A bronchoscope navigation correction system, method and computer readable storage medium

Abstract

The application discloses a bronchoscope navigation deviation correction system, method and computer readable storage medium, comprising: a data acquisition module for acquiring multi-source perioperative data at the current sampling time, including intraoperative ventilator parameters, intraoperative image data and navigation system data; a deviation minimum phase finding module for finding the deviation minimum phase in the respiratory cycle by using a navigation deviation prediction model when the bronchoscope tip approaches the lesion to perform lesion treatment; a deviation prediction module for performing deviation prediction according to the multi-source perioperative data to obtain a three-dimensional deviation vector of the bronchoscope tip relative to the actual position of the lesion, a spatial distance deviation and a deviation confidence; and a deviation adjustment module for performing path screening and parameter adjustment according to the three-dimensional deviation vector, the spatial distance deviation, the deviation confidence and the deviation minimum phase to perform navigation deviation correction. The application improves the accuracy and efficiency of bronchoscope navigation positioning, and improves the safety and success rate of surgery.

Description

Technical Field

[0001] This application relates to the field of navigation bronchoscopy technology, and in particular to a bronchoscopy navigation correction system, method and computer-readable storage medium. Background Technology

[0002] Navigation-guided bronchoscopy has become an important tool for the diagnosis and treatment of small pulmonary nodules. Currently, it typically involves constructing a three-dimensional airway map based on preoperative chest computed tomography (CT) images, combined with electromagnetic or shape sensing technologies to guide the bronchoscope to the target lesion. However, because the navigation system relies on preoperative CT data, discrepancies often exist between the patient's actual intraoperative anatomical position and the preoperative images, leading to navigation positioning errors.

[0003] One of the main reasons for discrepancies between CT scans and intraoperative positioning is changes in the patient's respiratory movements and anesthesia ventilation patterns. Intraoperative atelectasis is another common problem affecting navigation accuracy. Currently, these issues are mainly addressed through postoperative correction, requiring multiple imaging scans, which inevitably increases surgical time and X-ray exposure. Even with CBCT assistance, intraoperative imaging is still necessary to adjust catheter position. Improvements in anesthesia ventilation strategies have also been used to reduce navigation errors, but these are pre-defined, standardized protocols and lack universal applicability.

[0004] In summary, how to effectively address the issues of passively remedying errors after they occur, or adopting uniform preventative strategies that affect the accurate localization of lesions and the success rate of sampling, thus becoming a bottleneck limiting the diagnostic efficacy of navigation bronchoscopy, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] The purpose of this application is to provide a bronchoscopy navigation and correction system that improves the accuracy and efficiency of bronchoscopy navigation and positioning, and enhances the safety and success rate of surgery. Another purpose of this application is to provide a bronchoscopy navigation and correction method and a computer-readable storage medium.

[0006] To solve the above-mentioned technical problems, this application provides the following technical solution: A bronchoscopy navigation and correction system, comprising: The data acquisition module is used to acquire multi-source perioperative data at the current sampling time; wherein, the multi-source perioperative data includes intraoperative ventilator parameters, intraoperative imaging data, and navigation system data; The minimum deviation phase search module is used to find the minimum deviation phase in the respiratory cycle by using the navigation deviation prediction model based on the ventilator parameters when the tip of the bronchoscope reaches the preset distance range of the actual location of the lesion, and outputs the minimum deviation phase so as to perform lesion treatment in the minimum deviation phase. The deviation prediction module is used to predict deviations based on the multi-source perioperative data using the navigation deviation prediction model, and to obtain the three-dimensional deviation vector, spatial distance deviation, and deviation confidence level of the bronchoscope tip relative to the actual position of the lesion. The deviation adjustment module is used to perform path filtering and parameter adjustment based on the three-dimensional deviation vector, the spatial distance deviation, the deviation confidence level, and the time phase of minimum deviation, so as to perform navigation correction.

[0007] In one specific embodiment of this application, the deviation adjustment module is specifically used to obtain the simulation deviations corresponding to each feasible path obtained from the preoperative simulation; determine the airway status from each feasible path to the actual location of the lesion based on the spatial distance deviation; and filter the paths from each feasible path based on the airway status and each simulation deviation; wherein, each simulation deviation is a deviation obtained by simulation based on historical data in the database using the navigation deviation prediction model.

[0008] In one specific embodiment of this application, the deviation adjustment module is specifically used to output a prompt message indicating an increase in tidal volume or positive end-expiratory pressure when the predicted result corresponding to the spatial distance deviation is that the tidal volume is too low, causing insufficient lung volume and resulting in a shift in the actual location of the lesion.

[0009] In one specific embodiment of this application, the deviation adjustment module is specifically used to output a tidal volume reduction prompt when the prediction result corresponding to the spatial distance deviation is that the tidal volume is too high, causing lung overinflation and resulting in inaccurate positioning.

[0010] In one specific embodiment of this application, a data preprocessing module is also included; The data preprocessing module is used to perform time synchronization and preprocessing on the multi-source perioperative data to obtain processed multi-source perioperative data. The deviation prediction module is specifically used to predict deviations based on the processed multi-source perioperative data using the navigation deviation prediction model.

[0011] In one specific embodiment of this application, the deviation adjustment module is specifically used to maintain the current navigation path and prompt for continued operation when the spatial distance deviation is less than a preset threshold; when the spatial distance deviation exceeds the preset threshold and the deviation confidence meets the requirements, correct the lesion target position according to the deviation direction determined by the three-dimensional deviation vector, update the relative relationship between the bronchoscope tip and the lesion, and recalculate the candidate path; when the deviation confidence is insufficient, output a prompt message for intraoperative image review or re-registration.

[0012] In one specific embodiment of this application, it further includes: The prompt information output module is used to output calibration or image confirmation prompt information when it is determined that the spatial distance deviation is less than a preset safety threshold.

[0013] In one specific embodiment of this application, it further includes: The display module is used to highlight the selected target path through the navigation software interface after path filtering and parameter adjustment based on the spatial distance deviation, and to pop up parameter adjustment prompts in the display interface.

