A spatial localization uncertainty evaluation method for optimizing the precision of an MRI navigation system

Abstract

The application discloses a kind of spatial positioning uncertainty evaluation methods for optimizing the precision of MRI navigation system, first utilize 3DSlicer Software to realize the three-dimensional reconstruction of MRI data, obtain the possible area of target point position.By constructing Voronoi diagram to describe the spatial relationship between target object and surrounding environment, and select the reference point outside the cognitive range to establish collection.Edge points are obtained as new growth points by sampling method, and membership function is established based on fuzzy set theory to describe the distance probability of target object.Finally, with reference object as vertex, according to eight conical direction model, the spatial region is divided, and the direction probability of target object is calculated.The direction relationship and quantitative distance relationship are combined by using joint probability, so as to correct the positioning result.The application comprehensively evaluates and optimizes the difference between cognitive coordinates and real coordinates from quantitative and qualitative aspects, reduces the spatial position uncertainty introduced by manual calibration, and is beneficial to subsequent MRI navigation system registration and application scheme design and implementation.

Description

TECHNICAL FIELD

[0001] The present application relates to a spatial positioning uncertainty evaluation method, which is suitable for clinical spatial positioning analysis and research of oral and maxillofacial and cranial and brain under magnetic resonance imaging (MRI) navigation. BACKGROUND

[0002] The MRI navigation system is a highly integrated technical platform that deeply integrates knowledge and technology in multiple disciplines such as medicine, imaging, computer vision, advanced robot technology, precise spatial positioning technology, and virtual reality interaction. The core function of the system is to achieve high-precision surgical navigation by using multi-modal image registration technology combined with manual pose calibration steps. In actual operation, medical personnel need to accurately identify and locate multiple key anatomical positions on the patient's face, such as left and right inner and outer canthi, nose tip, and left and right corners of the mouth, to complete the pose calibration work before navigation. However, there are significant individual differences in facial features and their spatial positions among different individuals, and these parts often lack clear and consistent spatial boundaries. This ambiguity leads the system to rely excessively on the subjective experience of medical personnel, and the manual calibration process inevitably introduces a certain degree of spatial position cognitive uncertainty. As uncertainty accumulates, it will directly affect the registration accuracy and application effect of the MRI navigation system, and thus pose a potential threat to the safety and accuracy of surgery.

[0003] Therefore, it is particularly important to develop a method that can effectively evaluate spatial positioning uncertainty (SPU). This method aims to provide more accurate spatial positioning feedback for the MRI navigation system by quantitatively analyzing the distribution characteristics of the key anatomical position recognition process. By continuously optimizing this evaluation mechanism, the spatial positioning accuracy of the system can be significantly improved, and the reliability and safety of surgical navigation can be effectively enhanced, which has very important significance for the development of medical imaging and surgical fields. SUMMARY

[0004] For the problem of spatial positioning uncertainty of oral and maxillofacial and cranial and brain MRI image navigation system, the present application builds an SPU evaluation method model based on the relative position description of the target point in the direction relationship and quantitative distance relationship from the quantitative and qualitative aspects of clinical spatial positioning, so as to realize the error optimization of clinical positioning of the MRI navigation system.

[0005] A spatial positioning uncertainty evaluation method for optimizing the accuracy of an MRI navigation system, which adopts the technical scheme to solve the technical problem:

[0006] S1. 3D reconstruction of MRI data is achieved using 3D Slicer software, and the possible region of the target D position description is obtained based on the reference object and orientation relationship in the image.

[0007] S2. Based on the Voronoi diagram representing spatial proximity, select and establish possible reference points T for targets outside the cognitive range around the region where D is located. i (i = 1, 2, ..., n) and establish a set neigh(D) to describe the spatial relationship between the target object and its surrounding environment.

[0008] S3. To further refine and quantify the adjacent regions, traverse the set neigh(D) to select two consecutive points that are not collinear with D, combine point D as the vertex of the Delaunay triangle, and draw the circumcircle.

[0009] S4. Extract the outer contour of the region after taking the union of all circumcircles. Edge points are obtained through sampling. As a new growth point, the new growth point evenly shares a portion of the "occupied area" OA of the graphic area where the adjacent area is located.

[0010] S5. Based on the principle of fuzzy sets, a membership function describing the "neighborhood" state can be established based on Euclidean distance and OA. This is used to represent the distance probability of the target object, thereby determining the likelihood of the target location existing in each region.

