Post-processing of mandibular canal segmentation

Abstract

According to an embodiment, a method for post-processing mandibular canal segmentation data includes obtaining mandibular canal segmentation data including a plurality of possible mandibular canal voxels, forming a plurality of voxel structures by connecting adjacent voxels in the possible mandibular canal voxels having a probability value above a preconfigured probability value threshold, forming a plurality of roots by skeletonizing each voxel structure, for each root pair in the plurality of roots, checking whether the root pair satisfies a concatenation criterion, and forming a second plurality of roots by concatenating the root pair in response to the root pair satisfying the concatenation criterion, performing at least one check on at least one spatial feature of each root in 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

[Technical field]

[0001] The present disclosure relates to dental imaging, and more particularly to a method for post-processing mandibular canal segmentation data, corresponding computing devices, and computer program products. [Background technology]

[0002] The human mandible, also known as the mandible, is anatomically complex and is the only mobile bone in the facial region, facilitating the functions of mastication, speech, and facial expression. It also serves as a scaffold and platform for the lower dentition, muscle attachments, temporomandibular joints, nerves, and blood vessels. Important mandibular structures are the two mandibular canals, which are located on either side below the teeth in the premolar and molar regions. Each canal contains an artery and a vein as well as the inferior alveolar nerve, which is part of the marginal mandibular branch of the trigeminal nerve and provides motor innervation to muscles and sensory innervation to the teeth, chin, and lower lip. Accurately locating the mandibular canals within the mandible is important in dental implantology. Summary of the Invention

[0003] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or important features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.

[0004] It is an object of the present invention to provide a method for post-processing of mandibular canal segmentation data. These and other objects are achieved by the features of the independent claims. Further embodiment forms are evident from the dependent claims, the description and the figures.

[0005] According to a first aspect, a method for post-processing mandibular canal segmentation data includes acquiring mandibular canal segmentation data including a plurality of possible mandibular canal voxels during a computed tomography scan of a mandible, each voxel within the plurality of possible mandibular canal voxels being associated with a probability value that quantifies a probability that the voxel includes the mandibular canal; forming a plurality of voxel structures by connecting adjacent voxels within the possible mandibular canal voxels having probability values ​​above a preconfigured probability value threshold; and forming a plurality of roots by skeletonizing each voxel structure within the plurality of voxel structures. forming a second plurality of roots, each root in the plurality of roots including a spatial curve of a corresponding voxel structure in the plurality of voxel structures; for each root pair in the plurality of roots, checking whether the root pair satisfies a concatenation criterion, and forming a second plurality of roots by concatenating the root pair in response to the root pair satisfying the concatenation criterion; performing at least one check on at least one spatial characteristic of each root in 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. The method can, for example, provide the at least one candidate mandibular canal pair with improved accuracy.

[0006] In an implementation of the first aspect, the method further includes, after forming the plurality of roots, and for each root pair in the plurality of roots, sorting voxels in each root in the plurality of roots according to a relative position of each voxel in the root before checking whether the root pair satisfies a concatenation criterion. The method can provide at least one candidate mandibular canal pair with improved efficiency, for example, because other operations can be performed more efficiently when the roots are sorted.

[0007] In another implementation of the first aspect, performing at least one check on at least one spatial feature of each root in the second plurality of roots and selecting at least one candidate mandibular tube pair from the second plurality of roots based on the at least one check includes determining an orientation of a first segment of the root, checking for each segment in the plurality of segments of the root other than the first segment of the root whether the orientation of the segment is within a preconfigured tolerance from the orientation of the first segment of the root, calculating a monotonicity score based on the number of segments in the plurality of segments that are not within the tolerance, and selecting at least one candidate mandibular tube pair from the second plurality of roots based on at least the monotonicity score. The method can provide at least one candidate mandibular tube pair with further improved accuracy, for example, because it can utilize the orientation information of the root when selecting the candidate mandibular tube pair.

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

[0009] In another implementation of the first aspect, performing at least one check on at least one spatial feature of each root in 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 includes: selecting a subset of roots from the second plurality of roots based on the monotonicity score of each root; calculating at least one pair score for each root pair in the selected subset of roots; calculating a total score for each root pair in the selected subset based on the at least one pair score of the root pairs; and selecting a root pair with the highest total score in the selected subset of roots as a candidate mandibular canal pair. The method can provide at least one candidate mandibular canal pair with further improved accuracy, for example, because it can utilize root direction information and root pair symmetry information when selecting the candidate mandibular canal pair.

