A Mixed Reality-Based Intelligent Navigation Training Method and System for Bladder Irrigation

Abstract

This invention relates to the field of medical training technology, and more particularly to a mixed reality-based intelligent navigation training method and system for bladder irrigation. The method acquires catheter manipulation trajectory and eye-tracking data using mixed reality devices, calculates spatial deviations based on anatomical models and path planning, assesses attention allocation using eye-tracking mapping, generates navigation guidance with inverse adjustment based on detail and attention scores, and identifies skill deficiency types based on causal inference, adjusting the training difficulty accordingly. This method effectively improves the standardization and personalization of bladder irrigation operation training.

Description

Technical Field

[0001] This invention relates to the field of medical training technology, and in particular to a mixed reality-based intelligent navigation training method and system for bladder irrigation. Background Technology

[0002] In the field of bladder irrigation skills training, traditional teaching methods mainly rely on instructor demonstrations, simulated mannequin practice, and post-operative video playback for error correction. Trainees typically need to repeatedly operate on a standard silicone model, while instructors point out deviations based on subjective experience. Some training institutions have introduced immersive training systems based on virtual reality (VR). These systems can record operation trajectories and calculate errors between them and preset paths, but the feedback provided is mostly delayed numerical reports, unable to dynamically guide trainees to adjust their techniques during the operation. In addition, existing VR systems mostly use fixed difficulty settings, lack monitoring of trainees' attention allocation, and adjustments to the training scenario rely entirely on the instructor's subjective judgment.

[0003] Existing technologies suffer from two main drawbacks. Conventional training methods cannot provide adaptive navigation guidance based on trainees' real-time performance and cognitive state. When trainees encounter operational difficulties, they often resort to mechanically repeating incorrect actions, while the standard guidance provided by the system lacks specificity—when trainees are distracted due to anxiety, overly detailed navigation prompts can actually exacerbate cognitive load; conversely, when trainees have basically mastered the operational skills, redundant guidance can limit the development of their self-correction abilities. This "one-size-fits-all" feedback model severely restricts training efficiency.

[0004] Another drawback lies in the crudeness of skill defect diagnosis. Existing systems typically only count the total deviation of the operation trajectory from the standard path or the completion time, judging whether a trainee is "qualified" based on a simple threshold. This evaluation method cannot distinguish the root cause of the deviation—whether it is an error in anatomical structure identification, an improper operating angle, or insufficient hand-eye coordination. For example, a trainee's repeated problem of inserting catheters too deeply may be due to a fuzzy memory of the anatomical location of the bladder neck, or it may be a mechanical error caused by a lack of focus on the critical access area. The lack of causal identification of skill defect types renders subsequent targeted training ineffective, resulting in widespread waste of training resources. Summary of the Invention

[0005] The present invention provides a mixed reality-based intelligent navigation training method and system for bladder irrigation, which can solve the problems in the prior art.

[0006] A first aspect of the present invention provides a mixed reality-based intelligent navigation training method for bladder irrigation, comprising:

[0007] Acquire catheter manipulation trajectory and eye movement data of trainees during bladder irrigation training using mixed reality devices;

[0008] Based on the anatomical model of the bladder-urethra, the optimal catheter insertion path is generated through a path planning algorithm, and the spatial deviation sequence between the catheter operation trajectory and the optimal path is calculated.

[0009] The eye-tracking data is mapped to the anatomical structure region of the gaze identified by the anatomical model. Matching events and gaze duration between the gaze anatomical structure region and the target operation region set are extracted. An attention allocation score is calculated based on the matching events and gaze duration.

[0010] Based on the spatial deviation sequence and the attention allocation score, bladder irrigation operation navigation guidance information is generated, and the level of detail of the navigation guidance information is inversely modulated by the attention allocation score.

[0011] Using the time series of operational error events, the spatial deviation sequence, and the attention allocation score as observation data, a structural causal graph is constructed. Latent variables with the greatest causal effect on the operational error events are identified through causal inference and mapped to the bladder irrigation operation skill defect type.

[0012] The training scenario difficulty parameters are adjusted based on the statistical characteristics of the skill deficiency type and the spatial deviation sequence.

[0013] Based on the bladder-urethra anatomical model, an optimal catheter insertion path is generated using a path planning algorithm. The spatial deviation sequence between the catheter manipulation trajectory and the optimal path is calculated, including:

[0014] Based on the aforementioned anatomical model, the urethra is divided into the external urethral orifice region, the bulbous urethra region, the membranous urethra region, and the bladder neck region;

[0015] For each anatomical region, a region-specific path planning strategy is constructed. The path planning strategy includes tissue safety constraint weight coefficients and path efficiency constraint weight coefficients. The tissue safety constraint weight coefficients for the bulbous urethra and membranous urethra are higher than those for the external urethral orifice and bladder neck regions.

[0016] The catheter manipulation trajectory is divided into multiple regional trajectory segments. For each regional trajectory segment, the minimum safe distance threshold between the path and the urethral wall is determined according to the tissue safety constraint weight coefficient of the corresponding region. The trade-off ratio between path length and path curvature is determined according to the path efficiency constraint weight coefficient. The optimal path segment for the region is calculated based on the minimum safe distance threshold and the trade-off ratio as constraints.

[0017] Connect all optimal path segments in the region according to the anatomical order of the urethra and smooth the junctions to generate the optimal catheter insertion path;

[0018] Align the catheter manipulation trajectory with the optimal catheter insertion path in the time dimension and calculate the spatial deviation sequence.

[0019] The eye-tracking data is mapped to the anatomical structure region of the gaze identified by the anatomical model. Matching events and gaze durations between the gaze anatomical structure region and the target operation region set are extracted. An attention allocation score is calculated based on the matching events and gaze durations, including:

[0020] Extract the two-dimensional coordinate sequence, timestamp sequence, and pupil diameter sequence of the gaze fixation point from the eye-tracking data; obtain the viewpoint position parameters and viewpoint direction parameters of the mixed reality device; establish the coordinate transformation relationship from the screen coordinate system to the anatomical model space coordinate system; and map the two-dimensional coordinate sequence into a set of three-dimensional gaze rays.

[0021] Calculate the coordinates of the intersection points between the three-dimensional gaze ray set and the surfaces of each anatomical structure in the anatomical model, and use the anatomical structure identifiers corresponding to the intersection coordinates as the gaze anatomical structure regions;

[0022] When the gaze anatomical region belongs to the corresponding predefined set of target operation regions, it is marked as a target region gaze event;

[0023] Based on the timestamp sequence, the gaze duration for each gaze event in the target region is calculated. The pupil diameter sequence corresponding to the gaze event is extracted to calculate the pupil diameter change rate. The gaze duration and the pupil diameter change rate are weighted and fused to obtain the effective attention weight.

[0024] The ratio of the cumulative effective attention weights of all gaze events in the target region to the total training time is taken as the effective attention percentage of the target region. The effective attention percentage is then weighted and combined with the frequency of gaze events in the target region to calculate the attention allocation score.

[0025] Bladder irrigation operation navigation guidance information is generated based on the spatial deviation sequence and the attention allocation score. The level of detail in the navigation guidance information is inversely modulated by the attention allocation score, including:

[0026] Calculate the spatial deviation mean and spatial deviation variance for the spatial deviation sequence; identify key deviation moments based on the spatial deviation mean and spatial deviation variance, where the spatial deviation value corresponding to the key deviation moment exceeds the sum of the spatial deviation mean and spatial deviation variance, and extract the anatomical structure region identifier corresponding to the key deviation moment;

[0027] The attention allocation score is compared with a preset attention threshold to classify low attention states and high attention states.

[0028] For low-attention states, detailed navigation guidance information is generated, including the anatomical structure region identifier, safe operation constraints for the anatomical structure region, and path correction guidance direction; for high-attention states, concise navigation guidance information is generated, including the anatomical structure region identifier.

[0029] The detailed or concise navigation guidance information is presented to the trainee through the mixed reality device.

[0030] Using operational error events, the spatial deviation sequence, and the time series of attention allocation scores as observational data, a structural causal graph is constructed. Causal inference identifies latent variables with the greatest causal effect on the operational error events, mapping them to bladder irrigation operation skill deficiency types, including:

[0031] The time sequence of operational error events, the spatial deviation sequence, and the time sequence of attention allocation scores are aligned along a unified time axis to form a multidimensional dataset;

[0032] A structural causal graph is constructed based on the aforementioned multidimensional dataset. The nodes in the structural causal graph include observed variable nodes and latent variable nodes. When there is a conditional dependency between two observed variable nodes that cannot be explained by the existing observed variable nodes, the latent variable node is introduced as a common cause.