[0011] S6. Using the reference object as the vertex, divide the spatial region according to the eight conical direction models, and utilize... S represents the cone-shaped region occupied by the target object D relative to the reference object. j (N) represents a closed region in a different conical region. upper edge point The area OA of the occupied region is calculated, and the edge point OA with the largest proportion in the same conical region N is normalized to obtain the orientation probability of the target object.

[0012] S7. By combining the directional relationship and the quantitative distance relationship using joint probability, we obtain... By selecting the edge point with the highest probability value, the spatial boundary range of the target point is defined, and the positioning result is corrected.

[0013] Currently, the mainstream navigation and positioning method uses optical probes to select the location of soft and hard tissues. However, the spatial location of tissue entities without clearly defined boundaries is highly uncertain, making precise positioning difficult. Automatic localization and registration methods that extract features from panoramic images and point clouds are computationally intensive and time-consuming, and do not fundamentally optimize the cognitive errors introduced by manual calibration. This invention addresses the shortcomings of existing research by assessing the uncertainty of spatial location to improve the accuracy of MRI navigation and positioning. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of a clinical positioning optimization process based on spatial uncertainty.

[0015] Taking the outer canthus as an example, the spatial relationship is explained as follows: Figures 2-8 As shown.

[0016] Figure 2 The target object D and the area within the cognitive range.

[0017] Figure 3 This represents the location distribution of reference objects outside the cognitive range boundary of the target object D, obtained from the Voronoi diagram.

[0018] Figure 4 Delaunay triangles are created for any two adjacent reference objects that are not collinear with the target object and the target object.

[0019] Figure 5 Let be the circumcircle constructed using the three vertices of the Delaunay triangle.

[0020] Figure 6 The set of adjacent regions of the reference object based on the circumcircle of the Delaunay triangle.

[0021] Figure 7 The outer boundary of the set of adjacent regions based on the circumcircle of the Delaunay triangle is used as a reference object.

[0022] Figure 8 An example of solving for the occupied area of ​​the reference object after inserting a new growth point. Detailed Implementation

[0023] Taking the outer canthus region as an example, the present invention will be further described with reference to the accompanying drawings.

[0024] S1. Import the acquired human head MRI data into the 3D Slicer software. Use the Threshold tool in the Segment Editor module to select the epidermal layer, use the Paint tool to fill in missing parts, and the Show3D button to display the reconstructed 3D head model. Based on the reference objects and directional relationships in the images, select the possible area described by the target D position in the images according to clinical understanding, such as... Figure 2 As shown.

[0025] S2. Outside the cognitive range boundary of D, establish 7 reference points T1, T2, T3, T4, T5, T6, T7 around the Voronoi polygon of the target object D, and establish the set neigh(D) = {T1, T2, T3, T4, T5, T6, T7}, as follows: Figure 3 As shown. Because the size and shape of the region where D is located are different, the derived neighboring regions neigh(D) are also different.

[0026] S3. To address the discrepancies between adjacent regions generated by the model and the doctor's perception of these regions, two consecutive points that are not collinear with D are selected from the set neigh(D). Taking T1 and T2 as an example, point D is combined and labeled as the three vertices of a Delaunay triangle, as shown below. Figure 4 As shown. The Delaunay triangle does not contain any other reference points. Draw the circumcircle based on the three vertices of the Delaunay triangle, as follows. Figure 5 As shown. Traverse all adjacent reference points in neigh(D) and draw all circumcircles, as follows. Figure 6 As shown.

[0027] S4. Extract the outer contour of the region after taking the union of all circumcircles. like Figure 7 As shown. Edge points are obtained through sampling. As a new growth point, the new growth point evenly shares a portion of the "occupied area" OA of the graphic area where the adjacent area is located. Figure 8 China-Israel D * Taking point T7 as an example, the region enclosed by the intersection of the circumcircle of the Delaunay triangle and the Voronoi region is V[D(T7)]. * The area of ​​the region is A[D] * (T7)], the occupied area OA(T7)=V[D(T7)]∩A[D * (T7)].

[0028] S5. Based on the principle of fuzzy sets, construct a membership function representing the distance probability of the target object based on Euclidean distance and the occupied region.

[0029]

[0030] Among them, D min It is the square of the shortest distance between the new growth point and the reference object.

[0031] S6. Based on the spatial dimension, using the reference object as the vertex, uniformly expand the spatial region into eight conical directional models, utilizing... S represents the conical region occupied by the target object D relative to the reference object. j (N) represents a closed region in a different conical region. upper edge point The area OA value of the occupied region is used to quantify the distribution probability of the target object in different conical regions. The OA values ​​of all edge points within the same conical region are calculated sequentially, and the maximum value is identified and normalized to obtain the directional probability of the target object D in a specific direction.

