Systems and methods for verifying one or more segmentation models

Abstract

Systems and methods for validating one or more segmented models are provided. Image data depicting anatomical elements may be received, and the image data may be input into one or more segmentation models. The one or more segmentation models may be configured to segment image data and output segmented image data. At least one numerical measurement and at least one model measurement of an anatomical element may be received. The at least one model measurement may be compared to the at least one model measurement, and an error may be determined based on a difference between the at least one numerical measurement and the at least one model measurement. As a result of determining that the error is less than a predetermined error threshold, one or more segmentation models may be validated.

Description

Background Technology

[0001] This disclosure generally relates to segmentation models, and more specifically to the validation of one or more segmentation models.

[0002] Imaging can be used by medical providers for diagnostic, therapeutic, and / or surgical planning purposes. In some cases, multiple anatomical elements in an image can be segmented to obtain information and / or measurements about those elements. This information and / or measurements can then be used for the aforementioned diagnostic, therapeutic, and / or surgical planning purposes. Summary of the Invention

[0003] Examples of aspects of this disclosure include:

[0004] A method for verifying one or more segments according to at least one embodiment of this disclosure includes:

[0005] In any of the aspects of this document, the following steps are taken: receiving image data depicting anatomical elements; inputting the image data into one or more segmentation models configured to segment the image data and output segmented image data; receiving at least one numerical measurement result of the anatomical element; receiving at least one model measurement result of the anatomical element, wherein the at least one model measurement result is obtained by a segmentation measurement model using the segmented image data; comparing the at least one numerical measurement result with the at least one model measurement result; determining an error based on the difference between the at least one numerical measurement result and the at least one model measurement result; and validating the one or more segmentation models as a result of determining that the error is less than a predetermined error threshold.

[0006] In any aspect of this article, the at least one numerical measurement includes at least one measurement of cortical bone thickness.

[0007] Any aspect of this paper, wherein the at least one numerical measurement result and the at least one model measurement result are each obtained along one or more trajectories.

[0008] In any aspect of this article, the one or more trajectories are at least one of predefined trajectories or random trajectories.

[0009] Any aspect of this article, wherein the at least one numerical measurement result is obtained by one or more users.

[0010] Any aspect of this article, wherein the one or more users include at least three users.

[0011] Any aspect of this article, wherein the at least one numerical measurement result and the at least one model measurement result are obtained in one or more of the axial view, sagittal view and coronal view.

[0012] Any aspect of this document, wherein the at least one numerical measurement result and the at least one model measurement result are obtained for one or more test groups that are applicable, pediatric and / or contain metals.

[0013] Any aspect of this article, wherein the image data is received from at least one of a CT imaging device or an O-arm imaging device.

[0014] Any aspect of this article further includes averaging the at least one numerical measurement result prior to the comparison step.

[0015] Any aspect of this document further includes: inputting the image data into a segmentation model configured to segment the image data and output a segmented image; and inputting the segmented image data, the at least one numerical measurement result, and one or more trajectories into the segmentation measurement model, wherein the segmentation measurement model measures the at least one numerical measurement result in the segmented image.

[0016] A system for validating one or more segmentation models according to at least one embodiment of this disclosure includes: an imaging device; a processor; and a memory storing data for processing by the processor, the data, when processed, causing the processor to: receive image data depicting anatomical elements from the imaging device; input the image data into the one or more segmentation models configured to segment the image data and output segmented image data; receive at least one numerical measurement result of the anatomical element; receive at least one model measurement result of the anatomical element, wherein the at least one model measurement result is obtained by segmentation measurement models; compare the at least one model measurement result with the at least one model measurement result; determine an error based on the difference between the at least one numerical measurement result and the at least one model measurement result; and validate the one or more segmentation models as a result of determining that the error is less than a predetermined error threshold.

[0017] Any aspect of this article, wherein the imaging device includes at least one of a CT imaging device or an O-arm imaging device.

[0018] In any of the aspects of this paper, the at least one model measurement result is received from the segmentation measurement model, which is configured to receive segmented image data and the at least one numerical measurement result, and output the at least one model measurement result.

[0019] In any aspect of this article, the at least one numerical measurement includes at least one measurement of cortical bone thickness.

[0020] Any aspect of this paper, wherein the at least one numerical measurement result and the at least one model measurement result are each obtained along one or more trajectories.

[0021] Any aspect of this article, wherein the at least one numerical measurement result is obtained by one or more users.

