lacks needle-tip-level in-situ tissue property feedback, resulting in a sampling process that is essentially a blind sampling state lacking real-time tissue property perception.

[0004] This blind sampling method, lacking in-situ microscopic verification, directly leads to high clinical costs and patient risks. Because the effectiveness of the tissue touched by the needle tip cannot be confirmed in real time, the operator is very likely to collect necrotic or healthy tissue. This not only increases the operation time and additional trauma to the patient, but may also lead to "false negative" results in pathological diagnosis, forcing the patient to undergo a second operation, or even delaying the best treatment window due to diagnostic misdiagnosis.

[0005] Furthermore, the endoscopic sampling environment faces severe challenges in maintaining dynamic spatiotemporal consistency. Strong physiological movements during the procedure (such as respiration and heartbeat) and vibrations at the endoscope tip keep the target constantly in dynamic displacement. Existing pathological image analysis technologies are mostly designed for ex vivo, static pathological slide images, and generally lack the ability to perform millisecond-level real-time processing of dynamic intraoperative video streams and convert them into deterministic hardware execution signals. This disconnect between perception and execution results in an invisible time lag between "target identification" and "physical sampling," making it difficult to ensure the physical consistency of the sampling position in dynamic environments.

[0006] Therefore, how to construct a closed-loop guidance system with in-situ microscopic sensing capabilities and the ability to achieve "verification before sampling" physical intervention through intelligent algorithms to solve the problems of low sampling accuracy and unstable positive rate in endoscopic biopsy is a technical challenge that urgently needs to be solved in this field. Summary of the Invention

[0007] This invention addresses the problems mentioned in the background section by providing an endoscopic biopsy guidance system and control method that can intelligently and physically control biopsy sampling operations based on real-time microscopic image analysis results, thereby improving the accuracy and positive rate of biopsies.

[0008] At the system level, this invention provides an endoscopic biopsy guidance system. The system includes: an endoscope front probe assembly, comprising a microscopic image acquisition unit for acquiring microscopic images of target tissue, and a biopsy tool for sampling the target tissue; wherein the imaging field of view of the microscopic image acquisition unit is pre-aligned with the physical sampling path of the biopsy tool; The image processing host is communicatively connected to the microscopic image acquisition unit; An operation control terminal includes an operation restriction unit and a sampling execution component, wherein the sampling execution component is configured to control the sampling action of the biopsy tool, and the operation restriction unit is configured to impose restrictions on the operation of the sampling execution component; The image processing host is configured as follows: Acquire the microscopic images acquired by the microscopic image acquisition unit; The microscopic images are processed to generate an evaluation signal characterizing the pathological value of the target tissue; Determine whether the evaluation signal meets the preset triggering conditions; When the preset triggering condition is met, an enable signal is sent to the operation control terminal to drive the operation restriction unit to release the operation restriction on the sampling execution component.

[0009] The system consists of an endoscopic front-end probe assembly with in-situ microscopic sensing capabilities, an image processing host responsible for high-performance inference calculations, and an intelligent control handle integrating physical interlocking logic. The probe assembly integrates a set of microscopic imaging units (such as a confocal fiber bundle) and sampling tools (such as a biopsy needle), preferably arranged coaxially to ensure a high degree of overlap between the imaging field of view and the sampling physical path. The host establishes a high-speed electrical signal connection with the microscopic imaging units to receive and analyze dynamic microscopic video streams in real time. The intelligent control handle embeds an electromagnetic locking mechanism controlled by the host, which physically intervenes in the biopsy push rod in the default state. The host continuously compares the real-time acquired microscopic images with a pre-set pathological feature library by running specific deep learning algorithms (such as instance-aware algorithms), thereby generating an evaluation signal reflecting the pathological value of the tissue. When this signal meets a preset trigger threshold, the host immediately sends an unlocking command to the handle, driving the locking mechanism to release the physical restriction on the sampling execution component. Regarding the evaluation logic and control precision, to address the interference caused by complex physiological movements during surgery, the image processing host pre-executes image sharpness evaluation logic before performing pathological assessment. The system only triggers subsequent pathological value analysis when the image sharpness value is greater than a preset threshold and the image quality is sufficient to support feature extraction. Furthermore, the host introduces a temporal confidence superposition mechanism to transform the feature matching degree identified by the image into a pathological assessment signal with high confidence. To ensure the real-time performance and accuracy of the sampling operation, the system establishes a deterministic low-latency control link to ensure that the end-to-end processing latency from image acquisition to physical lock unlocking adapts to the frame rate requirements of endoscopic navigation. On this basis, the system further introduces dynamic logic arbitration based on temporal continuity: by maintaining a rolling time window to statistically analyze the cumulative value of pathological confidence within the window, the unlocking signal is finally triggered only when the confidence remains stable within a preset time and the pose change rate of the probe relative to the target tissue is lower than a preset motion threshold. This effectively avoids erroneous locking caused by instantaneous noise or violent probe sliding, ensuring precise spatiotemporal coupling between physical sampling actions and high-value pathological areas.