[0014] In one specific embodiment of this application, the deviation prediction module includes: The data input unit is used to receive the multi-source perioperative data sent by the data acquisition module, and to perform structured processing on the multi-source perioperative data to obtain structured numerical features. The feature encoding unit is used to receive the structured numerical features sent by the data input unit, normalize the structured numerical features, and extract path numerical features, pose numerical features, device status numerical features and respiratory parameter numerical features through a fully connected network, and output the extracted numerical features to the feature fusion unit. The temporal modeling unit is used to receive airway motion data, endoscope pose change data and navigation error change data in the continuous respiratory cycle of the multi-source perioperative data provided by the data input unit, and input them into the long short-term memory network or temporal convolutional network according to time steps, output the corresponding temporal feature vector through the hidden state of the network, and output the temporal feature vector to the feature fusion unit. The image feature extraction unit is used to receive intraoperative fluoroscopic images, CBCT local reconstruction images or registration residual maps from the multi-source perioperative data provided by the data input unit, and to extract lesion area features, catheter tip area features and their relative spatial relationship features using a convolutional neural network, and output the extracted image features to the feature fusion unit. The feature fusion unit is used to stitch and fuse the numerical features, the temporal feature vector and the image features, and output the fused features to the deviation regression unit. The deviation regression unit is used to output the spatial distance deviation between the bronchoscope tip and the actual position of the lesion based on the fusion features.

[0015] A bronchoscopic navigation correction method includes: Acquire multi-source perioperative data at the current sampling time; wherein, the multi-source perioperative data includes intraoperative ventilator parameters, intraoperative imaging data, and navigation system data; When the tip of the bronchoscope reaches the preset distance range of the actual location of the lesion, the navigation deviation prediction model is used to find the phase with the smallest deviation in the respiratory cycle according to the ventilator parameters, and the phase with the smallest deviation is output so that the lesion can be treated in the phase with the smallest deviation. The navigation deviation prediction model is used to predict deviations based on the multi-source perioperative data, and the three-dimensional deviation vector, spatial distance deviation, and deviation confidence level of the bronchoscope tip relative to the actual position of the lesion are obtained. Path selection and parameter adjustment are performed based on the three-dimensional deviation vector, the spatial distance deviation, the deviation confidence level, and the time phase of minimum deviation to perform navigation correction.

[0016] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the bronchoscopic navigation correction method as described above.

[0017] The bronchoscopy navigation correction system provided in this application includes: a data acquisition module for acquiring multi-source perioperative data at the current sampling time; wherein, the multi-source perioperative data includes intraoperative ventilator parameters, intraoperative imaging data, and navigation system data; a minimum deviation phase search module for finding the minimum deviation phase in the respiratory cycle using a navigation deviation prediction model based on ventilator parameters when the bronchoscope tip reaches a preset distance range from the actual location of the lesion, and outputting the minimum deviation phase for lesion treatment during the minimum deviation phase; a deviation prediction module for predicting deviation using the navigation deviation prediction model based on the multi-source perioperative data, obtaining a three-dimensional deviation vector, spatial distance deviation, and deviation confidence level of the bronchoscope tip relative to the actual location of the lesion; and a deviation adjustment module for path selection and parameter adjustment based on the three-dimensional deviation vector, spatial distance deviation, deviation confidence level, and minimum deviation phase for navigation correction.

[0018] As can be seen from the above technical solution, by utilizing multi-source data during the surgical procedure, a navigation deviation prediction model predicts in real time the spatial distance deviation between the bronchoscope tip and the actual position of the lesion, and provides optimized navigation path recommendations and parameter adjustment suggestions based on the prediction results. This allows for dynamic correction of the navigation strategy without interrupting the surgical procedure, transforming passive intraoperative calibration into proactive prevention. It significantly improves the accuracy and efficiency of bronchoscopic navigation positioning, solves the problem of inaccurate positioning caused by respiratory displacement and anatomical changes, improves the biopsy hit rate of small lung lesions, reduces the number of additional image corrections and radiation exposure, and enhances the safety and success rate of the surgery.

[0019] Accordingly, this application also provides a bronchoscopic navigation correction method and a computer-readable storage medium corresponding to the above-mentioned bronchoscopic navigation correction system, which have the above-mentioned technical effects, and will not be elaborated here. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a structural block diagram of a bronchoscopic navigation and correction system according to an embodiment of this application; Figure 2 This is a flowchart illustrating the implementation of a bronchoscopic navigation correction method in an embodiment of this application.

[0022] The following labels are shown in the attached diagram: 1-Data acquisition module, 2-Minimum deviation phase search module, 3-Deviation prediction module, 4-Deviation adjustment module. Detailed Implementation

[0023] Navigation-guided bronchoscopy has become an important tool for the diagnosis and treatment of small pulmonary nodules. Current navigation systems include virtual bronchoscopy navigation, electromagnetic navigation (EMN), and robot-assisted surgical platforms. These systems typically construct a three-dimensional airway map based on preoperative chest CT images and combine electromagnetic or shape sensing technologies to guide the bronchoscope to the actual location of the lesion. Compared to traditional experience-based procedures, these navigation technologies significantly improve the success rate of biopsy diagnosis. However, because navigation systems rely on preoperative CT data, there is often a discrepancy between the patient's actual anatomical position during surgery and the preoperative images, a phenomenon known as "CT-to-body divergence," which leads to errors in navigation positioning.

[0024] One of the main reasons for discrepancies between CT scans and intraoperative location is the alteration of the patient's respiratory movements and anesthesia ventilation methods. Bronchial and lung tissues undergo displacement and deformation during respiratory movements: preoperative CT scans are often performed during deep inspiration, while bronchoscopy is usually conducted under general anesthesia with intubation and muscle relaxation. The ventilation mode and lung volume distribution differ from the preoperative values, causing the actual location of the lesion to deviate from the planned path. Studies have shown that the average positional change of peripheral pulmonary nodules between extreme respiratory phases (deep inspiration to deep expiration) can reach 17.6 mm, with some lower lobe nodules near the diaphragm moving by more than 60 mm. Even under quiet ventilation conditions, each respiratory cycle can cause a maximum displacement of approximately 25 mm in bronchial structures. This means that lower lobe lesions are significantly affected by diaphragmatic movement, and their localization diagnostic success rate is often lower than that of upper lobe lesions, which are less affected by respiratory movements.

[0025] Besides respiratory motion, intraoperative atelectasis is another common problem affecting navigation accuracy. General anesthesia and prolonged supine position can lead to lung tissue collapse, causing the target lesion to be obscured by airless lung tissue, increasing the difficulty of navigation and positioning. Intraoperative cone-beam computed tomography (CBCT) imaging studies have found that up to 75% of patients under routine ventilation experience varying degrees of atelectasis, with the lesion completely obscured by collapsed lung tissue in approximately 38% of cases. The severity of atelectasis is related to the duration of anesthesia and the patient's body size; the longer the anesthesia time and the higher the body mass index, the greater the probability of multiple lung segments collapsing. Therefore, displacement caused by respiratory motion and lung tissue collapse together constitute important causes of navigation errors and sampling failures.