[0010] In another implementation of the first aspect, performing at least one check on at least one spatial feature of each root in 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 includes: selecting a subset of roots from the second plurality of roots based on a monotonicity score of each root, which quantifies how well its orientation generally aligns with the orientation of the actual mandibular canal in terms of spatial monotonicity; calculating at least one pair score for each root pair in the selected subset of roots; calculating a total score for each root pair in the selected subset based on the at least one pair score of the root pair; and selecting the root pair with the highest total score in the selected subset of roots as the candidate mandibular canal pair. The method can provide at least one candidate mandibular canal pair with further improved accuracy, for example, because it can utilize root orientation information and root pair symmetry information when selecting the candidate mandibular canal pair.

[0011] In another implementation of the first aspect, the coordinate axis direction is parallel to the width direction of the computed tomography scan of the mandible. The method can, for example, utilize root width (left-right) symmetry information when selecting candidate mandibular canal pairs, thereby providing at least one candidate mandibular canal pair with further improved accuracy.

[0012] In another implementation of the first aspect, the symmetry score and / or the symmetry plane score are calculated based on a digest of the roots in the root pair, the digest of the roots including a root start position, a root end position, and a root average position. The method can, for example, efficiently utilize the root symmetry information and thus provide at least one candidate mandibular canal pair with further improved accuracy and efficiency.

[0013] In another implementation 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 the computed tomography scan of the mandible. The method can, for example, utilize the symmetry plane information of the root, thereby providing at least one candidate mandibular canal pair with further improved accuracy.

[0014] In another implementation of the first aspect, the concatenation criterion includes a distance between an end point of a first root in the root pair and an end point of a second root in the root pair that is less than a preconfigured maximum threshold distance. The method can, for example, efficiently determine whether two roots should be concatenated, and thus provide at least one candidate mandibular canal pair with further improved accuracy and efficiency.

[0015] In another implementation of the first aspect, for each root pair in the plurality of roots, checking whether the root pair satisfies a concatenation criterion includes calculating at least one of a minimum distance between the end points of the root pair, a minimum distance between extrapolated end points of the root pair, where the extrapolated end points are obtained by linearly extrapolating the end points by a preconfigured number of voxels, and / or a skew distance of the root pair, and the concatenation criterion includes a first criterion that the minimum distance between the end points satisfies, a second criterion that the minimum distance between the extrapolated end points satisfies, and / or a third criterion that the skew distance of the root pair satisfies. The method can, for example, determine whether to concatenate two roots based on these criteria, and thus provide at least one candidate mandibular canal pair with further improved accuracy.

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

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

[0018] According to a second aspect, a computing device includes at least one processor and at least one memory having computer program code, the at least one memory and the computer program code configured to cause the computing device to perform a method according to the first aspect, by the at least one processor.

[0019] According to a third aspect, a computer program product comprises program code configured to carry out the method according to the first aspect when executed on a computer.

[0020] Many of the attendant features will be more readily appreciated as the same becomes better understood by reference to the following detailed description considered in conjunction with the accompanying drawings, in which:

[0021] In the following, embodiments are explained in more detail with reference to the accompanying figures and drawings. [Brief description of the drawings]

[0022] [Figure 1] 1 shows a flowchart representation of a method according to an embodiment. [Diagram 2] 1 shows a schematic diagram of a computing device according to an embodiment. [Diagram 3] 1 shows a schematic diagram of multiple possible mandibular canal voxels and corresponding voxel structures, according to an embodiment. [Figure 4] 1 shows a schematic diagram of a voxel structure and a corresponding route according to an embodiment. [Diagram 5] 1 shows a schematic diagram of a voxel structure as connected components according to an embodiment; [Figure 6] 1 illustrates a flowchart representation of a concatenation procedure according to an embodiment. [Figure 7] 1 shows a schematic diagram of distances between routes according to an embodiment; [Figure 8] 1 illustrates a schematic diagram of a route segment according to an embodiment. [Figure 9]1 shows a schematic diagram of symmetry score calculation according to an embodiment. [Figure 10] 1 illustrates a flowchart representation of a candidate mandibular canal pair selection procedure, according to an embodiment. [Figure 11] 1 shows a schematic diagram of a convolutional neural network, according to an embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0023] In the following, like reference numbers are used to denote like parts in the accompanying drawings.

[0024] In the following detailed description of the present disclosure, reference is made to the accompanying drawings which form a part hereof, and in which specific aspects in which the present disclosure may be arranged are shown by way of example. It is understood that other aspects may be utilized and structural or logical changes may be made without departing from the scope of the present disclosure. Therefore, the following detailed description should not be taken in a limiting sense, as the scope of the present disclosure is defined by the appended claims.