[0010] At the methodological level, this invention provides a control method for an endoscopic biopsy guidance system. This method first captures dynamic microscopic images of the target tissue in real time using a microscopic imaging unit, and simultaneously performs a quality screening step to filter out motion-blurred frames caused by physiological activities such as breathing and heartbeat. For valid image frames that meet the clarity requirements, the system uses a trained intelligent strategy network to extract their pathological features and generate an evaluation signal. When the evaluation signal indicates that the tissue currently contacted by the probe has high diagnostic value, the system automatically generates and issues an unlocking command to the locking mechanism on the handle side, thereby allowing the sampling execution component to complete a controlled, evidence-supported physical sampling action. The invention also includes an intelligent control handle comprising a locking mechanism and a sampling execution component. The sampling execution component is mechanically linked to the biopsy tool, and the locking mechanism is used to apply or release physical restrictions on the sampling execution component. When the system enters a locked state, the intelligent control handle is configured to generate specific force feedback, and the triggering of this force feedback is time-prior to the state reset at the software instruction level. In a preferred embodiment, the locking mechanism is an electromagnetic locking mechanism; the sampling execution component is a biopsy push rod. When the system enters a locked state, the intelligent control handle is configured to generate specific force feedback, and the triggering of this force feedback is time-prior to the reset at the software instruction level, thereby constructing a master-slave interlocked physical interaction feature. The invention also relates to related computer-readable storage media and program products for supporting the stable operation of the above-mentioned intelligent closed-loop control logic on medical device hardware.

[0011] The technical solution provided by this invention has the following beneficial effects: 1. This invention achieves a cross-dimensional improvement in biopsy navigation accuracy, filling the gap in tip-level in-situ verification. By integrating a microscopic imaging unit at the tip of the biopsy tool, this application elevates the accuracy of biopsy navigation from the traditional macroscopic anatomical level (centimeter-level) to the microscopic cellular pathology level (micrometer-level). This "tip-as-field-of-view" design allows the operator to acquire the microscopic morphological characteristics of the tissue at the sampling point in real time at the moment of sampling, thereby eliminating positioning deviations caused by anatomical heterogeneity at the physical contact level. 2. The "verify first, then sample" physically enforced interlocking mechanism constructed in this invention guarantees the biopsy positivity rate from the underlying logic. Unlike traditional methods that rely solely on visual cues to assist decision-making, this system uses an electromagnetic locking mechanism to physically control the biopsy execution components. This rigid constraint of "not unlocking except in high-value areas" fundamentally prevents invalid sampling in necrotic, inflamed, or normal tissue areas, thereby significantly reducing the "false negative" rate caused by blind sampling, minimizing the risk and suffering of patients requiring secondary surgery, and also greatly shortening the exploration time for finding effective lesions, thus optimizing the clinical surgical pathway.

[0012] 3. This invention significantly improves the robustness of the system in complex and dynamic environments through multi-dimensional image quality control and temporal feature enhancement. Considering the respiratory peristalsis and endoscopic vibration present in the in vivo environment, this application introduces image sharpness assessment as a preliminary step in pathological value analysis. This effectively filters out image frames with motion blur or quality degradation caused by physiological activities such as respiratory movements, ensuring the quality of input data for subsequent AI analysis. Simultaneously, by introducing temporal confidence evaluation logic, the system can effectively overcome single-frame noise interference, ensuring that the issuance of unlocking commands has extremely high determinism. This allows the system to maintain stable and reliable guidance performance in dynamic and changing in vivo environments. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a structural block diagram of an endoscopic biopsy guidance system provided in an embodiment of the present invention.