[0026] To address the aforementioned issues, some improvements have been implemented, but limitations remain. On one hand, navigation systems are increasingly incorporating intraoperative imaging techniques for error correction. For example, CBCT can provide real-time three-dimensional images, helping surgeons confirm the positional relationship between the bronchoscope and the lesion, thus allowing for correction when deviations are too large. Combining CBCT with navigation enhancement technology can promptly reposition the instrument when it is found to have misaligned the lesion, significantly improving positioning accuracy. However, this method is post-operative correction, requiring multiple imaging scans, inevitably increasing surgical time and X-ray exposure. Even with CBCT assistance, studies have found that approximately 67% of cases experience target deviation during initial intubation positioning, necessitating intraoperative imaging adjustments to the catheter position.

[0027] In addition, some navigation systems incorporate respiratory gating technology and biphasic CT imaging to reduce positioning errors caused by breathing. For example, 4D electromagnetic navigation systems acquire CT scans of the patient's end-inspiratory and end-expiratory phases preoperatively to create a more accurate airway "map," and use respiratory gating to track nodule movement with respiration in real time during surgery, achieving automatic registration and dynamic calibration. This "biphasic CT + real-time calibration" method reduces image drift caused by physiological motion to some extent, but it still cannot solve the influence of uncontrollable factors such as intraoperative atelectasis, and requires additional image acquisition procedures and equipment support.

[0028] On the other hand, improvements in anesthesia ventilation strategies have also been used to reduce navigation errors. For example, the "Lung Navigation Ventilation Protocol (LNVP)" and "Ventilation to Prevent Atelectasis (VESPA)" developed in recent years reduce the impact of atelectasis and respiratory motion by adjusting ventilator parameters. Comparative studies show that when using LNVP, the proportion of target lesions obscured by atelectasis is significantly reduced from about 36% under conventional ventilation conditions to only about 4%–8%, and the incidence of lung collapse is also significantly reduced. LNVP uses higher positive end-expiratory pressure (PEEP), larger tidal volume, and combined with brief breath-holding to simultaneously address lung collapse and respiratory motion. VESPA, on the other hand, mainly prevents atelectasis through moderate PEEP, with relatively limited control over motion. Although specific ventilation strategies help improve lesion visualization and diagnostic success rates, they are pre-set uniform protocols and lack a mechanism for real-time adjustment based on individual intraoperative conditions. Therefore, whether relying on passive correction of intraoperative images or passive prevention through fixed ventilation protocols, these technologies remain insufficient in addressing dynamically changing navigation errors.

[0029] In summary, current bronchoscopic navigation systems suffer from significant shortcomings in positioning accuracy: there is a lack of an intelligent method that can comprehensively utilize multi-source perioperative data, predict navigation errors in real time, and proactively propose optimization suggestions. Related technologies often resort to reactive remediation after errors occur or adopt a "one-size-fits-all" preventative strategy, failing to anticipate and optimize the specific circumstances of different patients and surgical stages. This limitation affects the accurate localization of lesions and the success rate of sampling, becoming a bottleneck restricting the diagnostic efficacy of navigation-guided bronchoscopy. New technological solutions are urgently needed to overcome these deficiencies and improve the reliability and effectiveness of navigation-guided surgery.

[0030] Therefore, the bronchoscopy navigation correction system provided in this application transforms passive intraoperative calibration into active prevention, significantly improving the accuracy and efficiency of bronchoscopy navigation and positioning, increasing the hit rate of small lung lesion biopsies, reducing the number of additional image corrections and radiation exposure, and improving the safety and success rate of the surgery.

[0031] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0033] See Figure 1 , Figure 1 This is a structural block diagram of a bronchoscopic navigation and correction system according to an embodiment of this application. The system may include: Data acquisition module 1 is used to acquire multi-source perioperative data at the current sampling time; the multi-source perioperative data includes intraoperative ventilator parameters, intraoperative imaging data and navigation system data; The minimum deviation phase search module 2 is used to find the minimum deviation phase in the respiratory cycle based on the ventilator parameters when the tip of the bronchoscope reaches the preset distance range of the actual location of the lesion, and outputs the minimum deviation phase so that the lesion can be treated in the minimum deviation phase. The deviation prediction module 3 is used to predict deviations based on multi-source perioperative data using the navigation deviation prediction model, and obtain the three-dimensional deviation vector, spatial distance deviation, and deviation confidence level of the bronchoscope tip relative to the actual position of the lesion. The deviation adjustment module 4 is used to perform path selection and parameter adjustment based on the three-dimensional deviation vector, spatial distance deviation, deviation confidence, and the time phase of minimum deviation in order to perform navigation correction.

[0034] The bronchoscopic navigation correction system provided in this application embodiment may include a data acquisition module 1, a minimum deviation phase search module 2, a deviation prediction module 3, and a deviation adjustment module 4. The data acquisition module 1 acquires multi-source perioperative data at the current sampling time, including intraoperative ventilator parameters, intraoperative imaging data, and navigation system data.

[0035] Ventilator parameters can include tidal volume, respiratory rate, positive end-expiratory pressure, peak airway pressure, and lung compliance indices, which are obtained from the anesthesia machine or respiratory monitoring equipment. These parameters reflect the patient's current lung ventilation status and respiratory movement amplitude.

[0036] Intraoperative imaging data primarily refers to CBCT 3D images acquired intraoperatively, combined with 2D fluoroscopic images or ultrasound images when necessary. By performing cone-beam CT scans during navigation, the system processes and analyzes the obtained CBCT images, such as reconstructing the airway tree and lesion location, and measuring the lesion's offset relative to the bronchoscope. For situations where frequent CBCT acquisition is inconvenient, the system can also utilize a small number of keyframe images intraoperatively (e.g., acquiring a 3D scan upon reaching the segmental airway) to extract key location information and infer data between image acquisitions through interpolation or modeling.

[0037] Navigation system data includes real-time bronchoscopic pose (position and orientation), virtual distance to the target lesion, and number of airway branch levels traversed, provided by electromagnetic or optical positioning systems. The navigation system also provides information on the preoperative planned route (e.g., the planned branch sequence) and kinematic data such as the current endoscope posture and acceleration. This data reflects the progress, position, and orientation of the bronchoscopy within the airway tree.

[0038] The output of the navigation deviation prediction model can be configured according to different usage scenarios. In the respiratory cycle optimization scenario, the navigation deviation prediction model outputs the recommended respiratory phase or time window with the smallest deviation based on the predicted deviation under different respiratory phases. This guides the system to perform positioning confirmation, puncture, or ablation operations in a stable phase. When the bronchoscope tip reaches the preset distance range of the actual lesion location, the minimum deviation phase search module 2 uses the navigation deviation prediction model to find the minimum deviation phase in the respiratory cycle based on ventilator parameters and outputs the minimum deviation phase so that critical lesion treatment can be performed during the minimum deviation phase. For example, it may suggest "advancing the biopsy needle at the next end-inspiratory pause" because the navigation deviation prediction model predicts that lung motion is temporarily still at this time, with the smallest deviation. For lesions with significant respiratory motion, the system may suggest a brief breath-hold to fix the lung position before performing a puncture biopsy. These suggestions can be combined with the rhythm of the navigation system. When the bronchoscope reaches the preset distance range, the system actively counts down and prompts the anesthesiologist to cooperate in performing a brief apnea, so that the sampling process is completed within the window of minimum motion. By identifying the phase with the smallest deviation in the respiratory cycle and treating the lesion during that phase, the success rate of the surgery was further improved.