[0025] For example, it is understood that disclosure related to a described method also applies to a corresponding device or system configured to perform the method, and vice versa. For example, where certain method steps are described, the corresponding device may include units for performing the described method steps, even if such units are not explicitly described or shown in the figures. Conversely, for example, where a particular apparatus is described in terms of functional units, the corresponding method may include steps for performing the described functions, even if such steps are not explicitly described or shown. Furthermore, it is understood that features of the various exemplary aspects described herein may be combined with each other, unless otherwise specified.

[0026] FIG. 1 shows a flow chart representation of a method according to an embodiment.

[0027] According to an embodiment, the method 100 includes acquiring 101 mandibular canal segmentation data during a computed tomography scan of the mandible, the mandibular canal segmentation data including a plurality of possible mandibular canal voxels, each voxel within the plurality of possible mandibular canal voxels being associated with a probability value that quantifies the probability that the voxel includes the mandibular canal.

[0028] The mandible is also called the lower jaw, jawbone, etc.

[0029] As used herein, a voxel may 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 may represent a single point on this grid. A data point may include a single datum, such as a probability value that quantifies the probability that the voxel has a mandibular canal, or multiple datums. A probability value may not include the probability itself. Rather, a probability value may include any quantity that quantifies a probability.

[0030] A mandibular canal voxel may refer to a voxel that corresponds to a region of the computed tomography scan that includes the mandibular canal. Similarly, any voxel within the plurality of possible mandibular canal voxels may correspond to a region of the computed tomography scan that includes the mandibular canal, with a probability associated with each voxel quantifying the probability that the region of the voxel includes the mandibular canal.

[0031] As used herein, the geometry of a computed tomography scan, mandible, and / or mandibular canal may be described by reference to depth, width, and height dimensions / directions. The depth dimension / direction may also be referred to as anterior-posterior dimension / direction, z dimension, etc. The depth dimension / direction may be perpendicular to the coronal plane. The width dimension / direction may also be referred to as left-right dimension / direction, x dimension, etc. The width dimension / direction may be perpendicular to the sagittal plane. The height dimension / direction may also be referred to as bottom-up dimension / direction, y dimension, etc. The height dimension / direction may be perpendicular to the horizontal / axial / transverse plane. The zero point of the x-axis, y-axis, and / or z-axis may be placed to be in the center of the scan of the mandible.

[0032] The method 100 may further include forming 102 a plurality of voxel structures by connecting adjacent voxels having probability values ​​above a preconfigured probability threshold within the possible mandibular canal voxels.

[0033] As used herein, a voxel structure may include any structure having multiple connected voxels. For example, a voxel structure may be implemented as a connected component as disclosed herein. A voxel structure may include multiple connected voxels. Thus, each voxel in a voxel structure has at least one other voxel neighbor in the same voxel structure.

[0034] The preconfigured probability thresholds may be, for example, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, or 0.9.

[0035] Small voxel structures may be discarded after a voxel structure is formed. For example, a voxel structure may be discarded because its size (i.e., the number of voxels in the voxel structure) is smaller than a pre-configured percentage of the largest voxel structure. For example, voxel structure C may be discarded because

number

[0036] The method 100 may further include forming a plurality of roots by skeletonizing (103) each voxel structure in the plurality of voxel structures, each root in the plurality of roots including a space curve of a corresponding voxel structure in the plurality of voxel structures.

[0037] Each root in the plurality of roots can be obtained by skeletonizing a corresponding voxel structure in the plurality of voxel structures.

[0038] The method 100 may further include, for each root pair in the plurality of roots, checking whether the root pair satisfies a concatenation criterion, and in response to a root pair that satisfies the concatenation criterion, forming a second plurality of roots by concatenating the root pairs.

[0039] As used herein, concatenating two routes may include creating a new route by joining the two routes end-to-end.

[0040] The method 100 may further include performing at least one check on at least one spatial feature of each root in the second plurality of roots (105) and selecting at least one candidate mandibular canal pair from the second plurality of roots based on the at least one check.

[0041] The mandibular canals are mostly directed downward and inward. At least one check of at least one spatial feature of each root in the second plurality of roots can use this information, as well as other information regarding the geometry of the mandibular canals, to select at least one candidate mandibular canal pair from the second plurality of roots.