[0022] Any aspect of this article, wherein the at least one numerical measurement result and the at least one model measurement result are obtained in one or more of the axial view, sagittal view and coronal view.

[0023] In any aspect of this document, the memory stores additional data for the processor to process, which, when processed, causes the processor to: average the at least one numerical measurement result prior to the comparison step.

[0024] A method for validating one or more segmentation models according to at least one embodiment of this disclosure includes: receiving image data depicting anatomical elements from a database; inputting the image data into the one or more segmentation models configured to segment the image data and output segmented image data; receiving at least one numerical measurement result of the anatomical element; inputting the segmented image data, the at least one numerical measurement result, and one or more trajectories into a segmentation measurement model configured to measure the at least one numerical measurement result from the segmented image data; receiving at least one model measurement result of the anatomical element from the segmentation measurement model; averaging the at least one numerical measurement result; comparing the average numerical measurement result with the at least one model measurement result; determining an error based on the difference between the average numerical measurement result and the at least one model measurement result; and validating the one or more segmentation models as a result of determining that the error is less than a predetermined error threshold.

[0025] Any one aspect in combination with any one or more other aspects.

[0026] Any one or more of the features disclosed in this article.

[0027] As is the case with any one or more of the features substantially disclosed herein.

[0028] Combinations of any one or more features as substantially disclosed herein with any one or more other features as substantially disclosed herein.

[0029] Any of these aspects / features / embodiments in combination with any one or more other aspects / features / embodiments.

[0030] Use of any one or more aspects or features as disclosed herein.

[0031] It should be understood that any feature described herein may be claimed in combination with any other feature(s) as described herein, regardless of whether such features are derived from the same embodiments described.

[0032] Details of one or more aspects of this disclosure are set forth in the accompanying drawings and the following description. Other features, objects, and advantages of the technology described in this disclosure will become clear from the specification, drawings, and claims.

[0033] The phrases “at least one,” “one or more,” and “and / or” are open-ended expressions that are both connective and discrete in operation. For example, each of the expressions “at least one of A, B, and C,” “at least one of A, B, or C,” “one or more of A, B, and C,” “one or more of A, B, or C,” and “A, B, and / or C” refers to a single A, a single B, a single C, A and B together, A and C together, B and C together, or A, B, and C together. When each of A, B, and C in the above expressions refers to an element (such as X, Y, and Z) or a class of elements (such as X1-Xn, Y1-Ym, and Z1-Zo), the phrase is intended to refer to a single element selected from X, Y, and Z, a combination of elements selected from the same class (e.g., X1 and X2), and a combination of elements selected from two or more classes (e.g., Y1 and Zo).

[0034] The term "a / an" refers to one or more of the entities in question. Thus, the terms "a" (or "an"), "one or more," and "at least one" are used interchangeably herein. It should also be noted that the terms "comprising," "including," and "having" are used interchangeably.

[0035] The foregoing is a simplified overview of this disclosure to provide an understanding of some aspects of it. This overview is neither a broad nor an exhaustive summary of this disclosure and its various aspects, embodiments, and configurations. It is not intended to identify key or essential elements of this disclosure, nor to define its scope, but rather to present selected concepts in a simplified form as an introduction to the more detailed description that follows. As will be understood, other aspects, embodiments, and configurations of this disclosure may utilize one or more features set forth above or described in detail below, either individually or in combination.

[0036] Many additional features and advantages of this disclosure will become apparent to those skilled in the art upon consideration of the embodiments described below. Attached Figure Description

[0037] The accompanying drawings are incorporated in and form a part of this specification to illustrate several examples of this disclosure. These drawings, together with the specification, explain the principles of this disclosure. The drawings simply illustrate preferred and alternative examples of how this disclosure can be performed and used, and should not be construed as limiting this disclosure to the examples shown and described. Further features and advantages will become apparent from the following more detailed description of various aspects, embodiments, and configurations of this disclosure, as illustrated in the accompanying drawings with reference below.

