Post-processing of mandibular canal segmentation

Abstract

According to an embodiment, a method for post-processing mandibular canal segmentation data comprises: acquiring mandibular canal segmentation data including a plurality of possible mandibular canal voxels; forming a plurality of voxel structures by connecting neighboring voxels among the possible mandibular canal voxels that have a probability value exceeding a pre-configured probability value threshold; forming a plurality of roots by skeletonizing each voxel structure; for each pair of roots among the plurality of roots, checking whether the root pair satisfies a connection criterion, and in response to the root pair satisfying the connection criterion, connecting the root pair to form a second plurality of roots; and performing at least one check on at least one spatial feature of each of the second plurality of roots, and selecting at least one candidate mandibular canal pair from the second plurality of roots based on the at least one check.

Description

Technology Field

[0001] The present disclosure relates to dental imaging, and more specifically, to a method for post-processing mandibular canal segmentation data, a corresponding computing device, and a computer program product. Background Technology

[0002] The human mandible, also known as the lower jaw, is anatomically complex and is the only movable bone in the facial region that facilitates functions such as mastication, speech, and facial expressions. It also serves as a scaffold and platform for the lower dentition, muscle insertions, the temporomandibular joint, nerves, and blood vessels. Important mandibular structures are the two mandibular canals located on either side beneath the teeth in the premolar and molar regions. Each canal contains arteries and veins, as well as the inferior alveolar nerve, which is part of the mandibular branch of the trigeminal nerve that supplies motor innervation to the muscles and sensory innervation to the teeth, jaw, and lower lip. Precise localization of the mandibular canals within the lower jaw is important in dental implantology.

[0003] This summary is provided to introduce, in a simplified form, selected concepts among those further described in the detailed description below. This summary is not intended to identify core or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.

[0004] The purpose is to provide a method for post-processing mandibular canal segmentation data. The aforementioned and other purposes are achieved by the features of the independent claims. Additional forms of implementation are apparent from the dependent claims, description, and drawings.

[0005] According to a first embodiment, a method for post-processing mandibular canal segmentation data comprises: acquiring mandibular canal segmentation data including a plurality of possible mandibular canal voxels within a computerized tomographic scan of the mandible — each voxel among the plurality of possible mandibular canal voxels is associated with a probability value quantifying the probability that the voxel contains the mandibular canal —; forming a plurality of voxel structures by connecting neighboring voxels among the possible mandibular canal voxels that have a probability value exceeding a pre-configured probability value threshold; forming a plurality of roots by skeletonizing each of the plurality of voxel structures — each of the plurality of roots includes a spatial curve of a corresponding voxel structure among the plurality of voxel structures —; for each pair of roots among the plurality of roots, checking whether the root pair satisfies a concatenation criterion, and in response to the root pair satisfying the concatenation criterion, concatenating the root pair to form a second plurality of roots; The method includes the step of performing at least one check on at least one spatial feature of each of the second plurality of roots, and selecting at least one candidate mandibular canal pair from the second plurality of roots based on at least one check. The method can, for example, provide at least one candidate mandibular canal pair with improved accuracy.

[0006] In an embodiment of the first aspect, the method further comprises the step of sorting the voxels within each of the plurality of roots according to the relative positions of each voxel within the root, after forming a plurality of roots and, for each pair of roots among the plurality of roots, before examining whether the root pair satisfies the junction criteria. The method can, for example, perform other operations more efficiently when the roots are sorted, and accordingly, can provide at least one candidate mandibular canal pair with improved efficiency.

[0007] In another embodiment of the first aspect, the step of performing at least one check on at least one spatial feature of each of the second plurality of roots and selecting at least one candidate mandibular canal pair from the second plurality of roots based on at least one check comprises: determining the orientation for a first segment of the root; checking, for each of the plurality of segments of the root other than the first segment of the root, whether the orientation of the segment is within a pre-configured tolerance from the orientation of the first segment of the root; calculating a monotonicity score based on the number of segments among the plurality of segments that are not within the tolerance; and selecting at least one candidate mandibular canal pair from the second plurality of roots based on at least the monotonicity score. The method may, for example, utilize orientation information of the roots when selecting the candidate mandibular canal pair, and accordingly, can provide at least one candidate mandibular canal pair with improved accuracy.

[0008] In another embodiment of the first aspect, the step of performing at least one check on at least one spatial feature of each of the second plurality of roots and selecting at least one candidate mandibular canal pair from the second plurality of roots based on at least one check comprises: for at least one root pair among the second plurality of roots, the step of calculating at least one pair score ― at least one pair score includes: an average coordinate score for the root pair ― the average coordinate score is calculated by comparing the average coordinates in the coordinate axis direction of each of the roots among the root pair ―, a symmetry score for the root pair quantifying the spatial symmetry of the root pair, and / or at least one of a symmetry plane score for the root pair ―; the step of calculating a total score for at least one root pair among the second plurality of roots based on at least one pair score of the pair; and the step of selecting at least one candidate mandibular canal pair from the second plurality of roots based on at least the total score. The method may, for example, utilize symmetry information of the root pairs when selecting the candidate mandibular canal pair, and accordingly, may provide at least one candidate mandibular canal pair with improved accuracy.