[0033] Calculate the total causal effect strength of the latent variable nodes in the causal graph on the operational error event;

[0034] The latent variable node with the strongest overall causal effect is identified as the dominant latent variable. The causal path pattern of the dominant latent variable is mapped to the type of bladder irrigation operation skill deficiency. When the causal path of the dominant latent variable is transmitted through the spatial bias sequence, it is mapped to a path planning skill deficiency. When the causal path of the dominant latent variable is transmitted through the attention allocation score, it is mapped to an attention allocation skill deficiency. When the causal path of the dominant latent variable is transmitted through both the spatial bias sequence and the attention allocation score, it is mapped to a coordination and integration skill deficiency.

[0035] Calculating the total causal effect strength of the latent variable nodes in the structural causal graph on the operational error event includes:

[0036] Extract all latent variable nodes that have a causal path to the operation error event node from the structural causal graph. The causal path is a path that starts from a latent variable node, passes through one or more directed edges, and finally points to the operation error event node.

[0037] The total causal effect strength is obtained by summing the causal effects of the latent variable node on the operational error event via different causal paths. The causal effect of each causal path is calculated by multiplying the causal strength coefficients of each directed edge on the path. The causal strength coefficients are determined based on the mutual information and time lag correlation of the corresponding observed variables in the multidimensional dataset.

[0038] Adjusting the training scenario difficulty parameters based on the statistical characteristics of the skill deficiency type and the spatial deviation sequence includes:

[0039] To address deficiencies in path planning skills, the anatomical regions where spatial deviation peaks occur in the spatial deviation sequence are extracted. The density of path planning guidance cues in these anatomical regions is increased, and the catheter manipulation speed limit threshold in these regions is reduced to extend the trainee's operational decision-making time.

[0040] To address the deficiencies in attention allocation skills, the number of visual distracting elements presented in non-target operation areas is reduced, the visual salience of target operation areas is enhanced, and phased attention checkpoints are introduced into the training scenario.

[0041] To address the deficiencies in coordination and integration skills, the fluctuation frequency of the spatial deviation sequence is calculated. When the fluctuation frequency exceeds a preset frequency threshold, it is determined to be an unstable operational state. In response to this state, the density of path guidance prompts and the salience of visual annotations are adjusted synchronously, and an operational rhythm prompt signal is set in the training scenario to help trainees establish a stable operational rhythm pattern.

[0042] A second aspect of the present invention provides a bladder irrigation intelligent navigation training system based on mixed reality, comprising:

[0043] The data acquisition unit is used to acquire catheter manipulation trajectory and eye movement data of trainees during bladder irrigation training using mixed reality devices;

[0044] The path analysis unit is used to generate the optimal catheter insertion path based on the bladder-urethra anatomical model and to calculate the spatial deviation sequence between the catheter operation trajectory and the optimal path through the path planning algorithm.

[0045] The attention assessment unit is used to map the eye movement data to the anatomical structure region of the gaze identified by the anatomical model, extract the matching events and gaze duration between the gaze anatomical structure region and the target operation region set, and calculate the attention allocation score based on the matching events and gaze duration.

[0046] A navigation generation unit is used to generate bladder irrigation operation navigation guidance information based on the spatial deviation sequence and the attention allocation score, wherein the level of detail of the navigation guidance information is inversely modulated by the attention allocation score.

[0047] The causal analysis unit is used to construct a structural causal graph using the time series of the operational error event, the spatial deviation sequence, and the attention allocation score as observation data. It identifies the latent variable with the greatest causal effect on the operational error event through causal inference and maps it to the bladder irrigation operation skill defect type.

[0048] The difficulty adjustment unit is used to adjust the training scenario difficulty parameters based on the skill defect type and the statistical characteristics of the spatial deviation sequence.

[0049] A third aspect of the present invention provides an electronic device, comprising:

[0050] processor;

[0051] Memory used to store processor-executable instructions;

[0052] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0053] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0054] This invention, based on mixed reality technology, enables the simultaneous acquisition and analysis of operational trajectories and eye-tracking data in bladder irrigation training, significantly enhancing the immersion and realism of the training. Through anatomical models and path planning algorithms, operational deviations are precisely quantified, and attention allocation is scored using eye-tracking data. This allows for reverse adjustment of the level of detail in navigation guidance, avoiding over-reliance on prompts, strengthening the trainee's active learning ability, and shortening the learning curve from novice to proficient operator.

[0055] By constructing a structural causal graph using causal inference, latent variables with the greatest causal effect on operational errors are identified from time series of operational error events, spatial deviation sequences, and attention scores. These variables are accurately mapped to skill deficit types such as insufficient hand-eye coordination or anatomical cognitive biases. This mechanism overcomes the limitation of traditional training that only assesses the final outcome, enabling precise diagnosis of the underlying causes of operational errors and providing a data foundation for targeted interventions.

[0056] Based on the statistical characteristics of skill deficit types and spatial deviations, the difficulty parameters of training scenarios are dynamically adjusted. Adaptive difficulty adjustment avoids the monotony of repetitive training, ensures that training challenges match the trainees' ability levels, maximizes skill improvement efficiency within limited training time, and ultimately significantly improves the success rate and safety of bladder irrigation procedures. Attached Figure Description

[0057] Figure 1A flowchart illustrating a mixed reality-based intelligent navigation training method for bladder irrigation;

[0058] Figure 2 A flowchart illustrating the process for identifying bladder irrigation technique skill deficiencies through causal inference. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0061] Figure 1 This is a flowchart illustrating the intelligent navigation training method for bladder irrigation based on mixed reality, as described in an embodiment of the present invention.

[0062] The mixed reality-based intelligent navigation training method for bladder irrigation includes:

[0063] Acquire catheter manipulation trajectory and eye movement data of trainees during bladder irrigation training using mixed reality devices;

[0064] Based on the anatomical model of the bladder-urethra, the optimal catheter insertion path is generated through a path planning algorithm, and the spatial deviation sequence between the catheter operation trajectory and the optimal path is calculated.

[0065] The eye-tracking data is mapped to the anatomical structure region of the gaze identified by the anatomical model. Matching events and gaze duration between the gaze anatomical structure region and the target operation region set are extracted. An attention allocation score is calculated based on the matching events and gaze duration.

[0066] Bladder irrigation operation navigation guidance information is generated based on the spatial deviation sequence and the attention allocation score, wherein the level of detail of the navigation guidance information is inversely modulated by the attention allocation score.

[0067] Using the time series of operational error events, the spatial deviation sequence, and the attention allocation score as observation data, a structural causal graph is constructed. Latent variables with the greatest causal effect on the operational error events are identified through causal inference and mapped to the bladder irrigation operation skill defect type.

[0068] The training scenario difficulty parameters are adjusted based on the statistical characteristics of the skill deficiency type and the spatial deviation sequence.

[0069] In one optional implementation, based on a bladder-urethra anatomical model, an optimal catheter insertion path is generated using a path planning algorithm, and the spatial deviation sequence between the catheter manipulation trajectory and the optimal path is calculated, including:

[0070] Acquire catheter manipulation trajectory and eye movement data of trainees during bladder irrigation training using mixed reality devices;

[0071] Based on the aforementioned anatomical model, the urethra is divided into the external urethral orifice region, the bulbous urethra region, the membranous urethra region, and the bladder neck region;

[0072] For each anatomical region, a region-specific path planning strategy is constructed. The path planning strategy includes tissue safety constraint weight coefficients and path efficiency constraint weight coefficients. The tissue safety constraint weight coefficients for the bulbous urethra and membranous urethra are higher than those for the external urethral orifice and bladder neck regions.

[0073] The catheter manipulation trajectory is divided into multiple regional trajectory segments. For each regional trajectory segment, the minimum safe distance threshold between the path and the urethral wall is determined according to the tissue safety constraint weight coefficient of the corresponding region. The trade-off ratio between path length and path curvature is determined according to the path efficiency constraint weight coefficient. The optimal path segment for the region is calculated based on the minimum safe distance threshold and the trade-off ratio as constraints.

