Digestive endoscope esophagus trajectory optimization method based on Kalman filtering

Abstract

The invention discloses a digestive endoscopic esophageal trajectory optimization method based on Kalman filtering, and belongs to the technical field of digestive endoscopic surgical navigation, and the method comprises the following steps: S1, based on an esophageal model of preoperative CT three-dimensional reconstruction, extracting a geometric center point sequence of an esophageal cavity, generating an esophageal center line, and extracting a geometric center point sequence of the esophageal cavity; the method comprises the following steps of S1, acquiring six-degree-of-freedom pose data of the front end of an endoscope in real time through an electromagnetic positioning system, and generating an original motion track, and S3, performing rigid body registration on the original motion track and an esophagus center line by utilizing an iterative closest point (ICP) algorithm, so as to realize coordinate system coarse alignment. The real-time data of an electromagnetic positioning system and a three-dimensional model reconstructed by CT are combined, the navigation robustness is improved, the pose is adaptively corrected through Kalman filtering, the influence of environmental interference is reduced, the space relation between the endoscope and surrounding organs is visually displayed through an optimized track, and accurate operation of a doctor is assisted.

Description

Technical Field

[0001] The present invention relates to the technical field of digestive endoscopic surgery navigation, and in particular to a digestive endoscopic esophageal trajectory optimization method based on Kalman filtering. Background Art

[0002] In existing digestive endoscopic surgeries, doctors can only observe the images taken by the endoscope during the operation. However, it is difficult for doctors to determine the specific position and orientation of the endoscope from an endoscopic image. Therefore, doctors can only estimate the current posture based on experience by moving the endoscope multiple times and observing its endoscopic images.

[0003] To determine the specific position of the endoscope, an electromagnetic positioning system was introduced. By attaching a sensor coil to the endoscope and generating a low-frequency electromagnetic field through an electromagnetic generator, the sensor coil induces a voltage in the field. By measuring these voltage changes, the coil's position and orientation can be calculated. A 3D model reconstructed from the CT image can visualize the current position of the endoscope and surrounding organs, enabling doctors to more accurately determine surgical plans.

[0004] However, the use of electromagnetic positioning systems for digestive endoscopic surgery currently has the following problems: Dynamic deformation of the esophagus: Due to peristalsis or contraction during surgery, the esophagus's shape may differ from the preoperative CT reconstruction model. Sensor noise interference: The electromagnetic positioning system is affected by electromagnetic interference, instrument jitter, and other factors, resulting in random errors in the posture measurement values.

[0005] These issues can cause the endoscope's trajectory, derived from information captured by the electromagnetic positioning system, to sometimes extend beyond the esophagus, contradicting the desired outcome. Traditional methods rely on single-sensor data or static models, making them difficult to adapt to the dynamic intraoperative environment. Therefore, a technical solution that integrates multi-source data and allows for real-time trajectory correction is urgently needed. Summary of the Invention

[0006] The present invention proposes a Kalman filter-based esophageal trajectory optimization method for digestive endoscopes. By fusing data obtained from an electromagnetic positioning system with a three-dimensional esophageal model reconstructed by CT, the endoscope posture is dynamically corrected to solve the trajectory deviation problem.

[0007] In order to achieve the above effects, the present invention provides the following technical solution: a digestive endoscope esophageal trajectory optimization method based on Kalman filtering, comprising the following steps: S1. Based on the esophageal model reconstructed by preoperative CT three-dimensional reconstruction, the geometric center point sequence of the esophageal cavity is extracted to generate the esophageal centerline.

[0008] S2. The electromagnetic positioning system is used to collect the six-degree-of-freedom posture data of the front end of the endoscope in real time to generate the original motion trajectory.

[0009] S3. Using the iterative closest point (ICP) algorithm, the original motion trajectory is rigidly registered with the esophageal centerline to achieve coarse alignment of the coordinate system.

[0010] S4. Based on the Kalman filter algorithm, the anatomical constraints of the esophageal centerline and the posture data of the electromagnetic positioning system are integrated to dynamically correct the real-time posture of the endoscope.

[0011] S5. Display the corrected endoscope posture in real time on the esophageal three-dimensional model to ensure that the trajectory is within the esophageal lumen.

[0012] Furthermore, according to the operation steps in S1, the method for extracting the esophageal centerline is: slicing the esophagus at equal intervals along the axial direction of the esophagus, calculating the geometric center points of the contours of each slice, and connecting the center points to generate a centerline.

[0013] Furthermore, according to the operation steps in S2, step 2 also includes intercepting the esophageal segment trajectory for optimization by matching the endoscopic image with the anatomical landmark of the cardia.

[0014] Furthermore, according to the operation steps in S3, in step three, the cardia at the junction of the esophagus and the stomach is used as the registration reference point, redundant trajectory segments are eliminated, and the initial registration is completed.