[0009] In another embodiment of the first aspect, the step of performing at least one check on at least one spatial feature of each of the second plurality of roots and selecting at least one candidate mandibular canal pair from the second plurality of roots based on at least one check comprises: selecting a root subset from the second plurality of roots based on the monotonicity score of each root; calculating at least one pair score for each root pair within the selected root subset; calculating a total score for each root pair within the selected subset based on the at least one pair score of the root pairs; and selecting a root pair having the highest total score within the selected root subset as a candidate mandibular canal pair. The method may, for example, utilize orientation information of the roots and symmetry information of the root pairs when selecting the candidate mandibular canal pair, and accordingly, can provide at least one candidate mandibular canal pair with improved accuracy.

[0010] In another embodiment of the first aspect, the step of performing at least one check on at least one spatial feature of each of the second plurality of roots and selecting at least one candidate mandibular canal pair from the second plurality of roots based on at least one check comprises: selecting a root subset from the second plurality of roots based on a monotonicity score of each root that quantifies how well the orientation of each root is generally aligned with the orientation of the actual mandibular canal in relation to spatial monotonicity; calculating at least one pair score for each root pair within the selected root subset; calculating a total score for each root pair within the selected subset based on at least one pair score of the root pairs; and selecting the root pair having the best total score within the selected root subset as a candidate mandibular canal pair. The method may, for example, utilize orientation information of the roots and symmetry information of the root pairs when selecting the candidate mandibular canal pair, and accordingly, can provide at least one candidate mandibular canal pair with improved accuracy.

[0011] In another embodiment of the first aspect, the coordinate axis direction is parallel to the width direction of the computerized tomographic scan of the mandible. The method may, for example, utilize width (left-right) symmetry information of the roots when selecting candidate mandibular canal pairs, and accordingly, provide at least one candidate mandibular canal pair with improved accuracy.

[0012] In another embodiment of the first aspect, the symmetry score and / or symmetry plane score are calculated based on digests of the roots among the root pair, wherein the digest of the roots includes the starting position of the roots, the ending position of the roots, and the average position of the roots. The method can, for example, efficiently utilize the symmetry information of the roots and, accordingly, provide at least one candidate mandibular canal pair with improved accuracy and efficiency.

[0013] In another embodiment of the first aspect, the symmetry plane score quantifies how well the symmetry plane of the root pair is aligned with the height direction of a computerized tomographic scan of the mandible. The method may, for example, utilize the symmetry plane information of the roots and, accordingly, provide at least one candidate mandibular canal pair with improved accuracy.

[0014] In another embodiment of the first aspect, the junction criterion includes the distance between the endpoint of the first root of the root pair and the endpoint of the second root of the root pair being less than a pre-configured maximum threshold distance. The method can efficiently determine, for example, whether two roots should be junctioned, and accordingly, can provide at least one candidate mandibular canal pair with improved accuracy and efficiency.

[0015] In another embodiment of the first aspect, for each root pair among a plurality of roots, the step of examining whether the root pair satisfies a junction criterion comprises: calculating at least one of the minimum distance between the endpoints of the root pair, the minimum distance between the extrapolated endpoints of the root pair — the extrapolated endpoints are obtained by linearly extrapolating the endpoints with a pre-configured number of voxels — and / or the skew distance of the root pair, wherein the junction criterion comprises the minimum distance between the endpoints satisfying a first criterion, the minimum distance between the extrapolated endpoints satisfying a second criterion, and / or the skew distance of the root pair satisfying a third criterion. The method can determine, for example, whether two roots should be junctioned based on these criteria, and accordingly, can provide at least one candidate mandibular canal pair with improved accuracy.

[0016] In another embodiment of the first aspect, each of the plurality of voxel structures is formed using a connected-component algorithm. The method can, for example, efficiently form the voxel structures using a connected-component algorithm and further utilize the characteristics of the connected components, and accordingly, provide at least one candidate mandibular canal pair with improved accuracy and efficiency.

[0017] In another embodiment of the first aspect, mandibular canal segmentation data is obtained as the output of a trained neural network. The method may, for example, post-process the mandibular canal segmentation data and, accordingly, provide at least one candidate mandibular canal pair with improved accuracy compared to the neural network alone.

[0018] According to a second embodiment, a computing device comprises at least one processor and at least one memory including computer program code, and the at least one memory and computer program code are configured together with at least one processor to enable the computing device to perform a method according to a first embodiment.

[0019] According to a third embodiment, a computer program product includes program code configured to perform a method according to a first embodiment when the computer program product is executed on a computer.

[0020] Many of the accompanying features will be better understood and more easily recognized by referring to the detailed description below, which is considered in relation to the attached drawings. Brief explanation of the drawing

[0021] In the following, embodiments are described in more detail with reference to the accompanying drawings and drawings. FIG. 1 illustrates a flowchart representation of a method according to an embodiment. FIG. 2 illustrates a schematic representation of a computing device according to an embodiment. FIG. 3 illustrates a schematic representation of a plurality of possible mandibular canal voxels and corresponding voxel structures according to an embodiment. FIG. 4 illustrates a schematic representation of voxel structures and corresponding roots according to an embodiment. FIG. 5 illustrates a schematic representation of voxel structures as connected components according to an embodiment. FIG. 6 illustrates a flowchart representation of a connection procedure according to an embodiment. FIG. 7 illustrates a schematic representation of the distances between the roots according to an embodiment. FIG. 8 illustrates a schematic representation of root segments according to an embodiment. FIG. 9 illustrates a schematic representation of the symmetry score calculation according to an embodiment. FIG. 10 illustrates a flowchart representation of a candidate mandibular canal pair selection procedure according to an embodiment. FIG. 11 illustrates a schematic representation of a convolutional neural network according to an embodiment. In the following, similar reference numbers are used to designate similar parts in the attached drawings. Specific details for implementing the invention