[0015] Figure 2 This is a schematic diagram of the main flow of an endoscopic biopsy guidance system control method provided in an embodiment of the present invention.

[0016] Figure 3 This is a detailed flowchart of the core control logic (EVE strategy) provided in the embodiments of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0018] See Figure 1 The endoscopic biopsy guidance system (100) mainly consists of three parts: an endoscope front probe assembly (101), an image processing host (102), and an intelligent control handle (103) for operation control terminal. In addition, the system also includes a power module (not shown) that provides power to the entire system and a display screen (not shown) for presenting images and system status to the operator.

[0019] The endoscope tip probe assembly (101) is the core actuator for achieving "needle-tip" in-situ sensing and sampling. Its overall outer diameter is designed to be less than 2.0 mm to fit the working channels of standard bronchoscopes or gastrointestinal endoscopes. In this embodiment, the assembly adopts a coaxial nested composite structure. Its central part is a biopsy tool, specifically a biopsy needle made of biocompatible medical-grade metal materials, such as medical stainless steel, nickel-titanium alloy, or metal-polymer composite materials. Its inner diameter is approximately 0.8 mm, used for puncture and obtaining tissue samples. A microscopic imaging unit, specifically a confocal fiber bundle, is tightly attached to the outer wall of the needle via a heat-shrink tubing. This structure ensures that the center of the field of view of the confocal imaging is spatially highly aligned with the axial puncture path of the biopsy needle, thereby eliminating the "parallax shift" common in macroscopic navigation. The confocal fiber bundle has a miniature objective lens at its end with a numerical aperture (NA) of not less than 0.7 to ensure high-resolution imaging; its working distance is set to 0 μm to support contact in-situ microscopy imaging mode; and the imaging field of view (FoV) diameter can be selected between 300 μm and 600 μm.

[0020] It should be noted that although this embodiment uses a confocal fiber bundle as the microscopic imaging unit, this application is not limited to this. In other embodiments, the microscopic imaging unit can also be other technologies capable of achieving real-time imaging at micrometer-level resolution, such as optical coherence tomography (OCT) probes or two-photon microscopy probes.

[0021] The image processing unit (102) is the "brain" of the system, responsible for image acquisition, processing, analysis, and decision-making. It is electrically connected to the confocal fiber bundle of the endoscope front probe assembly (101) via an optoelectronic interface. The unit contains a laser scanning unit (LSU) that provides excitation light at a wavelength of 488 nm and receives fluorescence signals in the 500-650 nm band through the same optical path. To achieve fluorescence imaging, the operation process may include a step of pre-application of a fluorescent contrast agent to the target area. The unit also has an embedded AI computing module, which can be built on the NVIDIA Jetson AGX platform or other platforms with equivalent computing power, running the core real-time control inference algorithm. The AI ​​computing module is also connected to a memory for storing a pre-set library of standard lesion examples, which consists of images of typical lesion features (such as adenocarcinoma nests, squamous cell carcinoma keratin beads, etc.) pre-annotated by pathologists. Meanwhile, the module also interacts with the handle through a deterministic low-latency communication link (such as CAN FD or real-time industrial Ethernet) to ensure that the end-to-end latency from image feature extraction to hardware action triggering is controlled within 33ms, adapting to an image update rate of 30fps.

[0022] The intelligent control handle (103) is the terminal for operator-system interaction, located outside the endoscope. The handle integrates a locking mechanism (1031) and a sampling execution component (1032). In this embodiment, the locking mechanism (1031) is specifically a high-frequency response solenoid lock, whose latch is mechanically engaged in the limiting groove of the sampling execution component (1032) in the default state (Fail-safe mode). This default state can be either power-off closed or power-on closed to adapt to different safety design requirements. The sampling execution component (1032) is specifically a biopsy push rod, which is mechanically linked to the biopsy needle at the front end via a flexible push rod. When the locking mechanism (1031) is locked, the forward movement of the biopsy push rod is physically restricted, thereby preventing sampling. The handle also integrates a haptic feedback motor to provide vibration cues to the operator under specific conditions. To construct an independently observable physical interlock feature, when the system enters the locked state, the intelligent control handle (103) is configured to generate specific force feedback (such as constant resistance or vibration), and the triggering timing of this physical action is strictly earlier than the state reset at the software instruction level of the image processing host (102). The image processing host (102) and the intelligent control handle (103) communicate via wired communication, such as RS-485 industrial bus, CAN bus, or other real-time communication interfaces, ensuring that the instruction response delay is less than 10ms.