[0032]

[0033] S7. To constrain the spatial location of the image target from both qualitative and quantitative perspectives and limit its boundary range, the directional and distance relationships are integrated to construct a more accurate joint probability model.

[0034]

[0035] By selecting the optimal value, the target position can be corrected, thereby improving the positioning accuracy of the MRI navigation system.

Claims

1. A method for assessing spatial positioning uncertainty to optimize the accuracy of an MRI navigation system, characterized in that, Includes the following steps: S1. 3D reconstruction of MRI data is achieved using 3D Slicer software, and the possible region of the target D position description is obtained based on the reference object and orientation relationship in the image. S2. Based on the Voronoi diagram representing spatial proximity, select and establish possible reference points T for targets outside the cognitive range around the region where D is located. i And establish a set neigh(D), i = 1, 2, ..., n, to describe the spatial relationship between the target object and its surrounding environment; S3. To further refine and quantify the adjacent regions, traverse the set neigh(D) to select two consecutive points that are not collinear with D, combine point D as the vertex of the Delaunay triangle, and draw the circumcircle. S4. Extract the outer contour of the region after taking the union of all circumcircles. Edge points are obtained through sampling. As a new growth point, j = 1, 2, ..., n, the new growth point uniformly shares a portion of the "occupied area" OA of the graphic region where the adjacent region is located; S5. Based on the principle of fuzzy sets, a membership function describing the "neighborhood" state can be established based on Euclidean distance and OA. This is used to represent the distance probability of the target object, thereby determining the likelihood of the target location existing in each region; S6. Using the reference object as the vertex, divide the spatial region according to eight conical models, and utilize... S represents the cone-shaped region occupied by the target object D relative to the reference object. j (N) represents a closed region in a different conical region. upper edge point The area OA of the occupied region is calculated, and the edge point OA value with the largest proportion in the same conical region is selected. This value is then normalized to obtain the orientation probability of the target object. S7. By combining the directional relationship and the quantitative distance relationship using joint probability, we obtain... By selecting the edge point with the highest probability value, the spatial boundary range of the target point is defined, and the positioning result is corrected.

2. The spatial positioning uncertainty assessment method for optimizing the accuracy of an MRI navigation system according to claim 1, characterized in that, In step S1, the target location region after 3D reconstruction of the medical image is selected: a. Import MRI data into medical imaging processing software; b. Preprocess the MRI data in image processing software, including denoising, image enhancement, and / or registration; c. Utilize the 3D reconstruction function of image processing software to generate a three-dimensional image model based on MRI data; d. In the three-dimensional image model, select one or more reference objects, which have well-defined and identifiable morphological features in the MRI image; e. Based on a selected reference object, determine one or more directional relationships in a 3D image model, the directional relationships being used to describe the spatial position of the target relative to the reference object; f. Based on the directional relationship and the characteristics of the target in the MRI data, determine the possible region of the target in the three-dimensional image model.

3. The spatial positioning uncertainty assessment method for optimizing the accuracy of an MRI navigation system according to claim 2, characterized in that, The reference objects selected in step d include anatomical structures and other prominent features in MRI images; the directional relationships determined in step e include distance, planar or spatial vectors.

4. A method for evaluating spatial positioning uncertainty to optimize the accuracy of an MRI navigation system according to claim 1 or 3, characterized in that, The generation of the Voronoi diagram in step S2 is based on the relative positional relationship between the target point and the reference object in the MRI data.

5. The method for evaluating spatial positioning uncertainty in optimizing the accuracy of an MRI navigation system according to claim 4, characterized in that, In step S4, the sampling method used is either uniform sampling or binary sampling.

6. The method for evaluating spatial positioning uncertainty in optimizing the accuracy of an MRI navigation system according to claim 5, characterized in that, The establishment of the membership function in step S5 is based on the concept of membership degree in fuzzy set theory; the division of the conical model in step S6 is based on the eight basic directions in three-dimensional space, and the opening angle and height of each conical region can be adjusted according to the actual situation.

7. The method for evaluating spatial positioning uncertainty in optimizing the accuracy of an MRI navigation system according to claim 6, characterized in that, Step S7 performs a joint calculation of the distance probability and the direction probability. By comprehensively considering the two key factors of distance and direction, the most likely location of the target point is accurately determined, providing an accurate and reliable basis for related positioning and navigation tasks.

Images