[0022] In the following description, reference is made to the accompanying drawings, which form part of the present disclosure, in which specific embodiments in which the present disclosure may be arranged are illustrated by way of example. It is understood that other embodiments may be utilized and that structural or logical modifications may be made without departing from the scope of the present disclosure. Accordingly, since the scope of the present disclosure is defined by the appended claims, the following detailed description should not be taken in a restrictive sense.

[0023] For example, it is understood that the disclosure relating to the method described may also be valid for a corresponding device or system configured to perform the method, and vice versa. For example, where a specific method step is described, the corresponding device may include a unit for performing the described method step even if such unit is not explicitly described or illustrated in the drawings. On the other hand, for example, where a specific device is described based on functional units, the corresponding method may include a step for performing the described function even if such step is not explicitly described or illustrated in the drawings. Additionally, it is understood that features of the various exemplary embodiments described herein may be combined with one another unless specifically stated otherwise.

[0024] FIG. 1 illustrates a flowchart representation of a method according to an embodiment.

[0025] According to an embodiment, the method (100) includes the step (101) of acquiring mandibular canal segmentation data including a plurality of possible mandibular canal voxels within a computerized tomographic scan of the mandible, wherein each of the plurality of possible mandibular canal voxels is associated with a probability value that quantifies the probability that the voxel contains the mandibular canal.

[0026] The mandible may also be referred to as the lower jaw, jawbone, or similar.

[0027] Here, a voxel can represent a value on a regular grid in three-dimensional space. A voxel represents a sample or data point on a regularly spaced three-dimensional grid. A voxel can represent a single point on this grid. A data point may include a single data piece or multiple data pieces, such as a probability value that quantifies the probability that the voxel includes the mandibular canal. The probability value may not include the probability itself. Rather, the probability value may include any quantity that quantifies the probability.

[0028] A mandibular canal voxel may refer to a voxel corresponding to an area of ​​a computed tomography scan containing the mandibular canal. Similarly, any of the multiple possible mandibular canal voxels may correspond to an area of ​​a computed tomography scan containing the mandibular canal, and the probability associated with each voxel quantifies the probability that the area of ​​the voxel contains the mandibular canal.

[0029] Here, the geometry of the mandible and / or mandibular canals in a computerized tomographic scan can be described by referring to depth, width, and height dimensions / directions. The depth dimension / direction may also be referred to as the anterior-posterior dimension / direction, the z dimension, or similar. The depth dimension / direction may be perpendicular to the coronal plane. The width dimension / direction may also be referred to as the left-right dimension / direction, the x dimension, or similar. The width dimension / direction may be perpendicular to the sagittal plane. The height dimension / direction may also be referred to as the inferior-superior dimension / direction, the y dimension, or similar. The height dimension / direction may be perpendicular to the horizontal / axial / transverse plane. The zero points of the x-axis, y-axis, and / or z-axis may be positioned so as to be in the middle of the scan of the mandible.

[0030] The method (100) may further include the step (102) of forming a plurality of voxel structures by connecting neighboring voxels having a probability value exceeding a pre-configured probability threshold among possible mandibular canal voxels.

[0031] Here, the voxel structure may include any structure comprising a plurality of connected voxels. For example, the voxel structure may be implemented as a connected component as disclosed herein. The voxel structure may include a plurality of connected voxels. Accordingly, each voxel within the voxel structure is adjacent to at least one other voxel within the same voxel structure.

[0032] The pre-configured probability threshold may be, for example, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, or 0.9.

[0033] Small voxel structures can be discarded after the voxel structures are formed. For example, voxel structures smaller in size (i.e., the number of voxels within the voxel structure) than the pre-configured percentage of the largest voxel structure can be discarded. For example, Voxel structures (C) having can be maintained, where T is a threshold such as 0.01, 0.02, 0.03, 0.05, or 0.1. A smaller threshold may be selected to include more candidate voxels and thus not ignore small voxel structures.

[0034] The method (100) may further include the step (103) of forming a plurality of routes by skeletonizing each of the plurality of voxel structures, wherein each of the plurality of routes includes a spatial curve of a corresponding voxel structure among the plurality of voxel structures.

[0035] Each of the multiple routes can be obtained by skeletonizing the corresponding voxel structure among the multiple voxel structures.

[0036] The method (100) may further include, for each pair of roots among a plurality of roots, a step (104) of checking whether the root pair satisfies a connection criterion, and in response to the root pair satisfying the connection criterion, a step of connecting the root pair to form a second plurality of roots.

[0037] Here, connecting two roots may include creating a new root by connecting two roots end-to-end.

[0038] The method (100) may further include the step (105) of performing at least one check on at least one spatial feature of each of the second plurality of roots, and the step of selecting at least one candidate mandibular canal pair from the second plurality of roots based on at least one check.

[0039] Most mandibular canals are directed downward and medially. At least one check for at least one spatial feature of each of the second plurality of roots can use this and other information regarding the geometry of the mandibular canals to select at least one candidate pair of mandibular canals from the second plurality of roots.