[0015] Further, according to the operation steps in S4, the step 4 includes the following sub-steps: S401. Construct a state space model of Kalman filtering and define the endoscope posture as a state vector, including three-dimensional position coordinates and posture angles.

[0016] S402. Define the covariance matrix of process noise and observation noise to respectively characterize the dynamic error caused by esophageal peristalsis and the measurement error of the electromagnetic positioning system.

[0017] S403: Calculate the optimal endoscope pose estimation value through the prediction and update iterative process of Kalman filtering.

[0018] Furthermore, according to the operation steps in S401, in the state space model, the state transfer matrix and the observation matrix are both 3×3 unit matrices.

[0019] Furthermore, according to the operation steps in S402, the process noise covariance matrix Q is a three-dimensional diagonal matrix with diagonal elements of the order of 10². The observation noise covariance matrix R is a three-dimensional diagonal matrix with diagonal elements of the order of 10¹.

[0020] Furthermore, according to the operation steps in S4, in the prediction step of the Kalman filter, the point on the esophageal centerline closest to the current electromagnetic positioning posture is used as a priori estimated value.

[0021] Furthermore, according to the operation steps in S5, in step five, the corrected posture display delay is less than 50 milliseconds.

[0022] Furthermore, according to the operation steps in S5, the correction result of the Kalman filter makes the endoscope motion trajectory fit the esophageal morphology, avoiding the trajectory from passing through the esophageal wall.

[0023] The present invention provides a digestive endoscope esophageal trajectory optimization method based on Kalman filtering, which has the following beneficial effects: (1) In the present invention, multi-source data fusion improves navigation robustness. By combining the esophageal centerline reconstructed by preoperative CT three-dimensional reconstruction with the real-time posture data of the electromagnetic positioning system, the static model and dynamic sensor information are effectively integrated, overcoming the limitation of traditional methods that rely on a single data source, and significantly improving the adaptability and reliability of the navigation system.

[0024] (2) In this invention, dynamic noise suppression and trajectory smoothness are achieved. The Kalman filter algorithm adaptively corrects the pose estimation by modeling process noise (esophageal peristalsis, instrument motion uncertainty) and observation noise. Experiments show that this method can suppress the random noise of the electromagnetic positioning system, reduce the trajectory jitter amplitude by about 80%, and make the motion trajectory smoother and more stable.

[0025] (3) In this invention, geometric fit optimization is used to avoid the risk of perforation. The anatomical constraints of the esophageal centerline are combined with the prediction update mechanism of the Kalman filter to ensure that the optimized trajectory always fits the esophageal morphology. The measured average distance error is <1mm, completely avoiding the trajectory from penetrating the esophageal wall and significantly reducing the risk of intraoperative perforation.

[0026] (4) In the present invention, the real-time performance meets clinical needs, the algorithm complexity is low, the single iteration time is <30ms, and the overall processing delay is <50ms, which meets the strict requirements of real-time navigation in endoscopic surgery and ensures the smooth operation of doctors.

[0027] (5) In the present invention, the initial registration accuracy is improved. The rigid body registration based on the ICP algorithm (with the cardia as the reference point) is used to achieve rough alignment of the coordinate system between the electromagnetic trajectory and the esophageal centerline, eliminate redundant trajectory segments, provide highly consistent initial conditions for subsequent optimization, and reduce cumulative errors.

[0028] (6) In the present invention, the calculation is simplified to improve efficiency. The design of the state transfer matrix (3×3 unit matrix) and the observation matrix (3×3 unit matrix) fully utilizes the standard trajectory characteristics of the esophageal centerline, avoids complex transformations, reduces the computational burden, and maintains high accuracy.

[0029] (7) In the present invention, parameter optimization enhances practicality. The process noise covariance Q and the observation noise covariance R are experimentally verified to balance the suppression effect of dynamic error and measurement noise, ensuring the stability of the algorithm in different scenarios.

[0030] (8) The present invention has significant clinical value. It displays the corrected posture in real time, assisting doctors in intuitively judging the spatial relationship between the endoscope and surrounding organs, thus shortening the operation time. Through trajectory constraints and dynamic corrections, it reduces the risk of misoperation and improves the success rate of the operation. It can be extended to other endoscopic surgeries and has broad application potential. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Schematic diagram of the process of the digestive endoscope esophageal trajectory optimization method based on Kalman filtering of the present invention; Figure 2 A schematic diagram of the esophageal model and its centerline reconstructed by CT three-dimensional reconstruction in the digestive endoscope esophageal trajectory optimization method based on Kalman filtering of the present invention; Figure 3 A schematic diagram of the motion trajectory of the digestive endoscope esophagus trajectory optimization method based on Kalman filtering of the present invention; Figure 4 Schematic diagram of the trajectory before filtering of the digestive endoscope esophagus trajectory optimization method based on Kalman filtering of the present invention; Figure 5 Schematic diagram of the filtered trajectory of the digestive endoscope esophagus trajectory optimization method based on Kalman filtering of the present invention. DETAILED DESCRIPTION

[0032] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to specific embodiments.