[0042] The method 100 may further include providing at least one candidate mandibular canal pair. Providing may include, for example, providing the at least one candidate mandibular canal pair to a user. The at least one candidate mandibular canal pair may be provided to the user, such as by, for example, displaying a corresponding root on a display of the user. The root may be overlaid, for example, on a computed tomography scan of the mandible. In some embodiments, key points of the root may be samples and the key points may be provided.

[0043] At least some embodiments of the method 100 may also take into account small route segments that would otherwise be ignored.

[0044] At least some embodiments of the method 100 can take into account the shape of the extracted root, and therefore can ignore other anatomical structures and even mislabeled data.

[0045] The method 100 can construct a correct segmentation of the mandibular nerve canal from incomplete segmentation data that may contain gaps and false positives. The method 100 can operate on both cone beam computed tomography (CBCT) and computed tomography (CT) data. Once all possible canals can be constructed, the method 100 can screen them, for example, based on root coordinate monotonicity and pairwise symmetry.

[0046] The method 100 can first construct all possible mandibular canals suggested by the mandibular canal segmentation data, and then use heuristic methods to screen out spurious canals. The method 100 can then use checks to screen out root / root pairs that are unlikely to correspond to mandibular canal roots. Checks such as these can take into account spatial characteristics of the mandibular canal roots.

[0047] FIG. 2 illustrates a schematic diagram of a computing device 200 according to an embodiment.

[0048] According to an embodiment, the computing device 200 includes at least one processor 201 and at least one memory 202 having computer program code.

[0049] The at least one memory 202 and computer program code may be configured, by the at least one processor 201 , to cause the computing device 200 to perform the method 100 .

[0050] The at least one processor 201 may comprise one or more of a variety of processing devices, such as, for example, a central processing unit (CPU), a graphical processing unit (GPU), a co-processor, a microprocessor, a processing unit, a digital signal processor (DSP), a processing circuit with or without an associated DSP, or a variety of other processing devices including integrated circuits, 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 dedicated computer chip, etc.

[0051] The at least one memory 202 may be configured to store, for example, computer programs, etc. The at least one memory 202 may comprise one or more volatile memory devices, one or more non-volatile memory devices, and / or a combination of one or more volatile and non-volatile memory devices. For example, the at least one memory 202 may be embodied as a magnetic storage device (such as a hard disk drive, a floppy disk, a magnetic tape, etc.), a magneto-optical storage device, and a semiconductor memory (such as a mask ROM, a PROM (programmable ROM), an EPROM (erasable PROM), a flash ROM, a RAM (random access memory), etc.).

[0052] Computing device 200 may further include other components not shown in the embodiment of Figure 2. Computing device 200 may include, for example, an input / output bus for connecting computing device 200 to other devices.

[0053] If the computing device 200 is configured to implement a function, then some components and / or elements of the computing device 200, such as the at least one processor 201 and / or the at least one memory 202, may be configured to implement the function. Furthermore, if the at least one processor 201 is configured to implement a function, then the function may be implemented using program code contained in the at least one memory 202, for example.

[0054] FIG. 3 shows a schematic diagram of multiple possible mandibular canal voxels and corresponding voxel structures, according to an embodiment.

[0055] A plurality of voxel structures can be formed by connecting adjacent voxels having a probability value that exceeds a preconfigured probability threshold in the possible mandibular canal voxel. For example, in the embodiment of Fig. 3, the probability threshold is set to 0.5. Thus, the voxel structures 302_1, 302_2 shown in the embodiment of Fig. 3 are formed by connecting adjacent voxels in the possible mandibular canal voxel 301 that have a probability value that exceeds 0.5 in the voxel structures 302_1, 302_2.

[0056] The rules used for whether two voxels are adjacent may be different in different embodiments. For example, in the embodiment of FIG. 3, for each voxel, the nearest eight voxels are considered to be adjacent voxels. If only the nearest four voxels are considered to be adjacent voxels, the second 302_2 voxel structure is not formed. In three dimensions, for each voxel, the other six voxels that touch that voxel, or the other 26 voxels within a 3×3×3 voxel cube, for example, may be considered to be adjacent voxels.

[0057] It should be understood that the embodiment of FIG. 3 is merely a two-dimensional representation of a simplified example showing only two voxel structures 302_1, 302_2.

[0058] FIG. 4 shows a schematic diagram of a voxel structure and a corresponding route according to an embodiment.

[0059] Each root may be obtained by skeletonizing the voxel structure, and therefore each root may include a skeleton of the corresponding voxel structure.

[0060] As used herein, a skeleton of a voxel structure may refer to a thin version of that voxel structure that is equidistant from the boundaries of the voxel structure. Thus, the root may highlight the geometric and topological properties of the voxel structure, such as its connectivity, topology, length, direction, and width. Together with the distance of its points to the shape boundaries, the skeleton may also serve as a representation of the voxel structure.