[0040] The method (100) may further include the step of providing at least one candidate mandibular canal pair. Providing may include, for example, providing at least one candidate mandibular canal pair to a user. At least one candidate mandibular canal pair may be provided to a user, for example, by displaying corresponding roots to the user on a display. The roots may be, for example, overlaid on a computerized tomographic scan of the mandible. In some embodiments, the major points of the roots may be samples, and the major points may be provided.

[0041] At least some embodiments of the method (100) may also consider small root segments, otherwise the small root segments will be ignored.

[0042] At least some embodiments of the method (100) may take into account the shapes of the extracted roots, and accordingly, other anatomical structures and even incorrectly labeled data may be ignored.

[0043] The method (100) can construct accurate segmentation of mandibular nerve canals from incomplete segmentation data that may include gaps and false positives. The method (100) can work on both cone beam computed tomography (CBCT) and computed tomography (CT) data. The method (100) can construct all possible canals, and then all possible canals can be selected based, for example, root coordinate monotonicity and pairwise symmetry.

[0044] The method (100) can first construct all possible mandibular canals proposed by mandibular canal segmentation data, and then use heuristic methods to screen out false canals. Subsequently, the method (100) can use checks to screen out roots / root pairs that are unlikely to correspond to mandibular canal roots. These checks can take into account the spatial characteristics of the mandibular canal roots.

[0045] FIG. 2 illustrates a schematic representation of a computing device (200) according to an embodiment.

[0046] According to an embodiment, the computing device (200) includes at least one processor (201) and at least one memory (202) containing computer program code.

[0047] At least one memory (202) and computer program code can be configured to enable a computing device (200) to perform a method (100) together with at least one processor (201).

[0048] At least one processor (201) may include one or more of various processing devices, such as a central processing unit (CPU), a graphics processing unit (GPU), a co-processor, a microprocessor, a processing unit, a digital signal processor (DSP), a processing circuit with or without a DSP, or various other processing devices such as an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), a microprocessor unit (MCU), a hardware accelerator such as a neural network accelerator, a special purpose computer chip, etc.

[0049] At least one memory (202) may be configured to store, for example, computer programs, etc. At least one memory (202) may include one or more volatile memory devices, one or more non-volatile memory devices, and / or a combination of one or more volatile memory devices and non-volatile memory devices. For example, at least one memory (202) may be implemented as magnetic storage devices (e.g., hard disk drives, floppy disks, magnetic tapes, etc.), optomechanical storage devices, and semiconductor memories (e.g., mask ROM, PROM (programmable ROM), EPROM (erasable PROM), flash ROM, RAM (random access memory), etc.).

[0050] The computing device (200) may include other components not exemplified in the embodiment of FIG. 2. The computing device (200) may include, for example, an input / output bus for connecting the computing device (200) to other devices.

[0051] When the computing device (200) is configured to implement some functions, some components of the computing device (200), such as at least one processor (201) and / or at least one memory (202), may be configured to implement these functions. Furthermore, when the at least one processor (201) is configured to implement some functions, these functions may be implemented, for example, using program code contained in at least one memory (202).

[0052] FIG. 3 illustrates a schematic representation of a plurality of possible mandibular canal voxels and corresponding voxel structures according to an embodiment.

[0053] Multiple voxel structures can be formed by connecting neighboring voxels among possible mandibular canal voxels that have a probability value exceeding a pre-configured probability threshold. For example, in the embodiment of FIG. 3, the probability threshold is set to 0.5. Accordingly, the voxel structures (302_1, 302_2) exemplified in the embodiment of FIG. 3 are formed by connecting neighboring voxels among possible mandibular canal voxels (301) that have a probability value exceeding 0.5.

[0054] The rule used to determine whether two voxels are adjacent may differ in different embodiments. For example, in the embodiment of FIG. 3, for each voxel, eight nearest voxels are considered to be adjacent voxels. If only four nearest voxels are considered to be adjacent voxels, the second voxel structure (302_2) will not be formed. In three dimensions, for each voxel, for example, six other voxels in contact with the voxel, or 26 other voxels within a 3x3x3 voxel cube, may be considered to be adjacent voxels.

[0055] It should be recognized that the embodiment of FIG. 3 is merely a two-dimensional representation of a simplified example illustrating only two voxel structures (302_1, 302_2).

[0056] FIG. 4 illustrates a schematic representation of voxel structures and corresponding roots according to an embodiment.

[0057] Each route can be obtained by skeletonizing a voxel structure. Therefore, each route may include a skeleton of a corresponding voxel structure.

[0058] Here, the skeleton of a voxel structure can refer to a thin version of the voxel structure that is equidistant from the boundaries of the voxel structure. Thus, the roots can highlight the geometric and topographical characteristics of the voxel structure, such as connectivity, topography, length, orientation, and width. Along with the distances of its points to the shape boundaries, the skeleton can also serve as a representation of the voxel structure.

[0059] In the embodiment of FIG. 4, each root (303_1, 303_2) is obtained by skeletonizing the corresponding voxel structure (302_1, 302_2).

[0060] According to an embodiment, the method (100) further includes the step of sorting the voxels within each of the plurality of roots according to the relative positions of each voxel within the root after forming a plurality of roots and, for each pair of roots among the plurality of roots, before checking whether the pair of roots satisfies the connection criteria.