[0033] The present invention provides a technical solution: please refer to Figure 1-5 , a digestive endoscope esophageal trajectory optimization method based on Kalman filtering, comprising the following steps: S1. Obtain the esophageal centerline. Based on the esophageal model reconstructed by preoperative CT 3D, segment the esophageal cavity using 3D computational graphics methods. Slice the esophagus at equal intervals along the axial direction (Z direction) of the esophagus, extract the geometric center point of each slice contour, and generate the esophageal centerline as the anatomical constraint benchmark for trajectory optimization.

[0034] S2. Obtain the endoscope motion trajectory based on the electromagnetic positioning system. Fix the sensor coil of the electromagnetic positioning system to the front end of the endoscope, collect its six-degree-of-freedom posture data (position and posture) in real time, generate the original motion trajectory, and intercept the esophageal segment trajectory for subsequent optimization by matching the endoscopic image with the anatomical landmarks of the cardia (such as the esophageal-gastric junction).

[0035] S3. Use ICP to initially align the esophageal centerline with the endoscope trajectory, using the iterative closest point algorithm (ICP). The ICP algorithm is currently a relatively mature rigid body registration algorithm. It calculates the optimal rotation and translation matrix by iteratively finding the nearest neighbor. The present invention uses this method to perform a rigid body transformation (rotation and translation) on the endoscope motion trajectory obtained by the electromagnetic positioning system and the esophageal centerline to achieve coarse alignment of the two coordinate systems, using the cardia as the registration reference point, eliminating redundant parts of the trajectory, and ensuring initial consistency.

[0036] S4. The Kalman filtering method is used to optimize the endoscope motion trajectory. Assuming that the posture of the endoscope in the digestive endoscopic surgical navigation system changes with time and satisfies the linear and Gaussian noise assumptions, the endoscope posture information obtained at time t by the electromagnetic positioning system can be combined with the center line of the esophagus. Considering factors such as esophageal peristalsis and the error of the electromagnetic positioning system in measuring the endoscope posture, the optimal posture at time t can be obtained, making it more suitable for actual surgical applications.

[0037] S401, State Space Modeling: State variables: define the endoscope posture as a state vector, which includes three-dimensional position coordinates and posture angles.

[0038] Process model: The process model is used to describe the standard endoscopic motion trajectory. The ideal motion trajectory can be calculated from the input motion model parameters using the state transfer matrix. The state transfer matrix is ​​the matrix form of the parameters required in the calculation process. The point in the esophageal centerline is used as the standard trajectory. No additional calculations or transformations are required, so the state transfer matrix is ​​a 3 x 3 unit matrix.

[0039] Observation model: The observation model is used to describe the actual motion trajectory of the endoscope. Since the endoscope posture obtained by the electromagnetic positioning system is not obtained in the coordinate system of the esophagus model, it is necessary to use the observation matrix to transform it to the esophageal coordinate system. The ICP algorithm is used to transform it to the inside of the esophagus and align it in the Z direction based on the point at the cardia. No additional transformation is required, so the observation matrix H is a 3x3 unit matrix.

[0040] S402, process noise ( ): Characterizes esophageal peristalsis and uncertainty of instrument movement, represents the process noise, express The distribution of is a Gaussian distribution with mean 0 and covariance matrix Q.

[0041] Observation noise ( ): reflects the measurement error of the electromagnetic positioning system, represents the observation noise, express The distribution of is a Gaussian distribution with mean 0 and covariance matrix R.

[0042] S403, iterative process of Kalman filtering: Prior estimates: Calculate the point among all points on the esophageal centerline that is closest to the endoscope position obtained by the electromagnetic positioning system at the current time t as the ideal value of the trajectory at the current time t .

[0043] Covariance prediction: , represents the prior estimated covariance of the endoscope pose at time t, and its value reflects the uncertainty of the current state estimate. It represents the posterior estimated covariance at time t-1, which is the result of the optimization update at time t-1. Q is the process noise covariance, which is a three-dimensional diagonal matrix representing the deviation of the three-dimensional coordinate measurement values ​​of the endoscope. After many experimental attempts, it was found that the three diagonal values ​​are 10 2 This order of magnitude is better. represents the transpose of the matrix F. The formula for calculating its covariance has been widely verified in past theory and practice and is considered to be a mature and effective method.

[0044] Measurement Update: Calculate Kalman gain : , The Kalman gain value at time t is used for subsequent calculations. R is the observation noise covariance, which is a three-dimensional diagonal matrix representing the deviation of the three-dimensional coordinates of the endoscope. After many experimental attempts, it was found that the three diagonal values ​​are 10 1 This order of magnitude has a better effect.