[0023] It should be noted that the locking mechanism (1031) is not limited to an electromagnetic locking mechanism, but can also be other devices that can achieve fast and reliable physical locking, such as a locking pin mechanism driven by a micro servo motor or a micro pneumatic locking device.

[0024] Those skilled in the art should understand that although this embodiment uses a "smart control handle" as an example to illustrate the specific form of the operation control terminal, this application is not limited thereto. The operation control terminal can be any device that allows the operator to issue sampling action commands and has physical feedback or limiting functions, including but not limited to handheld handles, foot pedal controllers, desktop consoles, or robot auxiliary operating arms connected to them.

[0025] Similarly, the physical limiting unit is not limited to the electromagnetic locking mechanism (such as an electromagnetic solenoid lock) described in the embodiments. Any device that can block, disengage, or increase damping of the movement of mechanical parts based on control signals is covered within the scope of protection of this application. For example, it can be a piezoelectric ceramic brake, a miniature hydraulic lock, a servo motor (which achieves locking by providing reverse torque), or a clutch mechanism (which mechanically disconnects the operating end from the actuating end when not unlocked, resulting in no-load operation).

[0026] Furthermore, regarding the preset triggering conditions, although this embodiment uses "the evaluation signal exceeds the threshold" as an example, in other embodiments, the triggering conditions may also be that the evaluation signal falls into a specific numerical range or matches a specific pathological feature classification label.

[0027] Based on the aforementioned hardware structure, the system in this embodiment operates according to a core control logic of "Explore-Verify-Exploit" (EVE). The specific flow of this logic can be combined with... Figure 2 and Figure 3 To understand.

[0028] Figure 2 The main flow of the control method of the present invention is shown. In step S201, when the operator moves the endoscope tip probe assembly (101) to the target lesion area and slides it in contact with the tissue surface, the system acquires microscopic images of the target tissue in real time through the microscopic imaging unit, forming a continuous video stream. The frame rate should be no less than 12 fps (preferably 30 fps) to complete dynamic image acquisition and quality pre-screening (Explore).

[0029] Subsequently, in step S202, the image processing host (102) processes the acquired microscopic images to generate an evaluation signal characterizing the pathological value of the target tissue. This processing step is the core of the EVE strategy, and it can be further subdivided into, for example... Figure 3 The two stages are shown.

[0030] The first stage is "Explore," which involves image quality perception and filtering (step S301). Due to breathing movements or operator hand tremors, the acquired video stream inevitably contains motion-blurred image frames. To ensure the accuracy of subsequent analysis, the system first processes each image frame... A sharpness assessment is performed. Specifically, the system quantifies the sharpness of the frame image by calculating its Laplacian variance. In a preferred embodiment, this calculation logic is implemented using the following formula:

[0031] in, It is an image sharpness value that characterizes image sharpness; It is the current image frame; It is a Gaussian smoothing kernel, used for preprocessing to reduce the impact of noise; It is the Laplacian operator, used to extract the second-order gradient information of an image. The clearer the image, the greater the gradient change. This indicates the variance calculation. The system has a preset resolution threshold. In step S302, the system determines the calculated... Is it greater than .like If the frame is deemed blurry, it is discarded (step S303) and not proceeded to subsequent processing. If a frame is clear, it is determined to be a clear frame and proceeds to the next stage. This step effectively filters out blurry frames, providing high-quality input for subsequent pathological value analysis.