[0061] For example, two nearest neighbor graphs for voxels within R can be constructed. The first graph (G s ) can have a smaller radius to preserve the root terrain, and the second graph (G l ) can have a larger radius for neighbor search. Here, the radius measures which voxels are considered neighbors of a given voxel.

[0062] For example, breadth-first search (BFS) is G s Based on, the longest path from the root(R)( It can be used to find the longest path. ) can be used as a sorted root. Path( When ) is not sufficiently long (e.g., In this case, here, is the length of the path / root and T r is the threshold), longest path( ) is the longest path( Until ) becomes sufficiently long (e.g., In the case of) repeatedly G l It can be extended by adding nearest neighbor voxels based on. Additional constraints may be used for the nearest neighbor voxels to be added. For example, the y component (v) of the voxel to be added y ) is the longest path( The y-component of the last voxel within ) It may need to be larger than ), where the y-axis follows the anterior-posterior direction of the computed tomography scan of the mandible, and larger y values ​​point in the posterior direction. If such voxels cannot be found, the constraint is, for example, It can be mitigated as, where, T y is the threshold value.

[0063] Additionally, the first and last voxels within the aligned roots can be guaranteed to be true endpoints by checking their neighbor voxels.

[0064] FIG. 5 illustrates a schematic representation of voxel structures as connected components according to an embodiment.

[0065] According to an embodiment, each of the plurality of voxel structures is formed using a linked component algorithm. This may also be referred to as linked component labeling (CCL) or linked component analysis (CCA).

[0066] The connected component algorithm can construct a graph (325) comprising vertices (321) and connected edges (320) based on a plurality of possible mandibular canal voxels (301). Each vertex within the graph (325) may correspond to a voxel. The edges (320) may represent connected neighboring voxels. The connected component algorithm can construct a plurality of connected components (312_1, 312_2). A connected component is a subgraph in which any two vertices (321) are connected to each other by paths and are not connected to any additional vertices in the rest of the graph (325). Each connected component may correspond to the voxel structure of the connected voxels.

[0067] The embodiment of FIG. 5 illustrates connected components (312_1, 312_2) corresponding to the voxel structures (302_1, 302_2) illustrated in the embodiments of FIG. 3 and FIG. 4, respectively.

[0068] FIG. 6 illustrates a flowchart representation of a connection procedure according to an embodiment.

[0069] The procedure can be started in operation (501).

[0070] In operation (502), a score for each pair of roots among a plurality of roots can be calculated. This score may be referred to as a contiguous score or similar.

[0071] In operation (503), it can be checked whether a valid proposal for a pair of roots to be connected is provided by score calculation (502). For example, if the pair of roots does not satisfy the connection criteria, no valid proposal is provided. If no valid proposal is provided, the procedure may proceed to operation (505) where the procedure ends. If a valid proposal is provided, the procedure may proceed to operation (504).

[0072] According to an embodiment, the connection criterion includes the distance between the endpoint of the first route among the route pairs and the endpoint of the second route among the route pairs being less than a pre-configured maximum threshold distance.

[0073] For example, in operation (502), a connection score may be calculated for each root pair based on the distance between the endpoint of the first root of the root pair and the endpoint of the second root of the root pair. Additionally or alternatively, other criteria, such as those disclosed herein, may be considered when calculating the connection score. Subsequently, in operation (503), whether the connection score of any root pair exceeds a pre-configured minimum connection score, that is, whether the calculation has provided a valid proposal of the roots to be connected. Alternatively or additionally, if no valid proposal is found, the score calculation (502) itself may return a special value, such as null, and this value may be detected in operation (503).

[0074] In operation (504), a pair of roots with the highest score may be connected. After connection, the procedure may return to operation (502), where the score for each pair of roots may be calculated. Due to the connection performed in operation (504), multiple roots now include different multiple roots. Therefore, operations (502-504) may be repeated until no valid connection proposal is provided.

[0075] As a result of the procedure (104), a second plurality of routes may be provided.

[0076] FIG. 7 illustrates a schematic representation of the distances between the roots according to an embodiment.

[0077] According to an embodiment, for each pair of roots among a plurality of roots, checking whether the root pair satisfies the connection criteria comprises: calculating at least one of the minimum distance between the endpoints of the root pair, the minimum distance between the extrapolated endpoints of the root pair — the extrapolated endpoints are obtained by linearly extrapolating the endpoints with a pre-configured number of voxels — and / or the skew distance of the root pair.

[0078] For each pair of roots, the minimum distance between endpoints belonging to different roots in the pair can be calculated. There are a total of 2 x 2 = 4 cases. The minimum distance can be denoted as d.

[0079] The minimum distance between the extrapolated endpoints of a pair of roots can be obtained, for example, by linearly extrapolating the endpoints of each root with k voxels. The minimum distance is calculated and d ext It can be written as.

[0080] The skew distance of a pair of roots can refer to the shortest distance between lines formed by connecting the endpoints of a root to its extrapolation. One purpose of skew distance is to check whether two roots intersect when extrapolated, while two other criteria quantify whether the two roots proceed in opposite directions. The skew distance is d skew It can be written as.

[0081] The connection criteria may include the minimum distance between endpoints satisfying a first criterion, the minimum distance between extrapolated endpoints satisfying a second criterion, and / or the skew distance of the root pair satisfying a third criterion.