[0045] Corrected prior estimates: , Indicates the distance between the esophageal centerline at time t The three-dimensional coordinates of the nearest point, It represents the three-dimensional coordinate after Kalman filter correction at time t, Represents the three-dimensional coordinates of the endoscope obtained by the electromagnetic positioning system at time t.

[0046] Update the posterior covariance: , represents the posterior estimated covariance of the endoscope pose at time t, which is the result of optimization, where Represents a 3x3 identity matrix.

[0047] S5. Real-time display: The corrected posture is displayed in real time in the esophagus model. Since the motion trajectory of the endoscope in the esophagus obtained by its electromagnetic navigation positioning system is constrained by the center line of the esophagus, it can ensure that the trajectory is always within the lumen.

[0048] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A Kalman filter-based esophageal trajectory optimization method for digestive endoscopy, characterized in that: The following steps are involved: S1. Based on the esophageal model reconstructed by preoperative CT 3D, the geometric center point sequence of the esophageal cavity is extracted to generate the esophageal centerline; S2, collecting six-degree-of-freedom pose data of the endoscope front end in real time through the electromagnetic positioning system to generate the original motion trajectory; S3. Using an iterative closest point (ICP) algorithm, the original motion trajectory is rigidly registered with the esophageal centerline to achieve coarse alignment of the coordinate system; S4. Based on the Kalman filter algorithm, the anatomical constraints of the esophageal centerline and the posture data of the electromagnetic positioning system are integrated to dynamically correct the real-time posture of the endoscope; S5. Display the corrected endoscope posture in real time on the esophageal three-dimensional model to ensure that the trajectory is within the esophageal lumen.

2. The Kalman filter-based esophageal trajectory optimization method for digestive endoscopes according to claim 1, characterized in that: The method comprises the following steps: according to the operation steps in S1, the method for extracting the esophageal centerline is: slicing the esophagus at equal intervals along the axial direction of the esophagus, calculating the geometric center points of the contours of each slice, and connecting the center points to generate a centerline.

3. The Kalman filter-based esophageal trajectory optimization method for digestive endoscopy according to claim 1, characterized in that: The method comprises the following steps: according to the operation steps in S2, the step 2 further comprises intercepting the trajectory of the esophagus segment for optimization by matching the endoscopic image with the anatomical landmark of the cardia.

4. The Kalman filter-based esophageal trajectory optimization method for digestive endoscopes according to claim 1, characterized in that: The following steps are involved: According to the operation steps in S3, in step three, the cardia at the junction of the esophagus and the stomach is used as the registration reference point, redundant trajectory segments are eliminated, and the initial registration is completed.

5. The Kalman filter-based esophageal trajectory optimization method for digestive endoscopes according to claim 1, characterized in that: The following steps are involved: According to the operation steps in S4, the step 4 includes the following sub-steps: S401, constructing a state space model of Kalman filtering, defining the endoscope posture as a state vector, including three-dimensional position coordinates and posture angle; S402, defining the covariance matrix of process noise and observation noise to respectively characterize the dynamic error caused by esophageal peristalsis and the measurement error of the electromagnetic positioning system; S403: Calculate the optimal endoscope pose estimation value through the prediction and update iterative process of Kalman filtering.

6. The Kalman filter-based esophageal trajectory optimization method for digestive endoscopes according to claim 5, characterized in that: The following steps are involved: According to the operation steps in S401, in the state space model, the state transfer matrix and the observation matrix are both 3×3 unit matrices.

7. The Kalman filter-based esophageal trajectory optimization method for digestive endoscopes according to claim 5, characterized in that: The following steps are involved: According to the operation steps in S402, the process noise covariance matrix Q is a three-dimensional diagonal matrix with diagonal elements of the order of 10²; the observation noise covariance matrix R is a three-dimensional diagonal matrix with diagonal elements of the order of 10¹.

8. The Kalman filter-based esophageal trajectory optimization method for digestive endoscopy according to claim 5, characterized in that: The following steps are involved: According to the operation steps in S4, in the prediction step of the Kalman filter, the point on the esophageal centerline closest to the current electromagnetic positioning posture is used as the prior estimation value.

9. The method for optimizing the esophageal trajectory of a digestive endoscope based on Kalman filtering according to claim 1, characterized in that: The following steps are involved: According to the operation steps in S5, in step five, the corrected posture display delay is less than 50 milliseconds.

10. The Kalman filter-based esophageal trajectory optimization method for digestive endoscopy according to claim 1, characterized in that: The following steps are involved: According to the operation steps in S5, the correction result of the Kalman filter makes the endoscope motion trajectory fit the esophageal morphology, avoiding the trajectory from penetrating the esophageal wall.

Images