[0061] In the embodiment of FIG. 4, each root 303_1, 303_2 is obtained by skeletonizing the corresponding voxel structure 302_1, 302_2.

[0062] According to an embodiment, the method further includes, after forming the multiple roots, and for each root pair in the multiple roots, sorting voxels in each root in the multiple roots according to a relative position of each voxel in the root before checking whether the root pair satisfies a concatenation criterion.

[0063] For example, two nearest neighbor graphs can be constructed for the voxels in R. The first graph G s The second graph G can have a smaller radius to maintain the root topology. l can have a large radius in the neighbor search, where the radius measures the voxels that are considered neighbors to a given voxel.

[0064] For example, we use breadth-first search (BFS) to find s Based on the longest path in route R

number

number

number

number

number

number

number

number

number

number

[0065] Furthermore, it can be ensured that the first and last voxels in an ordered route are true endpoints by checking their neighboring voxel counts.

[0066] FIG. 5 shows a schematic diagram of a voxel structure as connected components according to an embodiment.

[0067] According to an embodiment, each voxel structure in the plurality of voxel structures is formed using a connected component algorithm, which may also be referred to as connected component labeling (CCL) or connected component analysis (CCA).

[0068] The connected components algorithm can construct a graph 325 including vertices 321 and connecting edges 320 based on multiple possible mandibular canal voxels 301. Each vertex in the graph 325 can correspond to a voxel. The edges 320 can point to connected adjacent voxels. The connected components algorithm can construct multiple connected components 312_1, 312_2. A connected component is a subgraph in which any two vertices 321 are connected to each other by a path and to no additional vertices remaining in the graph 325. Each connected component can correspond to a voxel structure of connected voxels.

[0069] The embodiment of FIG. 5 shows connected components 312_1, 312_2 corresponding to the voxel structures 302_1, 302_2, respectively, shown in the embodiments of FIGS.

[0070] FIG. 6 illustrates a flow chart representation of a concatenation procedure, according to an embodiment.

[0071] The procedure may begin at operation 501 .

[0072] In operation 502, a score may be calculated for each pair of roots in the plurality of roots, which may be referred to as a concatenation score or the like.

[0073] In operation 503, it may be checked whether a valid suggestion for root pairs to be concatenated is provided by score calculation 502. For example, if there are no root pairs that satisfy the concatenation criteria, no valid suggestion is provided. If no valid suggestion is provided, the procedure may proceed to operation 505, where the procedure ends. If a valid suggestion is provided, the procedure may proceed to operation 504.

[0074] According to an embodiment, the concatenation criteria includes a distance between an end point of a first route in the route pair and an end point of a second route in the route pair being less than a preconfigured maximum threshold distance.

[0075] For example, in operation 502, a concatenation score may be calculated for each route pair based on the distance between the endpoint of the first route in the route pair and the endpoint of the second route in the route pair. Additionally or alternatively, other criteria as disclosed herein may be considered when calculating the concatenation score. Then, in operation 503, it may be checked whether the concatenation score of any route pair is greater than a preconfigured minimum concatenation score, i.e., whether the calculation provided a valid proposal for routes to be concatenated. Alternatively or additionally, if no valid proposal is found and this value can be detected in operation 503, the score calculation 502 itself may return a special value, such as null.

[0076] In operation 504, the route pairs with the highest scores may be concatenated. After concatenation, the procedure may return to operation 502, where a score may be calculated for each route pair. Due to the concatenation performed in operation 504, the multiple routes will include multiple different routes. Thus, operations 502-504 may be repeated until no valid concatenation suggestions are provided.

[0077] As a result of procedure 104, a second plurality of routes may be provided.

[0078] FIG. 7 shows a schematic diagram of distances between routes according to an embodiment.

[0079] According to an embodiment, for each root pair in the plurality of roots, checking whether the root pair satisfies the concatenation criterion includes calculating at least one of a minimum distance between the endpoints of the root pair, a minimum distance between extrapolated endpoints of the root pair, where the extrapolated endpoints are obtained by linearly extrapolating the endpoints by a preconfigured number of voxels, and / or a skew distance of the root pair.

[0080] For each route pair, we can calculate the minimum of the distance between the endpoints belonging to different routes in the route pair. There are 2×2=4 cases in total. The minimum distance can be denoted as d.