[0082] For example, d ext < T1, d - d ext T2 and d skew If < T3, the roots can be contiguous, where T iare pre-configured thresholds. If these criteria are satisfied, d ext This can be provided as a connection score in operation (502) of the embodiment of FIG. 6, for example. If these criteria are not satisfied, a special value such as null may be provided to indicate this. Subsequently, a pair of roots to be connected may be selected based on the connection score of each pair of roots, for example, as disclosed in the embodiment of FIG. 6.

[0083] FIG. 8 illustrates a schematic representation of root segments according to an embodiment.

[0084] According to an embodiment, performing at least one check on at least one spatial feature of each of the second plurality of roots (303) comprises: determining the orientation for the first segment (701) of the root (303); for each segment (702) among the plurality of segments of the root (303) other than the first segment (701) of the root (303), checking whether the orientation of the segment (702) is within a pre-configured tolerance from the orientation of the first segment (701) of the root (303); calculating a monotonicity score based on the number of segments among the plurality of segments that are not within the tolerance; and selecting at least one candidate mandibular canal pair from the second plurality of roots based on the monotonicity score.

[0085] The first segments (701) may correspond to any segment within the root (303). For example, in the embodiment of FIG. 8, the first segment (701) is located at one end of the root (303). In other embodiments, the first segment (701) may be located at any position along the root (303).

[0086] A plurality of segments may include any number of segments. For example, in the embodiment of FIG. 8, the plurality of segments includes all segments other than the first segment. In other embodiments, the plurality of segments may include only a smaller subset of all segments within the root (303), for example.

[0087] For example, near the start of the route (303), the x (width), y (height), and / or z (depth) coordinates may be checked in which directions they are progressing (increasing or decreasing). In some embodiments, only some of these coordinates may be considered. For example, only the x and z components may be of interest. To achieve greater robustness, some simple voting strategy may be applied here. Subsequently, for all other segments (702), it may be checked whether the route (303) is progressing along this direction. Some tolerances and some number of violations may be allowed.

[0088] FIG. 9 illustrates a schematic representation of the symmetry score calculation according to an embodiment.

[0089] According to an embodiment, performing at least one check for at least one spatial feature of each of the second plurality of routes comprises: for at least one pair of routes among the second plurality of routes, calculating at least one pair score, wherein the at least one pair score comprises: an average coordinate score for the route pair — the average coordinate score is calculated by comparing the average coordinates in the coordinate axis direction of each of the routes among the route pair —; a symmetry score for the route pair quantifying the spatial symmetry of the route pair; and / or at least one of a symmetry plane score for the route pair.

[0090] The total score for at least one pair of roots among the second plurality of roots can be calculated based on at least one pair score of the pair. At least one candidate pair of mandibular canals from the second plurality of roots can be selected based on at least the total score.

[0091] According to an embodiment, the symmetry score and / or symmetry plane score are calculated based on the digests of the roots among the root pair, wherein the digest of the roots includes the starting position of the roots, the ending position of the roots, and the average position of the roots.

[0092] A digest (801) of the first route, a digest (802) of the second route, and a symmetry plane (803) determined based on the digests (801, 802) are illustrated in the embodiment of FIG. 9.

[0093] The symmetry score can be calculated, for example, based on the average of the displacements (804) of the digests (801, 802) from the symmetry plane (803). Each displacement (804) can quantify how perpendicular a line (805) drawn from a point in the first digest (801) to a corresponding point in the second digest (802) is to the symmetry plane (803). For example, relative arrangements of acute angles or sine values ​​between the lines (805) and the symmetry plane (803) may be used. When the line (805) is perpendicular to the symmetry plane (803), the displacement may be zero. In other embodiments, the symmetry score may be calculated in a different way.

[0094] According to the embodiment, the symmetry plane score quantifies how well the symmetry plane (803) of the root pair is aligned with the height direction of the computerized tomography scan of the mandible. Thus, the symmetry plane score can quantify the left-right symmetry of the root pair.

[0095] According to the embodiment, the coordinate axis direction is parallel to the width direction of the computerized tomography scan of the mandible.

[0096] Therefore, the total score can reflect how well the root pair follows the following observations of the accurate mandibular canal root pair: the z-coordinates of the two roots should be close to the mean value, the two roots should be approximately symmetric, the plane of symmetry should follow the height (y) direction, and the normal vector of the plane of symmetry should follow the width (x) direction.

[0097] FIG. 10 illustrates a flowchart representation of a candidate mandibular canal pair selection procedure according to an embodiment.

[0098] According to an embodiment, selecting at least one candidate mandibular canal pair from a second plurality of roots based on at least one check includes performing at least some of the operations (901-904).

[0099] In operation (901), a subset of routes from the second plurality of routes can be selected based on the monotony score of each route.

[0100] A monotonicity score may be calculated for each of the second plurality of routes in the manner disclosed herein. A subset may be selected, for example, by selecting all routes having a monotonicity score greater than a critical monotonicity score.

[0101] If fewer than two routes remain after operation (901), for example, if fewer than two routes have monotonicity scores exceeding the threshold monotonicity score, some routes with monotonicity below the threshold may be selected for the next operations. Still, routes with the highest scores will be selected first.