[0032] The second stage is "Verification," which involves instance-aware matching of pathological value (step S304). For image frames that pass the sharpness screening, the system employs an image recognition algorithm, specifically an instance-aware algorithm based on Siamese Network, to determine whether the cell morphology within the current field of view matches the characteristics of the target lesion. This algorithm model contains two convolutional neural network (CNN) backbone branches with shared weights (such as the ResNet-50 model). Branch A receives the currently acquired sharp frame as input. Branch B takes a standard lesion instance image extracted from a pre-set lesion feature library as input. Both branches extract feature vectors from their respective input images. and Subsequently, the system generates a pathological value assessment signal by calculating the similarity between the two feature vectors. This signal is specifically a feature matching confidence score. In a preferred embodiment, the similarity calculation may use cosine similarity, which is then normalized using the Sigmoid function. The calculation logic is implemented using the following formula:

[0033] in, This is the feature matching confidence score, ranging from [0, 1]. A higher value indicates a higher degree of matching between the current real-time image and the lesion features. To smooth out random errors, the system introduces a temporal confidence score stacking mechanism. For example, it maintains a rolling time window and calculates the confidence score of consecutive valid frames within the window. The cumulative average value is used as the final evaluation signal.

[0034] Back Figure 2 After generating the pathological value assessment signal, the system proceeds to step S203 to determine whether the assessment signal exceeds a preset trigger threshold. (For example, 0.85). This judgment process belongs to the third stage of the EVE strategy, "Exploit," which is physical interlock control. This control logic is implemented by a state machine: - If the judgment result is negative, the system remains in the "Searching" state. At this time, the image processing host (102) either does not send or sends a locking signal to the smart control handle (103), keeping its locking mechanism (1031) locked. To enhance the perceptibility and unavoidability of the master-slave interlock, when the system enters the locked state, the smart control handle (103) will generate force feedback such as constant resistance or vibration, and the triggering timing of this force feedback hardware action is earlier than the state reset at the software instruction level. Simultaneously, the state of the red indicator box can be maintained on the display screen.

[0035]

[0036] - If the judgment result is yes, that is The system will then further determine whether the duration of this state exceeds a preset dwell time threshold. (For example, 200ms) to prevent false triggering caused by transient noise. When the duration of the high-confidence state exceeds At this point, the system enters the "Target Acquired" state and executes step S204.

[0037] In step S204, the image processing host (102) generates and sends an unlock signal (e.g., a GPIO high-level signal) to the smart control handle (103). This unlock signal drives the locking mechanism (1031) (electromagnetic solenoid lock) to retract its latch, thereby releasing the physical restriction on the sampling execution component (1032) (biopsy push rod). At this time, the operator can push the biopsy push rod to complete a valid sampling. Simultaneously, to provide clear feedback to the operator, the haptic feedback motor vibrates at a frequency of 200Hz, and the indicator box on the display screen turns green and highlights.

[0038] In summary, this embodiment constructs a closed-loop intelligent guidance system by combining microscopic imaging, real-time AI analysis, and a physical interlocking mechanism. It enforces a "verification before sampling" workflow, physically allowing sampling only when the system confirms that the probe tip is aligned with a tissue area of ​​high pathological value, thereby improving the accuracy and positive rate of biopsies. Example

[0039] This embodiment provides a control method for an endoscopic biopsy guidance system based on the system described in Embodiment 1. The execution subject of this method is the image processing host (102) described in Embodiment 1. Specific steps of this method can be found in [reference needed]. Figure 2 and Figure 3 The flowchart.

[0040] The method includes: Step S201: The dynamic microscopic image stream of the target tissue is acquired in real time at a preset frame rate (e.g., 30fps) through the microscopic imaging unit in the endoscope front probe assembly (101).

[0041] Steps S301-S303: Before processing the microscopic image stream to generate the evaluation signal, the sharpness of each acquired microscopic image frame is evaluated (the Laplacian variance is calculated) to obtain the image sharpness value. Determine whether the sharpness value is greater than a preset sharpness threshold. Only when the image frame exceeds this threshold is it considered a "valid frame" and used for subsequent processing steps; otherwise, it is discarded. This step establishes a pre-emptive quality assurance mechanism, corresponding to the technical features of claim 8, and effectively improves the system's anti-interference capability.

[0042] Step S304: For image frames that pass the resolution screening, the image processing host (102) runs a built-in image recognition algorithm to compare them with a preset lesion feature library, and processes the microscopic image to generate an evaluation signal characterizing the pathological value of the target tissue. This evaluation signal is the feature matching confidence level. .