[0081] The minimum distance between the extrapolated endpoints of a route pair can be obtained, for example, by linearly extrapolating the endpoints of each route by k voxels. Calculate the minimum distance, d ext It can be expressed as:

[0082] The skew distance of a root pair can refer to the shortest distance between the lines formed by connecting the endpoints of the root to its extrapolation. One of the purposes of the skew distance is to check whether two roots intersect when extrapolated, and the other two criteria are to quantify whether the two roots are progressing in opposite directions. The skew distance is d スキュー can be represented as follows.

[0083] The concatenation criteria can include a first criterion that the minimum distance between endpoints is met, a second criterion that the minimum distance between extrapolated endpoints is met, and / or a third criterion that the skew distance of the root pair is met.

[0084] For example, when d ext <T1, d - d ext > T2, and d スキュー < T3 (where T i is a preconfigured threshold), the roots can be concatenated. When these criteria are met, for example, in operation 502 of the embodiment of FIG. 6, d ext can be provided as the concatenation score. When these criteria are not met, a special value such as null can be provided to indicate this. Next, based on the concatenation score of each root pair, for example, as disclosed in the embodiment of FIG. 6, the root pairs to be concatenated can be selected.

[0085] FIG. 8 shows a schematic diagram of a root segment according to an embodiment.

[0086] According to an embodiment, performing at least one check on at least one spatial feature of each root 303 in the second plurality of roots includes determining a direction of a first segment 701 of the root 303, and for each segment 702 in the plurality of segments of the root 303 other than the first segment 701 of the root 303, checking whether the direction of the segment 702 is within a preconfigured tolerance from the direction of the first segment 701 of the root 303, calculating a monotonicity score based on the number of segments in 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 at least on the monotonicity score.

[0087] The first segment 701 may correspond to any segment within the route 303. For example, in the embodiment of FIG. 8, the first segment 701 is located at one end of the route 303. In other embodiments, the first segment 701 may be located anywhere along the route 303.

[0088] The plurality of segments may include any number of segments. For example, in the embodiment of Figure 8, the plurality of segments includes all segments except the first segment. In other embodiments, the plurality of segments may include only a smaller subset of all segments in the route 303, for example.

[0089] For example, near the start of the route 303, it can be checked in which direction the x (width), y (height), and / or z (depth) coordinates are going (increasing or decreasing). In some embodiments, only a portion of these coordinates can be considered. For example, only the x and z components may be of interest. To get more robustness, some simple voting strategies can be applied here. Then, for each other segment 702, it can be checked whether the route 303 is along this direction. Some tolerance and some number of violations can be tolerated.

[0090] FIG. 9 shows a schematic diagram of symmetry score calculation according to an embodiment.

[0091] According to an embodiment, performing at least one check on at least one spatial feature of each route in the second plurality of routes includes calculating at least one pair score for at least one route pair in the second plurality of routes, the at least one pair score including at least one of an average coordinate score of the route pair, calculated by comparing average coordinates in a coordinate axis direction of each route in the route pair; a route pair symmetry score quantifying the spatial symmetry of the route pair; and / or a symmetry plane score of the route pair.

[0092] A total score for at least one root pair in the second plurality of roots can be calculated based on the pair score of at least one of the pairs. At least one candidate mandibular canal pair can be selected from the second plurality of roots based at least on the total score.

[0093] According to an embodiment, the symmetry score and / or the symmetry plane score are calculated based on digests of the routes in the route pair, where the digests of the routes include a start position of the route, an end position of the route, and an average position of the route.

[0094] In the embodiment of FIG. 9, a first root digest 801, a second root digest 802, and a symmetry plane 803 determined based on the digests 801, 802 are shown.

[0095] The symmetry score may be calculated, for example, based on an average of the displacements 804 of the digests 801, 802 from the plane of symmetry 803. Each displacement 804 may 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 plane of symmetry 803. For example, the sine of the relative positioning or acute angle between the line 805 and the plane of symmetry 803 may be used. If the line 805 is perpendicular to the plane of symmetry 803, the displacement may be zero. In other embodiments, the symmetry score may be calculated in another manner.

[0096] According to an embodiment, the plane of symmetry score quantifies how well the plane of symmetry 803 of the root pair is aligned with the elevation of a computed tomography scan of the mandible. Thus, the plane of symmetry score can quantify the bilateral symmetry of the root pair.

[0097] According to an embodiment, the coordinate axis direction is parallel to the width direction of the computed tomography scan of the mandible.