[0074] Connect all optimal path segments in the region according to the anatomical order of the urethra and smooth the junctions to generate the optimal catheter insertion path;

[0075] Align the catheter manipulation trajectory with the optimal catheter insertion path in the time dimension and calculate the spatial deviation sequence.

[0076] For example, in bladder irrigation training, the catheter needs to pass sequentially through the external urethral orifice, bulbous urethra, membranous urethra, and bladder neck to reach the bladder. The tissue characteristics and risk of injury differ significantly among these anatomical structures. The bulbous and membranous urethra are rich in blood vessels and have thin walls, making them highly susceptible to bleeding or tearing if handled improperly. In contrast, the external urethral orifice and bladder neck have relatively stronger tissue tolerance and a higher tolerance for path deviations. Based on these anatomical characteristics, the urethra is spatially segmented into these four functional regions in the three-dimensional anatomical model presented by the mixed reality device. This allows subsequent path planning to adopt differentiated constraint strategies based on the physiological risk differences in different regions.

[0077] At the sensing level of the mixed reality device, electromagnetic tracking markers attached to the end of the conduit continuously report three-dimensional coordinates at a sampling frequency of no less than 50Hz. The collected conduit operation trajectory is stored as a spatial coordinate sequence in a timestamp-aligned manner. At the same time, the eye-tracking module built into the head-mounted mixed reality device synchronously collects the trainee's gaze direction vector. Both types of data are recorded with a unified time reference, providing a basis for subsequent trajectory segmentation and deviation calculation.

[0078] For each anatomical region, a region-specific path planning strategy is constructed. The planning strategy for each region is characterized by two types of weighting coefficients: tissue safety constraint weighting coefficients. and path efficiency constraint weight coefficient Organizational safety constraint weighting coefficients are used to quantify the importance of the safe distance between the path and the organizational wall, while path efficiency constraint weighting coefficients are used to quantify the importance of path length and curvature optimization. Both satisfy... The trade-off between safety and operational efficiency is balanced by adjusting the ratio of the two. For the bulbous urethra and membranous urethra, A higher value (range 0.7 to 0.85) means that the pathway planning prioritizes ensuring a safe distance between the catheter and the urethral wall. Even if the path curvature increases or the path length increases, it is necessary to ensure that the tissue is not compressed and damaged. For the urethral orifice area and the bladder neck area, Using lower values ​​(ranging from 0.4 to 0.55) allows the path to prioritize efficiency while maintaining a basic safe distance, reducing unnecessary detours. The weight coefficients for each region are pre-calibrated based on clinical operating procedures and the biomechanical characteristics of anatomical tissues, and are loaded into the path planning module before training begins.

[0079] The catheter manipulation trajectory is divided according to the spatial boundaries of the anatomical regions, resulting in trajectory segments corresponding one-to-one with the four anatomical regions. For each trajectory segment, based on the region... Determine the minimum safe distance threshold between the path and the urethral wall. . Specifically, and There is a positive correlation: The higher, The larger the urethra, the more pronounced the condition in the bulbous and membranous urethral regions. It is typically set at 15% to 20% of the urethral diameter; in the external urethral orifice and bladder neck region, It can be reduced to 8% to 12% of the urethral diameter. Path efficiency constraint weighting coefficient. This determines the trade-off between path length and path curvature: At higher levels, the optimization objective tends to shorten the total path length and reduce curvature; At lower levels, the optimized target allows for an appropriate increase in curvature of the path to maintain a safe distance.

[0080] Under the above constraints, a gradient-based path optimization method is used to calculate the optimal path segment for each region's trajectory segment. Using the anatomical model mesh within the region as environmental constraints, the path is represented as an ordered set of control points. Iterative optimization is then performed to ensure the path satisfies... Minimize the weighted cost function under constraints ,in , This is the total path length. Insufficient minimum distance between the path and the urethral wall The cumulative penalty value over time. The control point coordinates are updated in each iteration until the cost function converges or the number of iterations reaches a preset upper limit, at which point the optimal path segment for that region is output.

[0081] The optimal path segments of the four regions are connected end to end in the order of urethral anatomy. At the junction of adjacent regions, cubic B-spline interpolation is used for smoothing to eliminate abrupt changes in path direction caused by differences in planning strategies in different regions. This ensures that the final optimal catheter insertion path has a continuous tangential direction and bounded curvature throughout the entire process, avoiding acute angle inflection points at the junctions, thereby ensuring that the path has physical feasibility for guiding catheter movement.

[0082] When aligning the catheter manipulation trajectory with the optimal catheter insertion path in the time dimension, the moment the catheter enters the urethral orifice is taken as the time origin of both paths. A dynamic time warping algorithm is used to align the time series of the catheter manipulation trajectory with the arc length parameter series of the optimal path, obtaining the three-dimensional Euclidean distance between the manipulation position at each moment and the corresponding reference point on the optimal path, thus forming a spatial deviation sequence. ,in Indicates the first The spatial distance between the actual position of the catheter and the optimal path reference point at each moment was calculated. The deviation sequence was segmented and statistically analyzed by anatomical region, and the mean deviation, maximum deviation, and duration of deviation exceeding the threshold were calculated for each region. This provides a refined quantitative basis for subsequent navigation guidance information generation and skill defect identification.

[0083] In actual training, when the trainee's mean deviation sequence in the bulbourethral region exceeds that region... When the accuracy reaches 50%, the mixed reality device will overlay visual warning labels on the corresponding anatomical locations to help trainees perceive operational deviations in a timely manner and make corrections, thereby effectively transforming the regionally differentiated path planning results into real-time operational feedback signals.

[0084] In one optional implementation, the eye-tracking data is mapped to the anatomical structure region of the gaze identified by the anatomical model, and matching events and gaze durations between the gaze anatomical structure region and the target operation region set are extracted. An attention allocation score is calculated based on the matching events and gaze durations, including:

[0085] The eye-tracking data is used to extract the two-dimensional coordinate sequence, timestamp sequence, and pupil diameter sequence of the gaze fixation point. The viewpoint position and viewpoint direction parameters of the mixed reality device are obtained, and a coordinate transformation relationship is established from the screen coordinate system to the anatomical model spatial coordinate system. The two-dimensional coordinate sequence is mapped to a three-dimensional gaze ray set. The coordinates of the intersection points between the three-dimensional gaze ray set and the surfaces of each anatomical structure in the anatomical model are calculated, and the anatomical structure identifiers corresponding to the intersection coordinates are used as gaze anatomical structure regions. When a gaze anatomical structure region belongs to a corresponding predefined set of target operation regions, it is marked as a target region gaze event. Based on the timestamp sequence, the gaze duration of each target region gaze event is calculated, and the pupil diameter sequence corresponding to the gaze event is extracted to calculate the pupil diameter change rate. The gaze duration and the pupil diameter change rate are weighted and fused to obtain the effective attention weight. The ratio of the cumulative effective attention weight of all target region gaze events to the total training time is used as the effective attention percentage of the target region. The effective attention percentage is weighted and combined with the frequency of occurrence of target region gaze events to obtain an attention allocation score.

[0086] For example, from eye-tracking data collected by a mixed reality device, a two-dimensional coordinate sequence, a timestamp sequence, and a pupil diameter sequence are extracted from the gaze point. The two-dimensional coordinate sequence records the pixel coordinates of the trainee's gaze point on the device screen plane at each sampling moment. The timestamp sequence corresponds one-to-one, recording the absolute time of each sampling point. The pupil diameter sequence reflects the trainee's cognitive load at the corresponding moment. The eye-tracking acquisition frequency is typically no less than 60Hz to ensure that the temporal resolution of the gaze trajectory meets the requirements of subsequent analysis. After obtaining the above raw sequences, it is also necessary to read the viewpoint position parameters and viewpoint direction parameters of the mixed reality device in the current frame. The viewpoint position parameters describe the three-dimensional coordinates of the device's camera in the world coordinate system, and the viewpoint direction parameters describe the camera's orientation quaternion or rotation matrix. Together, they determine the coordinate transformation relationship from the screen coordinate system to the anatomical model space coordinate system.