[0102] In operation (902), at least one pair score can be calculated for each pair of roots within a selected root subset.

[0103] At least one pair score can be calculated for each pair of roots within a selected root subset in the manner disclosed herein.

[0104] In operation (903), a total score can be calculated for each root pair within a selected subset based on at least one pair score of the root pairs.

[0105] The total score can be calculated for each pair of roots within a selected root subset in the manner disclosed herein.

[0106] In operation (904), a root pair having the highest total score within a selected root subset can be selected as a candidate mandibular canal pair.

[0107] Accordingly, in the embodiment of FIG. 10, a monotonicity score can be used to filter the roots, and then, a candidate mandibular canal pair can be selected from a subset of the remaining roots based on the total score. Thus, the embodiment of FIG. 10 can select at least one candidate mandibular canal pair from a second plurality of roots based on the monotonicity score and the total score.

[0108] FIG. 11 illustrates a schematic representation of a convolutional neural network according to an embodiment.

[0109] According to an embodiment, mandibular canal segmentation data is obtained as the output of a trained neural network.

[0110] Here, the term "neural network" is used to refer to artificial neural networks.

[0111] In the embodiments of FIG. 11, the neural network includes a convolutional neural network (CNN). The CNN includes four types of convolutional layers. The first type (1001) includes a convolution, a convolutional kernel of size 3x3x3 with a stride of 1, batch normalization (BN), and a rectified linear unit (ReLU) activation function. The second type (1002) includes a convolution, a convolutional kernel of size 3x3x3 with a stride of 2, a BN, and a ReLU activation function. The third type (1003) includes a transpose convolution, a convolutional kernel of size 3x3x3 with a stride of 2, a BN, and a ReLU activation function. The fourth type (1004) includes a convolution, a convolutional kernel of size 1x1x1 with a stride of 1, and a sigmoid activation function. The CNN also includes skip connections as exemplified in Fig. 11. Some of the skip connections are element-wise sums, and some are feature connections. The number of channels within each layer is shown in Fig. 11.

[0112] The embodiment of FIG. 11 is merely an exemplary implementation of a CNN capable of providing mandibular canal segmentation data. Alternatively, mandibular canal segmentation data may be provided by any other type of data processing, such as any type of appropriately trained machine learning model, or an iterative machine learning model that may not need to be trained with data.

[0113] Any range or device value given herein may be extended or modified without loss of the desired effect. Additionally, any embodiment may be combined with other embodiments unless explicitly permitted.

[0114] Although the subject matter has been described in specific language regarding structural features and / or acts, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as examples embodying the claims, and other equivalent features and acts are intended to be within the scope of the claims.

[0115] It will be understood that the benefits and advantages described above may relate to one embodiment or to multiple embodiments. The embodiments are not limited to those that solve any or all of the stated problems or those that have any or all of the stated benefits and advantages. It will be further understood that a reference to 'one' item may refer to one or more of these items.

[0116] The steps of the methods described herein may be performed in any suitable order or, where appropriate, simultaneously. Additionally, individual blocks may be omitted from any of the methods without departing from the spirit and scope of the subject matter of the invention described herein. Aspects of any of the embodiments described above may be combined with aspects of any of the other embodiments described to form additional embodiments without losing the desired effect.

[0117] The term 'comprising' is used herein to mean including the identified methods, blocks, or elements, but such blocks or elements do not include an exclusive list, and the method or device may include additional blocks or elements.

[0118] The foregoing description is given merely as an example, and it will be understood that various modifications may be made by those skilled in the art. The foregoing specification, examples, and data provide a complete description of the structure and use of exemplary embodiments. Although various embodiments have been described above with some degree of specificity or by reference to one or more individual embodiments, those skilled in the art may make numerous modifications to the disclosed embodiments without departing from the spirit or scope of this specification.