[0098] Thus, the total score can reflect how well the root pair adheres to the following observations for a correct pairing of mandibular canal roots: the z coordinates of the two roots should be close in average value, the two roots should be approximately symmetrical, the plane of symmetry should be along the height (y) direction, and the normal vector of the plane of symmetry should be along the width (x) direction.

[0099] FIG. 10 illustrates a flowchart representation of a candidate mandibular canal pair selection procedure, according to an embodiment.

[0100] According to an embodiment, selecting at least one candidate mandibular canal pair from the second plurality of roots based on the at least one check includes performing at least a portion of operations 901-904.

[0101] Operation 901 may select a subset of routes from the second plurality of routes based on the monotonicity score of each route.

[0102] A monotonicity score may be calculated for each route in the second plurality of routes in the methods disclosed herein, and the subset may be selected, for example, based on selecting all routes having a monotonicity score greater than a monotonicity score threshold.

[0103] If there are less than two routes remaining after operation 901, for example, less than two routes with monotonicity scores greater than the monotonicity score threshold, some routes with monotonicity below the threshold may be selected for the next operation, with the highest scoring routes being selected first.

[0104] In operation 902, at least one pair score may be calculated for each route pair in the selected subset of routes.

[0105] The methods disclosed herein may calculate at least one pair score for each route pair in the selected subset of routes.

[0106] In operation 903, a total score may be calculated for each route pair in the selected subset based on at least one pair score of the route pair.

[0107] In the methods disclosed herein, a total score may be calculated for each route pair in the selected subset of routes.

[0108] In operation 904, the root pair having the highest total score within the selected subset of roots may be selected as the candidate mandibular canal pair.

[0109] Thus, in the embodiment of Figure 10, the monotonicity score can be used to filter the roots, and then a candidate mandibular canal pair can be selected from the subset of remaining roots based on the total score. Thus, the embodiment of Figure 10 can select at least one candidate mandibular canal pair from the second plurality of roots based on the monotonicity score and the total score.

[0110] FIG. 11 shows a schematic diagram of a convolutional neural network, according to an embodiment.

[0111] According to an embodiment, the mandibular canal segmentation data is obtained as the output of a trained neural network.

[0112] In this specification, the term "neural network" is used to refer to an artificial neural network.

[0113] In the embodiment 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 convolution kernel of size 3×3×3 with a stride of 1, a batch normalization (BN), and a rectified linear unit (ReLU) activation function. The second type 1002 includes a convolution, a convolution kernel of size 3×3×3 with a stride of 2, a BN, and a ReLU activation function. The third type 1003 includes a transposed convolution, a convolution kernel of size 3×3×3 with a stride of 2, a BN, and a ReLU activation function. The fourth type 1004 includes a convolution, a convolution kernel of size 1×1×1 with a stride of 1, and a sigmoid activation function. The CNN also includes skip connections as shown in FIG. 11. Some of the skip connections are element-wise sums and some are feature concatenations. The number of channels in each layer is shown in FIG.

[0114] The embodiment of Figure 11 is just an example implementation of a CNN that can provide mandibular canal segmentation data. Alternatively, the 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 by data.

[0115] Any range or device value given herein may be expanded or modified without losing the effect sought, and any embodiment may be combined with another embodiment unless expressly prohibited.

[0116] Although the subject matter has been described in language specific to structural features and / or acts, it is to 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 example forms of implementing the claims, and other equivalent features and acts are intended to be within the scope of the claims.

[0117] It will be understood that the benefits and advantages described above may relate to one embodiment or to several embodiments. The embodiments are not limited to those that solve any or all of the stated problems or have any or all of the stated benefits and advantages. It will be further understood that reference to "an" item may refer to one or more of those items.

[0118] The steps of the methods described herein may be performed in any suitable order, or simultaneously as appropriate. Additionally, individual blocks may be deleted from any of the methods without departing from the spirit and scope of the subject matter described herein. Aspects of any of the above-described embodiments may be combined with any of the other embodiments described to form further embodiments without losing the desired effect.

[0119] As used herein, the term "comprising" is used to mean including identified methods, blocks, or elements, but such blocks or elements do not constitute an exclusive list and a method or apparatus may comprise additional blocks or elements.

[0120] It will be understood that the above description is given by way of example only, and that various modifications may be made by those skilled in the art. The above specification, examples, and data provide a complete description of the structure and use of the exemplary embodiments. Although various embodiments have been described above with a certain degree of particularity, or with 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 the present specification.