[0039] The deviation prediction module 3 uses a navigation deviation prediction model to predict deviations based on multi-source perioperative data, obtaining the spatial distance deviation between the bronchoscope tip and the actual location of the lesion. The deviation adjustment module 4 performs path selection and parameter adjustment based on the spatial distance deviation, thereby enabling timely navigation correction.

[0040] The output of the navigation deviation prediction model can be configured according to different usage scenarios. In real-time navigation correction scenarios, the navigation deviation prediction model outputs a three-dimensional deviation vector, spatial distance deviation value, and deviation confidence score of the bronchoscope tip relative to the actual position of the lesion, used to indicate the current navigation error and correction direction. After acquiring multi-source perioperative data from the data acquisition module 1, the deviation prediction module 3 uses the navigation deviation prediction model to predict deviations based on the multi-source perioperative data, obtaining a three-dimensional deviation vector, spatial distance deviation, and deviation confidence score of the bronchoscope tip relative to the actual position of the lesion. This data can then be sent to the deviation adjustment module 4. The deviation adjustment module 4 performs path filtering and parameter adjustment based on the three-dimensional deviation vector, spatial distance deviation, deviation confidence score, and the time phase of minimum deviation.

[0041] By outputting a three-dimensional deviation vector, the system not only learns the magnitude of the spatial distance deviation but also clarifies its specific spatial direction, such as up / down, left / right, and front / back. This allows the deviation adjustment module 4 to accurately determine the actual location of the lesion relative to the bronchoscope tip, thereby enabling targeted correction of the navigation path, adjustment of target coordinates, or guidance device posture. Deviation confidence reflects the reliability of the navigation deviation prediction model for the predicted result. When the confidence is high, the system can perform automatic correction or provide explicit suggestions. When the confidence is low, the system can prompt the operator to perform intraoperative image review or re-registration, avoiding erroneous operations based on unreliable predictions. This significantly reduces the clinical risks caused by model misjudgments and improves the system's safety and clinical acceptability. By incorporating the time phase of minimum deviation into the decision-making criteria of the deviation adjustment module 4, the system can perform key operations such as path selection, target correction, and biopsy sampling within the time window of minimum deviation in the respiratory cycle. This minimizes the impact of respiratory motion on navigation accuracy, further improving the positioning hit rate and sampling success rate.

[0042] In the path adjustment scenario, the navigation deviation prediction model outputs the deviation risk level of the current path, recommended path branches, mirror advance depth adjustment amount, end attitude adjustment direction, and operation parameter correction suggestions, which are then used by the deviation adjustment module 4 for path selection and navigation parameter updates.

[0043] Offline training of the model can utilize a large amount of historical surgical data: for each case, ventilator parameter curves, navigation path information, intraoperative image calibration results, and the actual spatial distance deviation between the bronchoscope tip and the actual location of the lesion are collected during the operation. By extracting and labeling features from the historical dataset, a navigation error prediction model is trained using regression algorithms or deep learning networks.

[0044] During offline model training, historical surgical data, animal experiments, and phantom experiments can be used as training samples. Each sample includes preoperative CT planning data, intraoperative navigation trajectory, respiratory parameter curves, intraoperative image calibration results, and the actual location of the lesion and the actual location of the catheter tip confirmed by CBCT, fluoroscopic enhanced images, or postoperative review. The training label is the three-dimensional spatial deviation vector of the catheter tip relative to the actual location of the lesion and its spatial distance deviation. The model is trained through supervised learning to minimize the error between the predicted deviation and the actual deviation. The loss function may include three-dimensional coordinate deviation loss, spatial distance loss, and respiratory phase consistency loss.

[0045] The model can take various forms, such as multiple linear regression, decision tree regression, and random forest regression, to fit the functional relationship between the magnitude of spatial distance deviation (a continuous variable) and each input feature. Alternatively, a neural network model such as a multilayer perceptron (MLP), convolutional neural network (CNN), or long short-term memory network (LSTM) can be constructed to handle complex nonlinear relationships and temporal features, outputting an estimate of the spatial distance deviation. For cases involving image data, a CNN can be used to extract image features, which are then fused with other numerical features before being input into a fully connected network for regression.

[0046] The training objective of the navigation deviation prediction model is to ensure that the predicted spatial distance deviation of the bronchoscope tip relative to the actual lesion location closely approximates the true spatial distance deviation. Common evaluation metrics include mean squared error (MSE). After training, the navigation deviation prediction model is deployed in the navigation system for real-time inference. During actual surgery, the system inputs feature data collected at the current moment or within a recent window into the navigation deviation prediction model to calculate the predicted spatial distance deviation in real time. The model's inference results are continuously updated, with particular attention paid to when the bronchoscope penetrates deep into the lung and approaches the target lesion. By introducing the navigation deviation prediction model, the system can promptly capture the impact of multiple factors on positioning accuracy during surgery, enabling early quantitative assessment of potential errors.

[0047] This application breaks down the barriers between patient physiological parameters and navigation positioning systems. By integrating respiratory dynamics, imaging calibration results, and instrument position information, it comprehensively characterizes the factors affecting navigation accuracy. Multi-source data fusion technology enhances the model's adaptability to complex situations, enabling accurate error estimation even in cases of atelectasis and abnormal breathing patterns. Through training regression models or neural networks, a nonlinear mapping between navigation deviation and multiple factors is established, achieving real-time numerical prediction of deviations. Compared to manual experience or simple threshold judgments, machine learning models can comprehensively weigh the influence of multiple parameters and uncover hidden patterns, such as discovering the amplifying effect of the combination of "too many branch levels + low tidal volume" on deviation. The navigation deviation prediction model can quantify potential deviations seconds to minutes in advance, giving the surgical team ample time to respond. This early prediction elevates the navigation system from a passive positioning tool to an "active sensing" device, significantly improving the robustness and safety of navigation in complex situations.