[0087] When establishing coordinate transformation relationships, the two-dimensional pixel coordinates on the screen are back-projected into direction vectors in the camera coordinate system using the device intrinsic parameter matrix. Then, these direction vectors are transformed to the world coordinate system of the anatomical model using extrinsic parameters (viewpoint position parameters and viewpoint direction parameters), resulting in a three-dimensional gaze ray extending along the gaze direction from the device's optical center. This transformation is repeated for each gaze point in the two-dimensional coordinate sequence, forming a set of three-dimensional gaze rays. Each ray in this set carries a corresponding timestamp for easy alignment with subsequent time-series analysis.

[0088] When calculating the intersection coordinates of the 3D gaze ray set with the surfaces of each anatomical structure in the anatomical model, each ray is sequentially subjected to ray-patch intersection detection with triangular mesh patches of anatomical structures such as the bladder wall, urethra, prostate, and urethral sphincter. The intersection point closest to the optical center along the ray direction is taken as the coordinates of the current gaze point in 3D space. The anatomical structure to which this intersection point belongs is identified and recorded as the gaze anatomical structure region at this sampling time. If a ray does not intersect with any anatomical structure, the sampling point is marked as an invalid gaze and is not included in subsequent scoring calculations.

[0089] The predefined set of target operating areas is determined according to bladder irrigation procedures and typically includes the catheter insertion site (external urethral meatus region), key anatomical structures along the catheter advancement path (such as the membranous urethra and bladder neck), and the irrigation fluid injection observation area (bladder trigone). When the gaze location at a given moment belongs to any member of the above target operating area set, that moment is marked as a candidate starting point for a target area gaze event. When several consecutive sampling points fall within the same target area, they are merged into a single complete target area gaze event to avoid event fragmentation caused by eye movement tremors.

[0090] The duration of gaze in each target region is calculated based on the timestamp sequence. This is the difference between the timestamp of the last sampling point and the timestamp of the first sampling point within the event. Simultaneously, the pupil diameter sequence corresponding to this fixation event is extracted, and the pupil diameter change rate is calculated. The rate of change in pupil diameter reflects the intensity of cognitive processing during fixation on the target area: a larger pupil dilation indicates that the trainee has invested more cognitive resources in that area; a smaller rate of change, or even pupil constriction, may indicate attention wandering or that the operation has become automated. Weighted fusion of fixation duration and pupil diameter change rate yields the... Effective attention weight for gaze events in a target region The calculation method is as follows: ;in The fusion weighting coefficient is the fixation duration. This is the fusion weighting coefficient for the pupil diameter change rate; the sum of the two is 1, and the specific value can be configured according to the type of operation task and training phase. For example, during the initial catheter insertion phase, the trainee needs to continuously observe the urethral inlet position. It can be set to 0.7. The value was set to 0.3 to emphasize the contribution of fixation persistence; however, during the irrigation fluid injection observation phase, the trainee needs to quickly scan multiple observation points and make judgments about fluid recirculation. It can be reduced to 0.4. The weighting is increased to 0.6 to better capture dynamic changes in cognitive load. When trainees are in the initial learning stage, the weighting of fixation duration in effective attention is usually set higher; as training progresses, the weighting of pupil diameter change rate can be appropriately increased to capture subtle changes in cognitive load more precisely.

[0091] After obtaining the effective attention weights for all gaze events in the target region, sum them up and multiply by the total training time. The ratio is defined as the percentage of effective attention received by the target region. : ;in This represents the total number of fixation events occurring in the target region during the training process. The larger the value, the more effective attentional resources the trainee concentrates on the target operation area, and the closer the overall attention allocation pattern is to the expert operation standard.

[0092] Attention allocation score By increasing the percentage of effective attention Frequency of gaze events in the target area We obtain the following results through weighted combination calculation: ;in The combined weights for the effective attention percentage, A combined weight for the frequency of observed events. Introducing frequency. The reason is that relying solely on the percentage of effective attention may overlook the breadth of the trainee's saccades over the target area: if the trainee stares at a single target area for an extended period while ignoring other key structures, The estimate may be too high, but the overall operational awareness is insufficient; while It can reflect how frequently a trainee switches their attention between different target areas, and combining the two can provide a more comprehensive assessment of the quality of attention allocation. and The value needs to be configured differently depending on the specific stage of the bladder irrigation procedure, such as during catheter insertion. The weight should be higher than This emphasizes the need for continued attention to the urethral pathway; during the irrigation fluid injection phase, The weight can be appropriately increased to encourage trainees to make effective attentive patrols between the bladder trigone and the end of the catheter.

[0093] The calculation results will serve as one of the core inputs for generating subsequent navigation guidance information. When When the level is low, it indicates that the trainee is not focusing enough on key anatomical structures. Navigation guidance information will be presented in a more detailed form, such as highlighting the target operation area and increasing the frequency of voice prompts; when At higher levels, the level of detail in the navigation guidance information decreases accordingly, gradually guiding trainees to transition to independent operation mode, thereby achieving an adaptive training feedback mechanism based on attention state.

[0094] In one optional implementation, bladder irrigation operation navigation guidance information is generated based on the spatial deviation sequence and the attention allocation score, wherein the level of detail of the navigation guidance information is inversely modulated by the attention allocation score, including:

[0095] Calculate the mean and variance of the spatial deviation for the spatial deviation sequence;

[0096] Based on the mean and variance of the spatial deviation, key deviation moments are identified. If the spatial deviation value corresponding to a key deviation moment exceeds the sum of the mean and variance of the spatial deviation, the anatomical structure region identifier corresponding to the key deviation moment is extracted.

[0097] The attention allocation score is compared with a preset attention threshold to classify low attention states and high attention states.

[0098] Detailed navigation guidance information is generated for low-attention states. This detailed navigation guidance information includes the anatomical structure region identifier, the safe operation constraints for the anatomical structure region, and the path correction guidance direction.

[0099] Brief navigation guidance information is generated for high attention states, and the brief navigation guidance information includes the anatomical structure region identifiers;

[0100] The detailed or concise navigation guidance information is presented to the trainee through the mixed reality device.

[0101] For example, after obtaining the spatial deviation sequence, the spatial deviation values ​​at all times in the sequence are statistically calculated to obtain the mean spatial deviation. Spatial bias variance The mean spatial deviation reflects the average level of deviation between the trainee's overall operational trajectory and the optimal path, while the variance of spatial deviation characterizes the fluctuation characteristics of the deviation from the operational trajectory. When the spatial deviation value at a certain moment exceeds... At this point, the deviation moment is marked as a critical deviation moment. This criterion balances the trainee's overall operational baseline and operational stability, avoiding misjudging minor deviations as critical issues, while also effectively capturing significant abnormal nodes that occur during operation. For each critical deviation moment, its corresponding anatomical structural region identifier in the anatomical model is extracted, such as the urethral orifice region, membranous urethra region, bladder neck region, etc., thereby associating abstract trajectory deviation information with specific anatomical locations, providing a positioning basis for the generation of subsequent navigation guidance information.

[0102] After identifying the critical moments of deviation and their corresponding anatomical regions, attention allocation scoring is performed. Compared with the preset attention threshold Comparisons are made to differentiate the trainee's current cognitive state. When When a trainee is judged to be in a state of low attention, it indicates that they are not paying enough attention to key anatomical areas, and their operational decisions may lack sufficient visual information support; when When the trainee is judged to be in a state of high attention, it indicates that their focus on the operational area has reached a relatively sufficient level, and operational deviations are more likely to stem from skill execution rather than insufficient attention allocation. (Preset attention threshold) It can be configured according to the baseline level of different training stages and trainee groups to adapt to the needs of different training scenarios.

[0103] To address low attention levels, detailed navigation guidance information is generated. This detailed guidance information comprises three core components: anatomical region markers corresponding to critical deviation moments, safe operational constraints for those anatomical regions, and path correction guidance directions. Anatomical region markers are highlighted and overlaid in the mixed reality view, allowing trainees to intuitively perceive the specific anatomical areas involved in the current operation. Safe operational constraints are predefined region-level operational guidelines based on the bladder-urethra anatomy model. For example, when operating in the membranous urethra region, the angle between the catheter axis and the urethral direction must be kept within a safe range, and excessive lateral force on the urethral wall must be avoided. These constraints are presented in the mixed reality interface as text prompts or illustrations. Path correction guidance directions visually demonstrate the direction and magnitude of catheter movement to be adjusted using 3D arrows or path curves, helping trainees effectively combine visual attention with operational correction. The design logic of detailed navigation guidance information is that when trainees have insufficient attention, their perception of the operational environment has a significant gap, requiring rich external information prompts to compensate for the lack of cognitive resources, thereby reducing the risk of operational errors.