Claims

1. A method for post-processing of mandibular canal segmentation data using a computing device, comprising: acquiring the mandibular canal segmentation data comprising a plurality of possible mandibular canal voxels during a computed tomography scan of the mandible, wherein each voxel within the plurality of possible mandibular canal voxels is associated with a probability value that quantifies the probability that the voxel comprises the mandibular canal; forming a plurality of voxel structures by connecting adjacent voxels within the possible mandibular canal voxels having probability values ​​above a predetermined probability value threshold; forming a plurality of roots by skeletonizing each voxel structure in the plurality of voxel structures, each root in the plurality of roots comprising a space curve of a corresponding voxel structure in the plurality of voxel structures; for each root pair in the plurality of roots, testing whether the root pair satisfies a concatenation criterion, and in response to a root pair that satisfies the concatenation criterion, concatenating the root pair to form a second plurality of roots; performing at least one check on at least one spatial feature of each root in 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; The method comprising:

2. 2. The method of claim 1, further comprising: after forming the plurality of roots, and for each root pair within the plurality of roots, before checking whether the root pair satisfies the concatenation criterion, sorting voxels within each root within the plurality of roots according to a relative position of each voxel within the root.

3. performing the at least one check on the at least one spatial feature of each root in 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 includes: determining a direction for a first segment of the route; for each segment in the plurality of segments of the route other than the first segment of the route, checking whether the orientation of the segment is within a predetermined tolerance from the orientation of the first segment of the route; calculating a monotonicity score based on the number of segments in the plurality of segments that are not within the tolerance; selecting the at least one candidate mandibular canal pair from the second plurality of roots based at least on the monotonicity score; The method of claim 1 , comprising:

4. performing the at least one check on the at least one spatial feature of each root in 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 includes: calculating at least one pair score for at least one pair of routes in the second plurality of routes, the at least one pair score comprising: an average coordinate score of the route pair, the average coordinate score being calculated by comparing the average coordinates in the coordinate axis direction of each route in the route pair; a symmetry score for the root pair that quantifies the spatial symmetry of the root pair with respect to the plane of symmetry of the root pair; and / or a plane of symmetry score for the root pair, the plane of symmetry score quantifying how well the plane of symmetry for the root pair is aligned in the elevation direction with a computed tomography scan of the mandible; the calculating step comprising at least one of: calculating a total score for the at least one route pair in the second plurality of routes based on the at least one pair score of the route pair; selecting the at least one candidate mandibular canal pair from the second plurality of routes based at least on the total score; The method of claim 1 , comprising:

5. performing the at least one check on the at least one spatial feature of each root in 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 includes: selecting a subset of routes from the second plurality of routes based on a monotonicity score for each route; calculating the at least one pair score for each route pair in the selected subset of routes; calculating the total score for each route pair in the selected subset based on the at least one pair score of the route pair; selecting the root pair having the highest total score within the selected root subset as the candidate mandibular canal pair; The method of claim 4, comprising:

6. The method of claim 4 , wherein the coordinate axis direction is parallel to the width direction of the computed tomography scan of the mandible.

7. 5. The method of claim 4, wherein the symmetry score and / or the symmetry plane score are calculated based on digests of the routes in the route pair, the digests of the routes including a start position of the route, an end position of the route, and an average position of the route.

8. The method of claim 1 , wherein the concatenation criteria comprises a distance between an end point of a first route in the route pair and an end point of a second route in the route pair being less than a predetermined maximum threshold distance.

9. For each route pair in the plurality of routes, checking whether the route pair satisfies the concatenation criterion includes: the minimum distance between the endpoints of said route pair; a minimum distance between extrapolated endpoints of the route pair, the extrapolated endpoints being obtained by linearly extrapolating the endpoints by a predetermined number of voxels; and / or a skew distance for the route pair, the skew distance for the route pair comprising the shortest distance between lines formed by connecting endpoints of routes in the route pair to extrapolations of the routes; and calculating at least one of:

2. The method of claim 1, wherein the concatenation criteria include a first criterion satisfied by the minimum distance between the endpoints, a second criterion satisfied by the minimum distance between the extrapolated endpoints, and / or a third criterion satisfied by the skew distance of the root pair.

10. The method of claim 1 , wherein each voxel structure in the plurality of voxel structures is formed using a connected component algorithm.

11. The method of claim 1 , wherein the mandibular canal segmentation data is obtained as the output of a trained neural network.

12. at least one processor; and at least one memory having computer program code; 1. A computing device comprising: The computing device, wherein the at least one memory and the computer program code are configured to cause the computing device, by the at least one processor, to perform a method according to any of claims 1 to 11.

13. A computer program comprising instructions configured to carry out a method according to any one of claims 1 to 11 when the computer program is run on a computer.

Images