[0048] By utilizing a navigation deviation prediction model to predict in real time the spatial distance deviation and trend between the bronchoscope tip and the actual location of the lesion, the degree of deviation between the actual lesion location and the virtual planned location can be estimated in advance during the bronchoscope's approach to the target lesion, thus proactively alerting the surgeon to potential navigation error risks. Based on the prediction results, optimized navigation path recommendations and parameter adjustment suggestions are provided, enabling dynamic correction of the navigation strategy without interrupting the surgical procedure. This transforms passive intraoperative calibration into proactive prevention, significantly improving the accuracy and efficiency of bronchoscopic navigation positioning, resolving the problem of inaccurate positioning due to respiratory displacement and anatomical changes, increasing the biopsy hit rate of small lung lesions, reducing the number of additional image corrections and radiation exposure, and improving the safety and success rate of the surgery.

[0049] As can be seen from the above technical solution, by utilizing multi-source data during the surgical procedure, a navigation deviation prediction model predicts in real time the spatial distance deviation between the bronchoscope tip and the actual position of the lesion, and provides optimized navigation path recommendations and parameter adjustment suggestions based on the prediction results. This allows for dynamic correction of the navigation strategy without interrupting the surgical procedure, transforming passive intraoperative calibration into proactive prevention. It significantly improves the accuracy and efficiency of bronchoscopic navigation positioning, solves the problem of inaccurate positioning caused by respiratory displacement and anatomical changes, improves the biopsy hit rate of small lung lesions, reduces the number of additional image corrections and radiation exposure, and enhances the safety and success rate of the surgery.

[0050] It should be noted that, based on the above embodiments, this application also provides corresponding improvement solutions. In subsequent embodiments, steps that are the same as or corresponding to those in the above embodiments can be referred to each other, and the corresponding beneficial effects can also be referred to each other. These improvements will not be elaborated upon in the following improved embodiments.

[0051] In one specific embodiment of this application, the deviation adjustment module 4 is specifically used to obtain the simulation deviations corresponding to each feasible path obtained from the preoperative simulation; determine the airway status from each feasible path to the actual location of the lesion based on the spatial distance deviation; and filter the paths from each feasible path based on the airway status and each simulation deviation; wherein, each simulation deviation is a deviation obtained by simulation using a navigation deviation prediction model based on historical data in the database.

[0052] If the target lesion in the lung can be reached via different bronchial branch paths, the system will recommend the optimal path based on the deviation prediction assessment of each candidate path using a navigation deviation prediction model. For example, the model may predict that navigation via a certain path will result in less displacement error due to breathing because the airway is relatively straight and has fewer branch levels, while another path will have a larger positioning deviation because it needs to pass through too many small airways or large-angle bends. The system will combine the simulated deviation and success rate of each path to recommend the path plan with the least deviation and the highest probability of success to the operator. In actual execution, if the current path navigation error is found to be consistently high and an alternative path can reach the target, the system may also prompt to consider changing the path.

[0053] The route recommendation is based on simulation or rehearsal of the navigation process. Preoperatively, a navigation deviation prediction model is used to simulate the route based on historical data in the database, obtaining the simulated deviation for each feasible route. During the operation, the latest parameters are used for real-time correction and evaluation. After predicting the spatial distance deviation between the bronchoscope tip and the actual lesion location, the deviation adjustment module 4 obtains the simulated deviation for each feasible route obtained from the preoperative simulation. Based on the spatial distance deviation, it determines the airway condition from each feasible route to the actual lesion location, and selects a route from the feasible routes based on the airway condition and the simulated deviation. Through this mechanism, the navigation system no longer only provides a static route map but can dynamically suggest which route has the smallest error under current conditions, helping the surgeon make better decisions.

[0054] In one specific embodiment of this application, the deviation adjustment module 4 is specifically used to output a prompt message indicating an increase in tidal volume or positive end-expiratory pressure when the predicted result corresponding to the spatial distance deviation is that the tidal volume is too low, causing insufficient lung volume and resulting in a shift in the actual location of the lesion.

[0055] When the predicted result corresponding to the spatial distance deviation is that insufficient lung volume due to low tidal volume leads to lesion displacement (such as "overflow" displacement), the deviation adjustment module 4 outputs a prompt to increase tidal volume or positive end-expiratory pressure, thereby suggesting that the anesthesiologist appropriately increase the tidal volume or positive end-expiratory pressure to increase alveolar inflation and reduce the displacement of the lesion with respiration. By transforming complex information into concise action suggestions, the risk of errors caused by insufficient human experience or negligence is significantly reduced.

[0056] The navigation deviation prediction model can also estimate the quantitative impact of parameter adjustments on the error, providing hints such as "If the tidal volume is expected to increase from 300ml to 500ml, the prediction error can be reduced from 10mm to 5mm".

[0057] In one specific embodiment of this application, the deviation adjustment module 4 is specifically used to output a tidal volume reduction prompt when the prediction result corresponding to the spatial distance deviation is that the tidal volume is too high, causing lung overinflation and resulting in inaccurate positioning.

[0058] When the predicted result corresponding to the spatial distance deviation is that excessive tidal volume causes lung overinflation and leads to inaccurate positioning, the deviation adjustment module 4 outputs a tidal volume reduction prompt, thereby suggesting that the anesthesiologist appropriately reduce the tidal volume. By transforming complex information into concise action suggestions, the risk of errors caused by insufficient human experience or negligence is significantly reduced.

[0059] In one specific embodiment of this application, the system may further include a data preprocessing module; The data preprocessing module is used to perform time synchronization and preprocessing on multi-source perioperative data to obtain processed multi-source perioperative data. The deviation prediction module 3 is specifically used to predict deviations based on the processed multi-source perioperative data using the navigation deviation prediction model.

[0060] The bronchoscopy navigation correction system provided in this embodiment may further include a data preprocessing module. After the data acquisition module 1 acquires the multi-source perioperative data at the current sampling time, the data preprocessing module performs time synchronization and preprocessing on the multi-source perioperative data to obtain processed multi-source perioperative data. For example, the ventilator parameters are correlated with the bronchoscope position according to the timestamp, noise and outliers are filtered, and feature quantities (such as the current respiratory phase, cumulative insertion depth, branch turning angle, etc.) are extracted. For CBCT image data, the position coordinates of the lesion and bronchus are extracted through image segmentation. That is, the extracted position coordinates belong to the image coordinate system and are used to calculate the spatial distance deviation between the bronchoscope tip and the actual position of the lesion by comparing with the preoperative CT coordinate system. All preprocessed data will be used as part of the input feature vector of the navigation deviation prediction model. The deviation prediction module 3 uses the navigation deviation prediction model to predict the deviation based on the processed multi-source perioperative data. By performing time synchronization and preprocessing on the multi-source perioperative data, the accuracy of the deviation prediction results is improved.

[0061] In one specific embodiment of this application, the deviation adjustment module 4 is specifically used to maintain the current navigation path and prompt for continued operation when the spatial distance deviation is less than a preset threshold; when the spatial distance deviation exceeds the preset threshold and the deviation confidence meets the requirements, it corrects the lesion target position according to the deviation direction determined by the three-dimensional deviation vector, updates the relative relationship between the bronchoscope tip and the lesion, and recalculates the candidate path; when the deviation confidence is insufficient, it outputs a prompt message to perform intraoperative image review or re-registration.