[0104] For trainees in a state of high attention, concise navigation guidance information is generated. This concise guidance information only includes the anatomical structure regions corresponding to key deviation moments, presented in a lightweight visual annotation style, without detailed constraints or path correction arrows. This design is based on the following considerations: trainees in a state of high attention have already maintained sufficient focus on the manipulation area, and their operational deviations are more likely to stem from insufficient precision of movement execution or skill proficiency, rather than a lack of information acquisition; providing a large number of detailed prompts at this time may interfere with the trainee's autonomous decision-making process, which is not conducive to the internalization and consolidation of operational skills. By minimizing information intervention, concise navigation guidance information corrects operational deviations while preserving the trainee's autonomous operational space, which helps to promote the formation of higher-order operational skills.

[0105] Both detailed and concise navigation guidance information are presented to the trainee's field of vision through overlay rendering using a mixed reality device. The mixed reality device spatially registers the virtual navigation information with the physical operation scene seen by the trainee, ensuring a precise correspondence between the navigation annotations and the 3D position of the anatomical model. This allows the trainee to obtain the necessary navigation guidance without interrupting their operation. The presentation method supports dynamic updates; that is, as the trainee's operation trajectory changes in real time, the spatial deviation sequence is continuously updated, and the identification results and attention state judgment results at key deviation moments are also refreshed accordingly. The content and level of detail of the navigation guidance information can adaptively adjust to the trainee's operation state. For example, when the trainee receives a detailed navigation prompt at a key deviation moment, the attention allocation score... Rise to threshold In the above, the navigation guidance will automatically switch to a simplified level to reduce visual information interference and guide trainees to gradually transition to a more autonomous operation mode.

[0106] In the specific implementation, the extraction of anatomical structure region identifiers at critical deviation moments relies on the real-time spatial mapping relationship between the catheter position and the anatomical model. At each critical deviation moment, the actual position coordinates of the catheter are projected onto the region segmentation results of the anatomical model to determine its corresponding anatomical structure region, and the pre-bound safety operation constraints and standard path tangent direction of that region are read. The path correction guidance direction is calculated by pointing from the actual catheter position at the critical deviation moment to the nearest reference point on the optimal path, combined with the path tangent direction at that reference point, to ensure that the correction direction points to the optimal path and is consistent with the path travel direction, avoiding over-correction or directional misdirection.

[0107] The entire navigation guidance generation process runs continuously with low latency during training, ensuring that trainees receive corresponding navigation prompts within a very short time after an operational deviation occurs, thus enabling immediate feedback and correction. This mechanism, which adaptively adjusts the level of navigation detail based on attentional state, allows the training system to dynamically allocate information resources according to the trainee's cognitive load, improving operational safety while ensuring the effectiveness of skill development.

[0108] In one optional implementation, a structural causal graph is constructed using the operational error event, the spatial deviation sequence, and the time series of the attention allocation score as observational data. Causal inference is used to identify the latent variable with the greatest causal effect on the operational error event, which is then mapped to a bladder irrigation operation skill deficiency type, including:

[0109] The time sequence of operational error events, the spatial deviation sequence, and the time sequence of attention allocation scores are aligned along a unified time axis to form a multidimensional dataset;

[0110] A structural causal graph is constructed based on the aforementioned multidimensional dataset. The nodes in the structural causal graph include observed variable nodes and latent variable nodes. When there is a conditional dependency between two observed variable nodes that cannot be explained by the existing observed variable nodes, the latent variable node is introduced as a common cause.

[0111] Calculate the total causal effect strength of the latent variable nodes in the causal graph on the operational error event;

[0112] The latent variable node with the strongest overall causal effect is identified as the dominant latent variable. The causal path pattern of the dominant latent variable is mapped to the type of bladder irrigation operation skill deficiency. When the causal path of the dominant latent variable is transmitted through the spatial bias sequence, it is mapped to a path planning skill deficiency. When the causal path of the dominant latent variable is transmitted through the attention allocation score, it is mapped to an attention allocation skill deficiency. When the causal path of the dominant latent variable is transmitted through both the spatial bias sequence and the attention allocation score, it is mapped to a coordination and integration skill deficiency.

[0113] Combination Figure 2This paper illustrates the process of identifying bladder irrigation skill deficiencies through a flowchart using causal inference. During bladder irrigation skills training, trainee errors are often not isolated occurrences but are driven by multiple potential skill deficiencies. To trace the root causes of errors at the data level, it is necessary to integrate the time series of operational error events, spatial deviation sequences, and attention allocation scores into a unified multidimensional dataset for causal analysis. Specifically, operational error events are recorded as binary time series, with each discrete time point marked as indicating whether an error occurred. The spatial deviation sequence records the deviation between the actual catheter position and the optimal path at each time point. The attention allocation score time series is sampled using a sliding window method, reflecting the trainee's attention allocation status over a continuous time period. All three types of sequences are aligned with a unified time axis, and the time resolution is the time step corresponding to the greatest common divisor of the three data collection frequencies. This ensures that each type of data has a corresponding valid value under the same time index, thus forming a complete multidimensional dataset and providing a unified observational basis for the subsequent construction of a structural causal graph.

[0114] When constructing the structural causal graph based on the aforementioned multidimensional dataset, the nodes in the graph are divided into two categories: observed variable nodes and latent variable nodes. Observed variable nodes directly correspond to the dimensions in the multidimensional dataset, namely, operation error event nodes, spatial bias nodes, and attention allocation score nodes. When determining the directed edges between nodes, a causal discovery algorithm based on conditional independence testing is used to screen the dependencies between observed variables. When a significant conditional dependency still exists between two observed variable nodes given all other observed variables, i.e., this dependency cannot be explained by the common influence or mediation effect of existing observed variable nodes, it is considered that there is an unobserved common cause. In this case, a latent variable node is introduced as the common parent node of these two observed variable nodes. The number of latent variable nodes introduced follows the principle of minimum sufficiency, i.e., the number of latent variable nodes is minimized while still being able to explain all residual conditional dependencies, thus avoiding excessive model complexity. The parameters of each latent variable node are estimated using the expectation-maximization algorithm or variational inference method to make it statistically optimal in explaining the corresponding residual dependency structure.

[0115] After the structural causal graph is constructed, the total causal effect strength of each latent variable node on the operational error event node is calculated. The calculation of the total causal effect strength uses the do-coefficient calculus framework, which calculates the magnitude of change in the probability of the operational error event by intervening in the latent variable nodes. Let the latent variable nodes be... Operation error event node The overall causal effect strength is ,but Defined as in the When intervention is carried out, The change in the expected value relative to the baseline state, i.e. ,in and They are respectively High activation state values ​​and low activation state values, through... The posterior distribution is determined by the upper and lower quartiles. Since there are multiple directed paths from latent variable nodes to operational error event nodes in the structural causal graph, the overall causal effect strength integrates the direct and indirect effects on all paths, and can comprehensively reflect the actual driving force of latent variables on operational errors.

[0116] After identifying the latent variable node with the strongest overall causal effect, it is defined as the dominant latent variable. Further analysis of the causal path pattern of this dominant latent variable is then performed to determine the corresponding type of bladder irrigation operation skill deficiency. The identification of the causal path pattern is achieved through path-specific effect decomposition, which decomposes the overall causal effect of the dominant latent variable on the operational error event into several path-specific components according to the transmission path. Specifically, if the causal effect of the dominant latent variable on the operational error event is mainly transmitted through spatial deviation nodes, that is, if the proportion of the path-specific effect component via the spatial deviation path in the overall causal effect exceeds a preset threshold, then the causal path is considered dominant. If the trainee has a path planning skill deficiency, it is determined that the trainee has insufficient ability to track the optimal path during catheter insertion. There is a systematic deviation between the actual movement trajectory of the catheter and the optimal path under the constraints of the anatomical structure. This threshold is usually set at 0.6, which means that when the contribution of the spatial deviation path exceeds 60% of the total causal effect, it can be judged as a path planning dominant deficiency. This value is determined based on the clinical experts' experience in distinguishing different skill deficiency types.