[0043] Step S203: The host determines whether the preset triggering conditions are met based on both temporal continuity and spatial stability criteria. Specifically, this includes: pathological dimension determination—judging the evaluation signal ( Does it last for a period of time? Exceeding a preset trigger threshold ( )(like >0.85); Dwell Time Determination—Determine whether the duration of a high-confidence state reaches the dwell time threshold. T d (e.g., 200ms); Motion stability determination—The system uses feature point tracking or optical flow algorithms to synchronously evaluate the rate of change of the probe's pose relative to the target tissue. ,like Below the preset exercise threshold T m If the current probe is in a stable contact state, it is determined that it has the physical spatial consistency required for sampling.

[0044] Step S204: When the judgment result is yes, the host enters the "target locked" state, generates and sends an unlock signal to the locking mechanism (1031) in the smart control handle (103) to release the physical restriction on the sampling execution component (1032). The smart control handle (103) is configured to perform a physical security boundary priority confirmation timing sequence: at the moment T1 when the unlock signal is received, the locking mechanism (1031) immediately performs a physical unlocking action and drives the smart control handle (103) to generate corresponding physical force feedback characteristics (such as the instantaneous disappearance of handle damping or controlled vibration prompt); at the moment T2, which is later than the moment T1, the image processing host (102) updates the display status of its software interface (such as updating the locked status icon to the allowed sampling status). And the preset time difference The time difference is no less than 10ms. This timing difference ensures that the operator has confirmed the physical sampling admission boundary through haptic feedback before the visual logic state is aligned.

[0045] This method embodiment is an embodiment corresponding to the foregoing system embodiment and can be implemented in conjunction with the system embodiment. The relevant technical details (such as deterministic low-latency control, coaxial arrangement field of view alignment, etc.) and beneficial effects described in the foregoing system embodiment are also applicable to this method embodiment, and will not be repeated here. Explanation and further explanation of the terminology: 1. Regarding "generating assessment signals characterizing pathological value" and "determining preset trigger conditions": In a preferred embodiment, the system employs an EVE strategy (clarity assessment combined with temporal continuity confidence) to implement this logic; however, in other embodiments of this application, image processing and assessment signal generation can also be based on end-to-end deep learning classification models, traditional image template matching algorithms, or multispectral feature analysis algorithms; simultaneously, the preset trigger conditions, in addition to frame continuity and pose change rate, can also include auxiliary conditions such as probe contact pressure threshold or spectral absorbance. Any system that can output trigger commands based on image features falls within the scope of protection of this application.

[0046] 2. Regarding the “operation restriction unit”: In the preferred embodiment, it is mainly embodied as an electromagnetic lock or a mechanical damper, but in other embodiments, the operation restriction unit may also be a pneumatic locking valve, a shape memory alloy (SMA) brake, or an electrical hard isolation device that is achieved by cutting off electronic control signals in a fly-by-wire system. Those skilled in the art will understand that all or part of the steps in the above methods can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, such as a read-only memory (ROM), random access memory (RAM), a magnetic disk, or an optical disk. Accordingly, the present invention also provides a computer program product containing this program.

[0047] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An endoscopic biopsy guidance system, characterized in that, include: An endoscope front probe assembly includes a microscopic image acquisition unit for acquiring microscopic images of a target tissue, and a biopsy tool for sampling the target tissue; wherein the imaging field of view of the microscopic image acquisition unit is pre-aligned with the physical sampling path of the biopsy tool. The image processing host is communicatively connected to the microscopic image acquisition unit; An operation control terminal includes an operation restriction unit and a sampling execution component, wherein the sampling execution component is configured to control the sampling action of the biopsy tool, and the operation restriction unit is configured to impose restrictions on the operation of the sampling execution component; The image processing host is configured as follows: Acquire the microscopic images acquired by the microscopic image acquisition unit; The microscopic images are processed to generate an evaluation signal characterizing the pathological value of the target tissue; Determine whether the evaluation signal meets the preset triggering conditions; When the preset triggering condition is met, an enable signal is sent to the operation control terminal to drive the operation restriction unit to release the operation restriction on the sampling execution component.