[0062] After acquiring the three-dimensional deviation vector, spatial distance deviation, and deviation confidence level, the deviation adjustment module 4 maintains the current navigation path and prompts for continued operation when the spatial distance deviation is less than a preset threshold. When the spatial distance deviation exceeds the preset threshold and the deviation confidence level meets the requirements, it corrects the lesion target position based on the deviation direction determined by the three-dimensional deviation vector, updates the relative relationship between the bronchoscope tip and the lesion, and recalculates the candidate path. When the deviation confidence level is insufficient, it outputs a prompt for intraoperative image review or re-registration, thereby realizing path screening and parameter adjustment based on multi-source perioperative data. By determining different processing paths according to the magnitude of the spatial distance deviation and the level of deviation confidence, the system can automatically match the optimal response strategy according to different risk levels, i.e., maintaining operation when the deviation is low, actively correcting deviation when the deviation is reliable, and conservatively verifying when the deviation is unreliable, thus improving the intelligence level and clinical applicability of the navigation system. In one specific embodiment of this application, the system may further include: The prompt message output module is used to output calibration or image confirmation prompt messages when the spatial distance deviation is determined to be less than a preset safety threshold.

[0063] The bronchoscopy navigation and correction system provided in this application embodiment may further include a prompt information output module. When the spatial distance deviation is determined to be less than a preset safety threshold, the prompt information output module outputs calibration or image confirmation prompt information. For example, it may prompt another round of CBCT scans to update the location information, or use an ultrasound probe (r-EBUS) to reconfirm the lesion location, thereby preventing unreliable results due to blind sampling under high deviation conditions.

[0064] Another safety-related recommendation concerns intubation depth control: when the navigation deviation prediction model anticipates a risk of "forward drift" (i.e., the actual lesion location is closer to the bronchoscope than the predicted virtual target), the operator should be advised not to rely entirely on the distance provided by the virtual navigation, but rather to stop earlier, while still some distance from the virtual target (leaving a safety margin), to avoid the catheter tip overshooting the lesion. This "early braking" strategy helps prevent the catheter from being excessively inserted into the extrapleural space or colliding with vital tissues, improving operational safety. The system can automatically calculate a recommended stopping distance based on the prediction, such as "stop when 2mm from the virtual target is recommended," to offset the predicted forward drift error.

[0065] In one specific embodiment of this application, the system may further include: The display module is used to highlight the selected target path through the navigation software interface after path filtering and parameter adjustment based on spatial distance deviation, and to pop up parameter adjustment prompts in the display interface.

[0066] The bronchoscopic navigation and correction system provided in this embodiment may further include a display module. After path filtering and parameter adjustment based on spatial distance deviation, the display module highlights the filtered target path through the navigation software interface and displays parameter adjustment prompts on the display interface. By displaying the filtered target path and parameter adjustment prompts, the operator is provided with effective reference information.

[0067] The entire recommendation mechanism is closely linked to the navigation deviation prediction model, forming a closed-loop control: the navigation deviation prediction model makes real-time predictions → the system generates suggestions → the doctor adjusts the operation or parameters accordingly → the new parameters are fed back to the model to update the prediction. This dynamic cycle continues until the bronchoscope accurately reaches the lesion and completes sampling. Through this error prediction-driven decision support, this application can proactively reduce navigation deviation. Operators receive early warnings and optimization plans before significant deviations occur, thus adjusting their strategies in a timely manner and preventing potential positioning failures. This proactive recommendation method is significantly different from the traditional model where operators rely entirely on experience or postoperative image correction. It maximizes the use of intraoperative information to improve positioning accuracy and enhances navigation reliability.

[0068] In one specific embodiment of this application, the deviation prediction module 3 includes: The data input unit is used to receive multi-source perioperative data sent by the data acquisition module 1, and to perform structured processing on the multi-source perioperative data to obtain structured numerical features. The feature encoding unit is used to receive structured numerical features sent by the data input unit, normalize the structured numerical features, and extract path numerical features, pose numerical features, device status numerical features and respiratory parameter numerical features through a fully connected network. The extracted numerical features are then output to the feature fusion unit. The temporal modeling unit is used to receive airway motion data, endoscope pose change data and navigation error change data in the continuous respiratory cycle from the multi-source perioperative data provided by the data input unit, and input them into the long short-term memory network or temporal convolutional network according to time steps. The corresponding temporal feature vector is output through the hidden state of the network, and the temporal feature vector is output to the feature fusion unit. The image feature extraction unit is used to receive intraoperative fluoroscopic images, CBCT local reconstruction images or registration residual maps from the multi-source perioperative data provided by the data input unit, and to extract lesion area features, catheter tip area features and their relative spatial relationship features using a convolutional neural network, and output the extracted image features to the feature fusion unit. The feature fusion unit is used to stitch together and fuse numerical features, temporal feature vectors and image features, and output the fused features to the bias regression unit; The deviation regression unit is used to output the spatial distance deviation between the bronchoscope tip and the actual location of the lesion based on the fusion characteristics.

[0069] The deviation prediction module 3 includes a data input unit, a feature encoding unit, a temporal modeling unit, an image feature extraction unit, a feature fusion unit, and a deviation regression unit. The data input unit receives multi-source perioperative data from the data acquisition module 1 and performs structured processing on the multi-source perioperative data to obtain structured numerical features. The feature encoding unit receives the structured numerical features from the data input unit, normalizes the structured numerical features, and extracts path numerical features, pose numerical features, instrument status numerical features, and respiratory parameter numerical features through a fully connected network. The extracted numerical features are then output to the feature fusion unit. The temporal modeling unit receives airway motion data, endoscope pose change data, and navigation error change data within continuous respiratory cycles from the multi-source perioperative data provided by the data input unit. It inputs these data into a long short-term memory network or a temporal convolutional network at time steps, outputs the corresponding temporal feature vectors through the network's hidden states, and then outputs the temporal feature vectors to the feature fusion unit. The image feature extraction unit receives intraoperative fluoroscopic images, CBCT local reconstruction images, or registration residual maps from multi-source perioperative data provided by the data input unit. It then uses a convolutional neural network to extract lesion region features, catheter tip region features, and their relative spatial relationships. The extracted image features are output to the feature fusion unit. The feature fusion unit concatenates and fuses numerical features, temporal feature vectors, and image features, and outputs the fused features to the deviation regression unit. The deviation regression unit outputs the spatial distance deviation between the bronchoscope tip and the actual position of the lesion based on the fused features. Through the refined design of the internal structure of the deviation prediction module 3, in-depth utilization and complementary fusion of multi-source perioperative data are achieved, enhancing the accuracy, real-time performance, and clinical applicability of navigation deviation prediction, facilitating subsequent intelligent correction.