[0117] If the causal effect of the dominant latent variable on the operational error event is mainly transmitted through the attention allocation scoring node, that is, the proportion of the specific effect component through the attention allocation path to the total causal effect exceeds a preset proportion threshold. If the trainee exhibits a deficiency in attention allocation skills, it is determined that they have insufficient fixation time or low fixation frequency on key anatomical structures during the operation, resulting in a lack of perception and utilization of important visual information in the operating environment, and consequently, operational judgment errors. Identifying these attention allocation skill deficiencies helps in designing targeted attention guidance training modules. By applying visual cues to key anatomical areas in mixed reality scenarios, trainees can gradually develop correct attention allocation habits.

[0118] When the causal effect of the dominant latent variable on the operational error event is transmitted simultaneously through both the spatial bias node and the attention allocation scoring node, and the specific effect components of both paths do not individually exceed a preset proportion threshold. However, the sum of the two exceeds the combined proportion threshold of the total causal effect. If this occurs, the trainee is deemed to have a deficiency in coordination and integration skills. This deficiency means that during bladder irrigation, the trainee is unable to effectively plan and execute the catheter insertion path, nor can they properly allocate visual attention to anatomical structures; there is an obstacle to the synergistic integration of these two abilities. Such deficiencies typically appear in the early stages of training, when the trainee has not yet established a coordinated linkage mechanism between operational perception and action execution, requiring intervention through a comprehensive, multi-dimensional training program.

[0119] After the skill deficit type mapping is completed, the identification results will be output in the form of structured labels. The labels include the number of the dominant latent variable, the corresponding skill deficit type name, and the total causal effect strength value. The proportion of each path-specific effect component is also provided for use by the subsequent training scenario difficulty parameter adjustment module. For the skill deficit type sequence of the same trainee in multiple training sessions, the evolution trend of skill deficits can be further tracked through longitudinal comparative analysis to evaluate the effectiveness of training intervention, thereby providing data support for the dynamic optimization of personalized training programs.

[0120] In one alternative implementation, calculating the total causal effect strength of the latent variable nodes in the structural causal graph on the operational error event includes:

[0121] Extract all latent variable nodes that have a causal path to the operation error event node from the structural causal graph. The causal path is a path that starts from a latent variable node, passes through one or more directed edges, and finally points to the operation error event node.

[0122] The total causal effect strength is obtained by summing the causal effects of the latent variable node on the operational error event via different causal paths. The causal effect of each causal path is calculated by multiplying the causal strength coefficients of each directed edge on the path. The causal strength coefficients are determined based on the mutual information and time lag correlation of the corresponding observed variables in the multidimensional dataset.

[0123] For example, after the structural causal graph is constructed, it is necessary to calculate the total causal effect strength of each latent variable node on the operational error event node in order to subsequently identify the skill deficiency root cause with the greatest causal effect on the operational error. This calculation process consists of two steps: causal path extraction and path effect accumulation, which together constitute a complete causal effect quantification process.

[0124] When extracting causal paths from a structural causal graph, the graph structure is traversed backwards along directed edges, starting with the operation error event node as the endpoint, to find all latent variable nodes that can ultimately point to the operation error event node via one or more directed edges. Specifically, a directed reachability check is performed on each latent variable node in the graph: if a path exists that starts from the latent variable node, moves along a directed edge, and ultimately reaches the operation error event node, then the latent variable node is included in the set of valid causal source nodes, and all directed paths that satisfy the condition are recorded. Paths can be single-hop direct paths (latent variable nodes directly point to the operation error event node via a directed edge) or multi-hop indirect paths (latent variable nodes ultimately point to the operation error event node via directed edge chains of one or more intermediate nodes). In the bladder irrigation operation scenario, a typical multi-hop path is as follows: the catheter holding stability latent variable node first affects the spatial bias observation node, and the spatial bias observation node then affects the operation error event node; while the attention allocation latent variable node may have both indirect paths via gaze matching event nodes and direct paths directly pointing to the operation error event node. During the extraction process, the graph needs to be topologically sorted to avoid getting stuck in infinite traversal when there are loop structures. In practice, a hierarchical traversal strategy can be used for mixed graphs of latent variable nodes and observation nodes to trace the causal source layer by layer.

[0125] The causal effect of each causal path is calculated by multiplying the causal strength coefficients of the directed edges along the path. Let a path start from a latent variable node... To the operation error event node causal path A directed edge, the first The causality strength coefficient of a directed edge is Then the path causality corresponding to this path for: ; where subscript Indicates the first The path is numbered. The physical meaning of the chain multiplication operation is that the influence of the upstream node on the downstream node is transmitted step by step along the path. Each time a directed edge is passed, the effect is attenuated or amplified by the causal strength coefficient corresponding to that edge. Finally, the effect intensity when it reaches the operation error event node is equal to the product of the transmission coefficients of each segment. When the path is long, if the coefficients of each edge are all less than 1, the path effect decreases with the increase of the path length. This is consistent with the law of diminishing influence of remote factors in the transmission of practical operation skills.

[0126] Causality intensity coefficient The determination of mutual information utilizes two types of statistics: mutual information and time-lag correlation. It measures the statistical dependence between variables at both ends of a directed edge, is not limited by the linear assumption, and can capture nonlinear dependencies; time-lag correlation. Then, by calculating the Pearson correlation coefficient for the time series of the two variables at different lag steps, the correlation coefficient value corresponding to the lag step with the largest absolute value is taken to characterize the temporal leader relationship in the causal direction. The causal strength coefficient is obtained by fusing the two results. ;in The normalized mutual information value. This represents the absolute value of the time-lag correlation coefficient. and These are the mutual information fusion weights and the time-lag correlation fusion weights, respectively, and their sum is 1. Normalized mutual information is calculated by dividing the original mutual information by half the sum of the entropy values ​​of the variables at both ends, so that... The value range is constrained to between 0 and 1, facilitating fusion with the time lag correlation coefficient under the same dimension. In bladder irrigation operation data, catheter hand jitter and spatial deviation typically exhibit a strong nonlinear dependence, and mutual information weights... The weighting can be appropriately increased; however, the temporal leading relationship between attention allocation and operational errors is quite obvious, while the weighting of time-lag correlation is relatively low. The weights can be increased accordingly, and the specific weight values ​​can be determined through cross-validation based on the statistical characteristics of the training dataset.

[0127] After obtaining the path causal effects of each path, the latent variable nodes are... By summing all causal paths, we obtain the node corresponding to the operation error event node. Total causal effect strength : ;in For latent variable nodes To the operation error event node The total number of valid causal paths. The summation operation reflects the joint contribution of multiple paths to operational errors: when a latent variable node simultaneously influences operational errors through multiple paths, its total causal effect is stronger than that of a node that propagates through only a single path, indicating that the latent variable has a more comprehensive and profound impact on operational errors.

[0128] In practical calculations, to prevent the computational burden caused by the explosion of the number of paths, a pruning threshold can be set for path effects: when the path causality effect of a certain path is reduced... Below the preset path effect pruning threshold When this happens, the path should be excluded from the accumulation to reduce the interference of inefficient paths on the accuracy of the total effect estimation. For causal graphs containing loop structures, path enumeration should be performed after expanding the equivalent form of the directed acyclic graph to ensure the completeness and non-redundancy of path extraction.

[0129] The total causal effect strength of all latent variable nodes After the calculation is completed, the latent variable nodes with the largest total causal effect strength are sorted in descending order and identified as the root causes that contribute most significantly to the current trainee's operational errors. These are then mapped to the corresponding bladder irrigation operation skill deficiency types, such as unstable catheter entry angle control, improper adjustment of irrigation fluid injection rate, or anatomical structure positioning deviation, providing a basis for targeted adjustment of the difficulty parameters of subsequent training scenarios.

[0130] In one optional implementation, adjusting the training scenario difficulty parameters based on the skill deficiency type and the statistical characteristics of the spatial deviation sequence includes:

[0131] To address deficiencies in path planning skills, the anatomical regions where spatial deviation peaks occur in the spatial deviation sequence are extracted. The density of path planning guidance cues in these anatomical regions is increased, and the catheter manipulation speed limit threshold in these regions is reduced to extend the trainee's operational decision-making time.

[0132] To address the deficiencies in attention allocation skills, the number of visual distracting elements presented in non-target operation areas is reduced, the visual salience of target operation areas is enhanced, and phased attention checkpoints are introduced into the training scenario.