Claims

Claim 1 A method (100) for post-processing of mandibular canal segmentation data using a computing device, comprising the step (101) of acquiring mandibular canal segmentation data including a plurality of possible mandibular canal voxels (301) in a computerized tomographic scan of the mandible, wherein each of the plurality of possible mandibular canal voxels (301) is associated with a probability value that quantifies the probability that the voxel (301) includes the mandibular canal; The method comprises the step (102) of forming a plurality of voxel structures (302_1, 302_2) by connecting neighboring voxels (301) having a probability value exceeding a pre-configured probability value threshold among possible mandibular canal voxels (301), and the step (103) of forming a plurality of roots by skeletonizing each of the plurality of voxel structures (302_1, 302_2), wherein each of the plurality of roots (303, 303_1, 303_2) includes a spatial curve of the corresponding voxel structure (302_1, 302_2) among the plurality of voxel structures (302_1, 302_2); for each of the plurality of roots (303, 303_1, 303_2), the root pair A method (100) further comprising: a step (104) of examining whether a concatenation criterion is satisfied and, in response to the concatenation criterion being satisfied, concatenating the root pair to form a second plurality of roots (303, 303_1, 303_2); and a step (105) of performing at least one check on at least one spatial feature of each of the second plurality of roots (303, 303_1, 303_2) and selecting at least one candidate mandibular canal pair from the second plurality of roots (303, 303_1, 303_2) based on at least one check. Claim 2 A method (100) according to claim 1, further comprising the step of sorting the voxels (301) within each of the plurality of routes (303, 303_1, 303_2) according to the relative position of each voxel (301) within the route (303, 303_1, 303_2) after forming a plurality of routes (303, 303_1, 303_2) and before checking whether the route pair satisfies the connection criteria for each of the route pairs. Claim 3 In claim 1 or 2, the step of performing at least one check on at least one spatial feature of each of the second plurality of roots (303, 303_1, 303_2) and selecting at least one candidate mandibular canal pair from the second plurality of roots (303, 303_1, 303_2) based on at least one check comprises: determining the orientation for the first segment (701) of the root (303, 303_1, 303_2); for each segment (702) among the plurality of segments (702) of the root (303, 303_1, 303_2) other than the first segment (701) of the root (303, 303_1, 303_2), the orientation of the segment (702) of the root (303, 303_1, 303_2) A method (100) comprising: a step of checking whether there is a pre-configured tolerance from the direction of a first segment (701); a step of calculating a monotonicity score based on the number of segments (702) among a plurality of segments (702) that are not within the tolerance; and a step of selecting at least one candidate mandibular canal pair from a second plurality of roots (303, 303_1, 303_2) based on at least the monotonicity score. Claim 4 In claim 1 or 2, the step of performing at least one check on at least one spatial feature of each of the second plurality of roots (303, 303_1, 303_2) and selecting at least one candidate mandibular canal pair from the second plurality of roots (303, 303_1, 303_2) based on at least one check is the step of calculating at least one pair score for at least one root pair among the second plurality of roots (303, 303_1, 303_2), wherein the at least one pair score comprises an average coordinate score for the root pair — the average coordinate score is calculated by comparing the average coordinates in the coordinate axis direction of each of the roots (303, 303_1, 303_2) among the root pair — a symmetry score for the root pair quantifying the spatial symmetry of the root pair with respect to the symmetry plane (803) of the root pair, and / or symmetry for the root pair A method (100) comprising at least one of the following steps: a plane (803) score—the plane (803) score quantifies how well the plane (803) of the plane of the plane of the plane of the plane of the plane of the plane of the plane of the plane of the plane of the plane of the plane of the plane of the plane of the plane of the plane of the plane of the plane of the plane of the plane of the plane quantifies how well the plane of Claim 5 In claim 4, the step of performing at least one check on at least one spatial feature of each of the second plurality of roots (303, 303_1, 303_2) and selecting at least one candidate mandibular canal pair from the second plurality of roots (303, 303_1, 303_2) based on at least one check comprises: a step (901) of selecting a subset of roots (303, 303_1, 303_2) from the second plurality of roots (303, 303_1, 303_2) based on the monotonicity score of each root (303, 303_1, 303_2); a step (902) of calculating at least one pair score for each root pair within the selected root (303, 303_1, 303_2) subset; and, based on at least one pair score of the root pair, the selected subset A method (100) comprising: a step (903) of calculating a total score for each root pair within; and a step (904) of selecting a root pair having the best total score within a subset of selected roots (303, 303_1, 303_2) as a candidate mandibular canal pair. Claim 6 In paragraph 4, the coordinate axis direction is parallel to the width direction of the computerized tomography scan of the mandible, method (100). Claim 7 In claim 4, the symmetry score and / or symmetry plane (803) score is calculated based on digests (801, 802) of the roots (303, 303_1, 303_2) of the root pair, and the digest of the roots (303, 303_1, 303_2) includes the starting position of the roots (303, 303_1, 303_2), the ending position of the roots (303, 303_1, 303_2), and the average position of the roots (303, 303_1, 303_2), method (100). Claim 8 A method (100) according to claim 1 or 2, wherein the connection criterion comprises the distance between the endpoint of the first route (303, 303_1, 303_2) of the route pair and the endpoint of the second route (303, 303_1, 303_2) of the route pair being less than a pre-configured maximum threshold distance. Claim 9 In claim 1 or 2, for each pair of roots among a plurality of roots (303, 303_1, 303_2), checking whether the root pair satisfies the connection criteria comprises calculating at least one of the minimum distance between the endpoints of the root pair, the minimum distance between the extrapolated endpoints of the root pair — the extrapolated endpoints are obtained by linearly extrapolating the endpoints with a pre-configured number of voxels (301) — and / or the skew distance of the root pair — the skew distance of the root pair includes the shortest distance between lines formed by connecting the endpoints of the roots (303, 303_1, 303_2) of the root pair to the extrapolation of the roots (303, 303_1, 303_2); and the connection criteria are that the minimum distance between the endpoints satisfies the first criterion, the minimum distance between the extrapolated endpoints satisfies the second criterion, and / or the skew distance of the root pair satisfies the third criterion. Method (100) including. Claim 10 A method (100) in which, in claim 1 or 2, each of the plurality of voxel structures (302_1, 302_2) is formed using a connected-component algorithm. Claim 11 A method (100) in which, in claim 1 or 2, mandibular canal segmentation data is obtained as the output of a trained neural network. Claim 12 A computing device (200) comprising at least one processor (201); and at least one memory (202) including computer program code, wherein the at least one memory (202) and the computer program code are configured together with at least one processor (201) to enable the computing device (200) to perform the method (100) according to claim 1 or 2. Claim 13 A computer program product comprising program code, wherein the program code is configured to perform the method (100) according to claim 1 or 2 when the computer program product is executed on a computer. Claim 14 delete

Images