[0070] Multi-source perioperative data can include preoperative planning pathway, real-time bronchoscope position, relative distance from the bronchoscope tip to the lesion target, pathway curvature, bifurcation angle, catheter bending status, intraoperative fluoroscopy or CBCT image registration results, ventilator parameter curves, airway pressure, tidal volume, positive end-expiratory pressure, respiratory cycle phase, and actual deviation information during historical operations.

[0071] Corresponding to the above method embodiments, this application also provides a bronchoscopy navigation correction method. The bronchoscopy navigation correction method described below can be referred to in correspondence with the bronchoscopy navigation correction system described above.

[0072] See Figure 2 , Figure 2 This is a flowchart illustrating an implementation method for bronchoscopic navigation correction in this application. The method may include the following steps: S201: Obtain multi-source perioperative data at the current sampling time.

[0073] The multi-source perioperative data includes intraoperative ventilator parameters, intraoperative imaging data, and navigation system data.

[0074] S202: When the tip of the bronchoscope reaches the preset distance range of the actual location of the lesion, the navigation deviation prediction model is used to find the phase with the smallest deviation in the respiratory cycle based on the ventilator parameters, and the phase with the smallest deviation is output so that the lesion can be treated at the phase with the smallest deviation.

[0075] S203: Using a navigation deviation prediction model, deviation prediction is performed based on multi-source perioperative data to obtain the three-dimensional deviation vector, spatial distance deviation, and deviation confidence level of the bronchoscope tip relative to the actual position of the lesion.

[0076] S204: Path selection and parameter adjustment are performed based on the three-dimensional deviation vector, spatial distance deviation, deviation confidence, and the time phase of minimum deviation in order to perform navigation correction.

[0077] As can be seen from the above technical solution, by utilizing multi-source data during the surgical procedure, a navigation deviation prediction model predicts in real time the spatial distance deviation between the bronchoscope tip and the actual position of the lesion, and provides optimized navigation path recommendations and parameter adjustment suggestions based on the prediction results. This allows for dynamic correction of the navigation strategy without interrupting the surgical procedure, transforming passive intraoperative calibration into proactive prevention. It significantly improves the accuracy and efficiency of bronchoscopic navigation positioning, solves the problem of inaccurate positioning caused by respiratory displacement and anatomical changes, improves the biopsy hit rate of small lung lesions, reduces the number of additional image corrections and radiation exposure, and enhances the safety and success rate of the surgery.

[0078] In one specific embodiment of this application, path filtering based on spatial distance deviation may include the following steps: Step 1: Obtain the simulation deviations corresponding to each feasible path obtained from the preoperative simulation. Step 2: Determine the airway condition from each feasible path to the actual location of the lesion based on the spatial distance deviation; Step 3: Select a path from the feasible paths based on the conditions of each airway and each simulation deviation; where each simulation deviation is the deviation obtained by simulation using a navigation deviation prediction model based on historical data in the database.

[0079] In one specific embodiment of this application, adjusting parameters based on spatial distance deviation may include the following steps: When the predicted result corresponding to the spatial distance deviation is that the tidal volume is too low and the lung volume is insufficient, resulting in the actual location of the lesion shifting, an upward prompt message for tidal volume or positive end-expiratory pressure is output.

[0080] In one specific embodiment of this application, adjusting parameters based on spatial distance deviation may include the following steps: When the prediction result corresponding to the spatial distance deviation is that the lungs are overinflated due to excessive tidal volume, resulting in inaccurate positioning, a tidal volume reduction prompt message is output.

[0081] In one specific embodiment of this application, the method may further include the following steps: Time synchronization and preprocessing of multi-source perioperative data were performed to obtain processed multi-source perioperative data. Using a navigation deviation prediction model to predict deviations based on multi-source perioperative data may include the following steps: The navigation deviation prediction model is used to predict deviations based on processed multi-source perioperative data.

[0082] In one specific embodiment of this application, the method may further include the following steps: When the spatial distance deviation is determined to be less than the preset safety threshold, a calibration or image confirmation prompt message is output.

[0083] In one specific embodiment of this application, after path filtering and parameter adjustment based on spatial distance deviation, the method may further include the following steps: The target path is highlighted and filtered through the navigation software interface, and a parameter adjustment prompt message pops up in the display interface.

[0084] Corresponding to the above method embodiments, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the following steps: The system acquires multi-source perioperative data at the current sampling time, including intraoperative ventilator parameters, intraoperative imaging data, and navigation system data. When the bronchoscope tip reaches a preset distance range from the actual location of the lesion, the navigation deviation prediction model is used to find the phase with the smallest deviation in the respiratory cycle based on the ventilator parameters, and outputs the phase with the smallest deviation so that lesion treatment can be performed at the phase with the smallest deviation. The navigation deviation prediction model is used to predict deviations based on the multi-source perioperative data, obtaining the three-dimensional deviation vector, spatial distance deviation, and deviation confidence level of the bronchoscope tip relative to the actual location of the lesion. Path selection and parameter adjustment are performed based on the three-dimensional deviation vector, spatial distance deviation, deviation confidence level, and phase with the smallest deviation to perform navigation correction.

[0085] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0086] For a description of the computer-readable storage medium provided in this application, please refer to the above method embodiments; further details will not be repeated here.

[0087] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatuses, devices, and computer-readable storage media disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0088] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the technical solutions and core ideas of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A bronchoscopic navigation and correction system, characterized in that, include: The data acquisition module (1) is used to acquire multi-source perioperative data at the current sampling time; wherein, the multi-source perioperative data includes intraoperative ventilator parameters, intraoperative imaging data and navigation system data; The minimum deviation phase search module (2) is used to find the minimum deviation phase in the respiratory cycle by using the navigation deviation prediction model according to the ventilator parameters when the tip of the bronchoscope reaches the preset distance range of the actual location of the lesion, and output the minimum deviation phase so as to perform lesion treatment in the minimum deviation phase; wherein, in the respiratory cycle optimization scenario, the navigation deviation prediction model outputs the recommended respiratory phase or time window with the minimum deviation according to the predicted deviation under different respiratory phases; The deviation prediction module (3) is used to use the navigation deviation prediction model to predict deviations based on the multi-source perioperative data, and obtain the three-dimensional deviation vector, spatial distance deviation, and deviation confidence of the bronchoscope tip relative to the actual position of the lesion; wherein, in the real-time navigation correction scenario, the navigation deviation prediction model outputs the three-dimensional deviation vector, spatial distance deviation, and deviation confidence of the bronchoscope tip relative to the actual position of the lesion, which is used to indicate the current navigation error and the correction direction; The deviation adjustment module (4) is used to perform path selection and parameter adjustment based on the three-dimensional deviation vector, the spatial distance deviation, the deviation confidence level, and the deviation minimum time phase, so as to perform navigation correction; wherein, in the path adjustment scenario, the navigation deviation prediction model outputs the deviation risk level of the current path, recommended path branches, mirror propulsion depth adjustment amount, end attitude adjustment direction, and operation parameter correction suggestions for the deviation adjustment module to perform path selection and navigation parameter update; Specifically, the deviation adjustment module (4) is used to maintain the current navigation path and prompt for continued operation when the spatial distance deviation is less than a preset threshold; when the spatial distance deviation exceeds the preset threshold and the deviation confidence meets the requirements, it corrects the lesion target position according to the deviation direction determined by the three-dimensional deviation vector, updates the relative relationship between the bronchoscope tip and the lesion, and recalculates the candidate path; when the deviation confidence is insufficient, it outputs a prompt message to perform intraoperative image review or re-registration.