[0133] To address the deficiencies in coordination and integration skills, the fluctuation frequency of the spatial deviation sequence is calculated. When the fluctuation frequency exceeds a preset frequency threshold, it is determined to be an unstable operational state. In response to this state, the density of path guidance prompts and the salience of visual annotations are adjusted synchronously, and an operational rhythm prompt signal is set in the training scenario to help trainees establish a stable operational rhythm pattern.

[0134] For example, differentiating the difficulty parameters of training scenarios based on the statistical characteristics of skill deficiency types and spatial deviation sequences is a core element in achieving personalized adaptive training. Different skill deficiency types exhibit different behavioral patterns, therefore, the corresponding scenario adjustment strategies differ fundamentally in both their scope of application and direction of adjustment.

[0135] Adjustments to address path planning skill deficiencies focus on the anatomical regions where spatial deviation peaks are concentrated. Deviation values ​​at each time step are extracted from the spatial deviation sequence, identifying deviation values ​​exceeding a preset peak threshold. The time points were analyzed, and the corresponding catheter positions were mapped to the coordinate system of the bladder-urethra anatomy model. The spatial density distribution of deviation peak points within each anatomical region (such as the bladder neck, posterior urethral bend, and bulbar urethra) was statistically analyzed. Anatomical regions where the density of deviation peak points exceeded the regional density mean were marked as high-difficulty operation regions. For these high-difficulty anatomical regions, the presentation density of path planning guidance prompts in the mixed reality view was increased. Specifically, a visible path guidance marker was placed at shorter arc length intervals along the optimal path within the region, allowing trainees to obtain denser spatial reference information when traversing the region, thereby reducing path deviations caused by the complexity of the anatomical structure. At the same time, the catheter operation speed limit threshold in this region was lowered. When the advance speed of the catheter in this area is detected to exceed When the speed is too fast, a real-time alert is triggered, forcibly extending the trainee's decision-making time in that area. The magnitude of the speed limit threshold reduction can be adjusted proportionally based on the peak deviation density in that area: the higher the peak deviation density, the greater the reduction in the speed limit threshold, ensuring that the trainee has sufficient time for spatial judgment and path correction in areas with complex anatomical structures.

[0136] The adjustments to address attention allocation skill deficiencies focus on reducing ineffective visual competition and enhancing the salience of key areas. In normal training scenarios, the mixed reality field of view may simultaneously present multiple auxiliary information elements, including anatomical annotations, operational parameter values, and highlighted structural outlines of non-target areas. When a trainee's attention allocation skill deficiency is identified, the number of visual distractions in non-target operational areas is reduced sequentially, prioritizing the hiding of anatomical annotations and auxiliary information irrelevant to the current operational stage, preventing the trainee's visual attention from being diverted to irrelevant areas. Simultaneously, enhanced visual annotation salience processing is applied to the target operational area, such as increasing the brightness and contrast of the target area's boundary contours, increasing the size of the target area's annotation symbols, or adding periodically flashing visual attraction signals to drive the trainee's gaze towards the target area. On the training scenario's timeline, phased attention checkpoints are introduced before and after key operational nodes: when the catheter advances to a preset anatomical position checkpoint, the current operational process is paused, and an attention verification interface requiring active confirmation from the trainee is presented in the mixed reality field of view, requiring the trainee to identify the current anatomical position of the catheter or confirm the target area for the next operation. The operation process only resumes after the trainee completes the correct confirmation, thereby embedding mandatory attention-focusing nodes in the training process and gradually cultivating the trainee's habit of actively allocating attention to key anatomical areas during the operation.

[0137] Adjustments to address deficiencies in coordination and integration skills require, first and foremost, a quantitative description of the degree of operational instability. This can be achieved through spatial deviation sequences. Calculating the fluctuation frequency The specific method involves counting zero-crossings of the deviation sequence: subtracting the moving average from the deviation sequence yields the detrended sequence, and counting the number of times the detrended sequence crosses zero from a positive value to a negative value (or from a negative value to a positive value) per unit time; this number represents the fluctuation frequency. .when Exceeding the preset frequency threshold When this condition is triggered, the trainee is determined to be in an unstable operational state. Upon triggering this state, a joint adjustment is made to both the path guidance cue density and the visual annotation salience: the path guidance cue density increases proportionally to the magnitude of the fluctuation frequency exceeding a threshold. Higher than The smaller the spacing between path guidance markers, the more salience of visual annotations is improved, ensuring that trainees can clearly identify the boundaries of the target area even under shaking conditions. Based on this, rhythmic cue signals are introduced into the training scenario, presented as periodic audio beats or periodically expanding and contracting rhythmic indicator rings in mixed reality vision. The periodicity of the rhythmic signal... The recommended operating rhythm is pre-set according to the current training task phase. The trainee needs to try to synchronize the changes in the speed of catheter advancement with the period of the rhythm signal, gradually establishing a stable operating rhythm pattern. The duration of the rhythm cue signal is linked to the duration of the unstable operating state: when continuous detection... Below The duration reaches the preset stability determination time. Afterward, the unstable state of operation is resolved, and the intensity of the rhythm cue signal gradually weakens until it disappears, in order to avoid the trainee from becoming overly dependent on the external rhythm signal.

[0138] The scenario adjustment strategies corresponding to the three types of skill deficiencies mentioned above are not mutually exclusive in actual training. When causal inference results show that the trainee has multiple skill deficiencies simultaneously, various adjustment strategies can be implemented in combination. When implementing these strategies in combination, attention should be paid to the presentation priority of visual elements: when path planning guidance markers and target area visual annotations coexist, the salience parameter of the target area annotations should be prioritized, and the increase in the density of path guidance markers should not obscure the visibility range of the target area annotations. Under the combined effect of the operation rhythm cues and the speed limit threshold, the trainee's progress rhythm in high-difficulty anatomical areas is constrained by both the speed limit and the periodic guidance of the rhythm signals. Together, they constitute a dual regulation mechanism for the operation sequence, which helps the trainee form a standardized and stable operational behavior pattern in complex anatomical areas.

[0139] A second aspect of the present invention provides a bladder irrigation intelligent navigation training system based on mixed reality, comprising:

[0140] The data acquisition unit is used to acquire catheter manipulation trajectory and eye movement data of trainees during bladder irrigation training using mixed reality devices;

[0141] The path analysis unit is used to generate the optimal catheter insertion path based on the bladder-urethra anatomical model and to calculate the spatial deviation sequence between the catheter operation trajectory and the optimal path through the path planning algorithm.

[0142] The attention assessment unit is used to map the eye movement data to the anatomical structure region of the gaze identified by the anatomical model, extract the matching events and gaze duration between the gaze anatomical structure region and the target operation region set, and calculate the attention allocation score based on the matching events and gaze duration.

[0143] A navigation generation unit is used to generate bladder irrigation operation navigation guidance information based on the spatial deviation sequence and the attention allocation score, wherein the level of detail of the navigation guidance information is inversely modulated by the attention allocation score.

[0144] The causal analysis unit is used to construct a structural causal graph using the time series of the operational error event, the spatial deviation sequence, and the attention allocation score as observation data. It identifies the latent variable with the greatest causal effect on the operational error event through causal inference and maps it to the bladder irrigation operation skill defect type.

[0145] The difficulty adjustment unit is used to adjust the training scenario difficulty parameters based on the skill defect type and the statistical characteristics of the spatial deviation sequence.

[0146] A third aspect of the present invention provides an electronic device, comprising:

[0147] processor;

[0148] Memory used to store processor-executable instructions;

[0149] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0150] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0151] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A training method for intelligent navigation of bladder irrigation based on mixed reality, characterized in that, include: Acquire catheter manipulation trajectory and eye movement data of trainees during bladder irrigation training using mixed reality devices; Based on the anatomical model of the bladder-urethra, the optimal catheter insertion path is generated through a path planning algorithm, and the spatial deviation sequence between the catheter operation trajectory and the optimal path is calculated. The eye-tracking data is mapped to the anatomical structure region of the gaze identified by the anatomical model. Matching events and gaze duration between the gaze anatomical structure region and the target operation region set are extracted. An attention allocation score is calculated based on the matching events and gaze duration. Based on the spatial deviation sequence and the attention allocation score, bladder irrigation operation navigation guidance information is generated, and the level of detail of the navigation guidance information is inversely modulated by the attention allocation score. Using the time series of operational error events, the spatial deviation sequence, and the attention allocation score as observation data, a structural causal graph is constructed. Latent variables with the greatest causal effect on the operational error events are identified through causal inference and mapped to the bladder irrigation operation skill defect type. The training scenario difficulty parameters are adjusted based on the statistical characteristics of the skill deficiency type and the spatial deviation sequence.