2. The endoscopic biopsy guidance system according to claim 1, characterized in that: The microscopic image acquisition unit is an in-situ microscopic image acquisition unit, and the center of the imaging field of view is aligned with the physical sampling path of the biopsy tool at a preset sampling depth; The operation restriction unit is configured to impose restrictions on the operation of the sampling execution unit in the default state; The image processing host is configured as follows: Receive the real-time video stream acquired by the microscopic image acquisition unit; The image frames in the video stream are analyzed for sharpness, and the pathological features of valid image frames that meet the sharpness threshold are extracted to generate the evaluation signal. The evaluation signal is determined based on the continuity in the time domain to determine whether it meets the preset triggering conditions. The preset triggering conditions include at least the following: the cumulative value of the confidence of the evaluation signal within a preset time window reaches a confidence threshold, and the pose change rate of the endoscope front probe assembly relative to the target tissue is lower than a preset motion threshold.

3. The endoscopic biopsy guidance system according to claim 1, characterized in that, The sampling execution component and the biopsy tool are mechanically linked through a mechanical transmission component; the operation restriction unit is a physical restriction unit, used to apply physical restrictions to the movement of the sampling execution component by at least one of mechanical blocking, increasing damping, electrical disconnection or logical interception.

4. The endoscopic biopsy guidance system according to claim 3, characterized in that, The operation control terminal is an intelligent control handle; the physical restriction unit is a locking mechanism, which is used to apply or release mechanical locking to the sampling execution component; the enable signal is an unlock signal; wherein, when the system enters the locked state, the intelligent control handle is configured to generate specific force feedback, and the action trigger of the force feedback is earlier in timing than the state reset at the software instruction level.

5. The endoscopic biopsy guidance system according to claim 2, characterized in that, The in-situ microscopic image acquisition unit is a confocal fiber bundle; the biopsy tool is a biopsy needle; the confocal fiber bundle is attached to the outer wall of the biopsy needle and the two extend side by side along the axial direction, or the confocal fiber bundle and the biopsy needle are coaxially sleeved.

6. The endoscopic biopsy guidance system according to claim 4, characterized in that, The locking mechanism is an electromagnetic locking mechanism; the sampling execution component is a biopsy push rod.

7. The endoscopic biopsy guidance system according to claim 1 or 2, characterized in that, The image processing host is configured to compare the microscopic image with a preset lesion feature library by running an image recognition algorithm to generate the pathological value assessment signal; wherein the pathological value assessment signal is the feature matching confidence level.

8. The endoscopic biopsy guidance system according to claim 7, characterized in that, The image recognition algorithm is an instance-aware algorithm or a semantic segmentation algorithm.

9. The endoscopic biopsy guidance system according to claim 2, characterized in that, The image processing host is further configured to: before generating the pathological value assessment signal, use a gradient operator or frequency domain analysis algorithm to first assess the sharpness of the microscopic image to obtain an image sharpness value; wherein, the operation of generating the pathological value assessment signal is performed only when the image sharpness value is greater than a preset sharpness threshold.

10. A control method for an endoscopic biopsy guidance system, characterized in that, Includes the following steps: Acquire microscopic images of the target tissue by the microscopic image acquisition unit; The microscopic images are processed to generate an evaluation signal characterizing the pathological value of the target tissue; Determine whether the evaluation signal meets the preset triggering conditions; When the preset triggering condition is met, an enable signal is generated and sent to the operation restriction unit of the operation control terminal to release the operation restriction on the sampling execution component.

11. The control method according to claim 10, characterized in that: The acquisition of microscopic images specifically involves: real-time acquisition of dynamic microscopic video streams of the target tissue; The specific steps of processing the microscopic image are as follows: performing a sharpness assessment on the video stream to screen out valid image frames; extracting pathological features from the valid image frames to generate the assessment signal; The determination of whether the evaluation signal meets the preset triggering conditions specifically involves: determining whether the evaluation signal meets the preset triggering conditions based on the continuity of the time domain. The preset triggering conditions include at least the cumulative value of the confidence of the evaluation signal within a preset time window reaching a confidence threshold, and the pose change rate of the probe relative to the target tissue being lower than a preset motion threshold.

12. The control method according to claim 11, characterized in that, The length of the preset time window or the confidence threshold is dynamically adjusted based on the pose change rate.

13. A computer-readable storage medium having a computer program stored thereon, the computer program implementing the control method according to any one of claims 10-12 when executed by a processor.

14. A computer program product comprising a computer program that, when executed by a processor, implements the control method according to any one of claims 10-12.

Images