2. The bronchoscopic navigation and correction system according to claim 1, characterized in that, The deviation adjustment module (4) is specifically used to obtain the simulation deviations corresponding to each feasible path obtained from the preoperative simulation; determine the airway status from each feasible path to the actual location of the lesion based on the spatial distance deviation; and filter the paths from each feasible path based on the airway status and each simulation deviation. Among them, each simulation deviation is a deviation obtained by simulation based on historical data in the database using the navigation deviation prediction model.

3. The bronchoscopic navigation and correction system according to claim 1, characterized in that, The deviation adjustment module (4) is specifically used to output a prompt message to increase tidal volume or positive end-expiratory pressure when the predicted result corresponding to the spatial distance deviation is that the tidal volume is too low and the lung volume is insufficient, resulting in the actual position of the lesion shifting.

4. The bronchoscopic navigation and correction system according to claim 1, characterized in that, The deviation adjustment module (4) is specifically used to output a tidal volume reduction prompt when the prediction result corresponding to the spatial distance deviation is that the tidal volume is too high, causing the lungs to overinflate and resulting in inaccurate positioning.

5. The bronchoscopic navigation and correction system according to claim 1, characterized in that, It also includes a data preprocessing module; The data preprocessing module is used to perform time synchronization and preprocessing on the multi-source perioperative data to obtain processed multi-source perioperative data. The deviation prediction module (3) is specifically used to predict deviations based on the processed multi-source perioperative data using the navigation deviation prediction model.

6. The bronchoscopic navigation and correction system according to claim 1, characterized in that, Also includes: The prompt information output module is used to output calibration or image confirmation prompt information when it is determined that the spatial distance deviation is greater than a preset threshold.

7. The bronchoscopic navigation and correction system according to claim 1, characterized in that, Also includes: The display module is used to highlight the selected target path through the navigation software interface after path filtering and parameter adjustment based on the spatial distance deviation, and to pop up parameter adjustment prompts in the display interface.

8. The bronchoscopic navigation and correction system according to any one of claims 1 to 7, characterized in that, The deviation prediction module (3) includes: The data input unit is used to receive the multi-source perioperative data sent by the data acquisition module (1) and perform structured processing on the multi-source perioperative data to obtain structured numerical features. The feature encoding unit is used to receive the structured numerical features sent by the data input unit, normalize the structured numerical features, and extract path numerical features, pose numerical features, device status numerical features and respiratory parameter numerical features through a fully connected network, and output the extracted numerical features to the feature fusion unit. The temporal modeling unit is used to receive airway motion data, endoscope pose change data and navigation error change data in the continuous respiratory cycle of the multi-source perioperative data provided by the data input unit, and input them into the long short-term memory network or temporal convolutional network according to time steps, output the corresponding temporal feature vector through the hidden state of the network, and output the temporal feature vector to the feature fusion unit. The image feature extraction unit is used to receive intraoperative fluoroscopic images, CBCT local reconstruction images or registration residual maps from the multi-source perioperative data provided by the data input unit, and to extract lesion area features, catheter tip area features and their relative spatial relationship features using a convolutional neural network, and output the extracted image features to the feature fusion unit. The feature fusion unit is used to stitch and fuse the numerical features, the temporal feature vector and the image features, and output the fused features to the deviation regression unit. The deviation regression unit is used to output the spatial distance deviation between the bronchoscope tip and the actual position of the lesion based on the fusion features.

9. A bronchoscopic navigation and correction method, characterized in that, include: Acquire multi-source perioperative data at the current sampling time; wherein, the multi-source perioperative data includes intraoperative ventilator parameters, intraoperative imaging data, and navigation system data; When the tip of the bronchoscope reaches the preset distance range of the actual location of the lesion, the navigation deviation prediction model is used to find the phase with the smallest deviation in the respiratory cycle according to the ventilator parameters, and outputs the phase with the smallest deviation, so that the lesion can be treated in the phase with the smallest deviation; wherein, in the respiratory cycle optimization scenario, the navigation deviation prediction model outputs the recommended respiratory phase or time window with the smallest deviation according to the predicted deviation under different respiratory phases. The navigation deviation prediction model is used to predict deviations based on the multi-source perioperative data, and obtains the three-dimensional deviation vector, spatial distance deviation, and deviation confidence level of the bronchoscope tip relative to the actual position of the lesion. In real-time navigation correction scenarios, the navigation deviation prediction model outputs the three-dimensional deviation vector, spatial distance deviation, and deviation confidence level of the bronchoscope tip relative to the actual position of the lesion to indicate the current navigation error and the direction of correction. Based on the three-dimensional deviation vector, the spatial distance deviation, the deviation confidence level, and the time phase of minimum deviation, path selection and parameter adjustment are performed to correct navigation deviations. In the path adjustment scenario, the navigation deviation prediction model outputs the deviation risk level of the current path, recommended path branches, mirror advance depth adjustment amount, end attitude adjustment direction, and operation parameter correction suggestions for path selection and navigation parameter updates. The path selection and parameter adjustment based on the three-dimensional deviation vector, the spatial distance deviation, the deviation confidence level, and the time phase with the minimum deviation include: When the spatial distance deviation is less than a preset threshold, the current navigation path is maintained and a prompt to continue operation is displayed; when the spatial distance deviation exceeds the preset threshold and the deviation confidence level meets the requirements, the lesion target position is corrected according to the deviation direction determined by the three-dimensional deviation vector, the relative relationship between the bronchoscope tip and the lesion is updated, and the candidate path is recalculated; when the deviation confidence level is insufficient, a prompt message for intraoperative image review or re-registration is output.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the bronchoscopic navigation correction method as described in claim 9.

Images