2. The method according to claim 1, characterized in that, Based on the bladder-urethra anatomical model, an optimal catheter insertion path is generated using a path planning algorithm. The spatial deviation sequence between the catheter manipulation trajectory and the optimal path is calculated, including: Based on the aforementioned anatomical model, the urethra is divided into the external urethral orifice region, the bulbous urethra region, the membranous urethra region, and the bladder neck region; For each anatomical region, a region-specific path planning strategy is constructed. The path planning strategy includes tissue safety constraint weight coefficients and path efficiency constraint weight coefficients. The tissue safety constraint weight coefficients for the bulbous urethra and membranous urethra are higher than those for the external urethral orifice and bladder neck regions. The catheter manipulation trajectory is divided into multiple regional trajectory segments. For each regional trajectory segment, the minimum safe distance threshold between the path and the urethral wall is determined according to the tissue safety constraint weight coefficient of the corresponding region. The trade-off ratio between path length and path curvature is determined according to the path efficiency constraint weight coefficient. The optimal path segment for the region is calculated based on the minimum safe distance threshold and the trade-off ratio as constraints. Connect all optimal path segments in the region according to the anatomical order of the urethra and smooth the junctions to generate the optimal catheter insertion path; Align the catheter manipulation trajectory with the optimal catheter insertion path in the time dimension and calculate the spatial deviation sequence.

3. The method according to claim 1, characterized in that, The eye-tracking data is mapped to the anatomical structure region of the gaze identified by the anatomical model. Matching events and gaze durations between the gaze anatomical structure region and the target operation region set are extracted. An attention allocation score is calculated based on the matching events and gaze durations, including: Extract the two-dimensional coordinate sequence, timestamp sequence, and pupil diameter sequence of the gaze fixation point from the eye-tracking data; obtain the viewpoint position parameters and viewpoint direction parameters of the mixed reality device; establish the coordinate transformation relationship from the screen coordinate system to the anatomical model space coordinate system; and map the two-dimensional coordinate sequence into a set of three-dimensional gaze rays. Calculate the coordinates of the intersection points between the three-dimensional gaze ray set and the surfaces of each anatomical structure in the anatomical model, and use the anatomical structure identifiers corresponding to the intersection coordinates as the gaze anatomical structure regions; When the gaze anatomical region belongs to the corresponding predefined set of target operation regions, it is marked as a target region gaze event; Based on the timestamp sequence, the gaze duration for each gaze event in the target region is calculated. The pupil diameter sequence corresponding to the gaze event is extracted to calculate the pupil diameter change rate. The gaze duration and the pupil diameter change rate are weighted and fused to obtain the effective attention weight. The ratio of the cumulative effective attention weights of all gaze events in the target region to the total training time is taken as the effective attention percentage of the target region. The effective attention percentage is then weighted and combined with the frequency of gaze events in the target region to calculate the attention allocation score.

4. The method according to claim 1, characterized in that, Based on the spatial deviation sequence and the attention allocation score, bladder irrigation operation navigation guidance information is generated, including: Calculate the spatial deviation mean and spatial deviation variance for the spatial deviation sequence; identify key deviation moments based on the spatial deviation mean and spatial deviation variance, where the spatial deviation value corresponding to the key deviation moment exceeds the sum of the spatial deviation mean and spatial deviation variance, and extract the anatomical structure region identifier corresponding to the key deviation moment; The attention allocation score is compared with a preset attention threshold to classify low attention states and high attention states. For low-attention states, detailed navigation guidance information is generated, including the anatomical structure region identifier, the safe operation constraints of the anatomical structure region, and the path correction guidance direction; for high-attention states, concise navigation guidance information is generated, including the anatomical structure region identifier.

5. The method according to claim 1, characterized in that, Using operational error events, the spatial deviation sequence, and the time series of attention allocation scores as observational data, a structural causal graph is constructed. Causal inference identifies latent variables with the greatest causal effect on the operational error events, mapping them to bladder irrigation operation skill deficiency types, including: The time sequence of operational error events, the spatial deviation sequence, and the time sequence of attention allocation scores are aligned along a unified time axis to form a multidimensional dataset; A structural causal graph is constructed based on the aforementioned multidimensional dataset. The nodes in the structural causal graph include observed variable nodes and latent variable nodes. When there is a conditional dependency between two observed variable nodes that cannot be explained by the existing observed variable nodes, the latent variable node is introduced as a common cause. Calculate the total causal effect strength of the latent variable nodes in the causal graph on the operational error event; The latent variable node with the strongest overall causal effect is identified as the dominant latent variable. The causal path pattern of the dominant latent variable is mapped to the type of bladder irrigation operation skill deficiency. When the causal path of the dominant latent variable is transmitted through the spatial bias sequence, it is mapped to a path planning skill deficiency. When the causal path of the dominant latent variable is transmitted through the attention allocation score, it is mapped to an attention allocation skill deficiency. When the causal path of the dominant latent variable is transmitted through both the spatial bias sequence and the attention allocation score, it is mapped to a coordination and integration skill deficiency.

6. The method according to claim 5, characterized in that, Calculating the total causal effect strength of the latent variable nodes in the structural causal graph on the operational error event includes: Extract all latent variable nodes that have a causal path to the operation error event node from the structural causal graph. The causal path is a path that starts from a latent variable node, passes through one or more directed edges, and finally points to the operation error event node. The total causal effect strength is obtained by summing the causal effects of the latent variable node on the operational error event via different causal paths. The causal effect of each causal path is calculated by multiplying the causal strength coefficients of each directed edge on the path. The causal strength coefficients are determined based on the mutual information and time lag correlation of the corresponding observed variables in the multidimensional dataset.

7. The method according to claim 6, characterized in that, Adjusting the training scenario difficulty parameters based on the statistical characteristics of the skill deficiency type and the spatial deviation sequence includes: To address deficiencies in path planning skills, the anatomical regions where spatial deviation peaks occur in the spatial deviation sequence are extracted. The density of path planning guidance cues in these anatomical regions is increased, and the catheter manipulation speed limit threshold in these regions is reduced to extend the trainee's operational decision-making time. To address the deficiencies in attention allocation skills, the number of visually distracting elements presented in non-target operation areas is reduced, and phased attention checkpoints are introduced into the training scenario. To address the deficiencies in coordination and integration skills, the fluctuation frequency of the spatial deviation sequence is calculated. When the fluctuation frequency exceeds a preset frequency threshold, it is determined to be an unstable operational state. In response to this state, the density of path guidance prompts and the salience of visual annotations are adjusted synchronously, and an operational rhythm prompt signal is set in the training scenario.

8. A mixed reality-based intelligent navigation training system for bladder irrigation, used to implement the method as described in any one of claims 1-7, characterized in that, include: The data acquisition unit is used to acquire catheter manipulation trajectory and eye movement data of trainees during bladder irrigation training using mixed reality devices; The path analysis unit is used to generate the optimal catheter insertion path based on the bladder-urethra anatomical model and to calculate the spatial deviation sequence between the catheter operation trajectory and the optimal path through the path planning algorithm. The attention assessment unit is used to map the eye movement data to the anatomical structure region of the gaze identified by the anatomical model, extract the matching events and gaze duration between the gaze anatomical structure region and the target operation region set, and calculate the attention allocation score based on the matching events and gaze duration. A navigation generation unit is used to generate bladder irrigation operation navigation guidance information based on the spatial deviation sequence and the attention allocation score, wherein the level of detail of the navigation guidance information is inversely modulated by the attention allocation score. The causal analysis unit is used to construct a structural causal graph using the time series of the operational error event, the spatial deviation sequence, and the attention allocation score as observation data. It identifies the latent variable with the greatest causal effect on the operational error event through causal inference and maps it to the bladder irrigation operation skill defect type. The difficulty adjustment unit is used to adjust the training scenario difficulty parameters based on the skill defect type and the statistical characteristics of the spatial deviation sequence.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images