[0038] Figure 1 This is a block diagram of a system according to at least one embodiment of the present disclosure;

[0039] Figure 2A This is a flowchart of at least one embodiment of this disclosure;

[0040] Figure 2B This is a flowchart of at least one embodiment of this disclosure;

[0041] Figure 3 This is a flowchart of at least one embodiment of this disclosure;

[0042] Figure 4 This is a flowchart of at least one embodiment of this disclosure;

[0043] Figure 5A These are example images of anatomical elements with marked trajectories according to at least one embodiment of this disclosure;

[0044] Figure 5B These are example images of anatomical elements with marked trajectories and measurement results at locations according to at least one embodiment of this disclosure;

[0045] Figure 5CThese are example images of anatomical elements with marked trajectories according to at least one embodiment of this disclosure;

[0046] Figure 5D These are example images of anatomical elements with marked trajectories and measurement results according to at least one embodiment of this disclosure;

[0047] Figure 6 It is a graph showing the distribution of the mean difference of numerical measurement results obtained according to at least one embodiment of this disclosure;

[0048] Figure 7 These are example evaluations of numerical measurement results and model measurement results based on at least one embodiment of this disclosure;

[0049] Figure 8A These are example images of numerical measurement results and model measurement results according to at least one embodiment of this disclosure; and

[0050] Figure 8B This is another example image of numerical measurement results and model measurement results according to at least one embodiment of this disclosure. Detailed Implementation

[0051] It should be understood that the various aspects disclosed herein can be combined in combinations different from those specifically presented in the specification and drawings. It should also be understood that, depending on the examples or embodiments, certain actions or events of any process or method described herein may be performed in a different order, and / or may be added, combined, or completely omitted (e.g., implementing the technology disclosed according to different embodiments of this disclosure may not necessarily require all the described actions or events). Furthermore, although for clarity some aspects of this disclosure are described as being performed by a single module or unit, it should be understood that the technology of this disclosure can be performed by a combination of units or modules associated with, for example, computing devices and / or medical devices.

[0052] In one or more examples, the described methods, processes, and techniques may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, these functions may be stored as one or more instructions or code on a computer-readable medium and executed by a hardware-based processing unit. Alternatively or additionally, the functions may be implemented using machine learning models, neural networks, artificial neural networks, or combinations thereof (alone or in combination with instructions). Computer-readable media may include non-transitory computer-readable media, which corresponds to tangible media such as data storage media (e.g., RAM, ROM, EEPROM, flash memory, or any other medium that can be used to store desired program code in the form of instructions or data structures and that can be accessed by a computer).

[0053] The instructions can be executed by one or more processors, such as one or more digital signal processors (DSPs), general-purpose microprocessors (e.g., Intel Core i3, i5, i7, or i9 processors; Intel Celeron processors; Intel Xeon processors; Intel Pentium processors; AMD Ryzen processors; AMD Athlon processors; AMD Phenom processors; Apple A10 or 10X Fusion processors; Apple A11, A12, A12X, A12Z, or A13 Bionic processors; or any other general-purpose microprocessors), graphics processing units (e.g., Nvidia GeForce RTX 2000 series processors, Nvidia GeForce RTX 3000 series processors, AMD Radeon RX 5000 series processors, AMD Radeon RX 6000 series processors, or any other graphics processing units), application-specific integrated circuits (ASICs), field-programmable logic arrays (FPGAs), or other equivalent integrated or discrete logic circuits. Accordingly, the term "processor" as used herein may refer to any of the above-described structures or any other physical structures suitable for implementing the described techniques. Furthermore, these techniques may be fully implemented in one or more circuit or logic elements.

[0054] Before explaining any embodiment of this disclosure in detail, it should be understood that the application of this disclosure is not limited to the construction details and component arrangements set forth in the following description or shown in the accompanying drawings. This disclosure can have other embodiments and can be practiced or implemented in various ways. Furthermore, it should be understood that the wording and terminology used herein are for descriptive purposes and should not be considered limiting. The use of “including,” “comprising,” or “having,” and variations thereof herein is intended to cover items listed below and their equivalents, as well as additional items. Further, this disclosure may use examples to illustrate one or more aspects thereof. Unless otherwise expressly stated, the use or enumeration of one or more examples (which may be expressed as “for example,” “by way of example,” “e.g.,” “for instance,” or similar language) is not intended to, and will not, limit the scope of this disclosure.

[0055] Segmentation of cortical and cancellous vertebral layers from images obtained from, for example, CT imaging equipment and / or O-arm imaging equipment has a variety of applications: bone mineral density assessment, force feedback prediction for robotic osteotomy, and / or providing data for automated surgical planning. The segmentation algorithm is routinely tested using manual segmentation of ground truth by human annotators. Manual cortical bone segmentation is very time-consuming and can result in blurry images in different vertebral regions.

[0056] A novel testing method for testing segmentation models according to at least one embodiment of this disclosure is provided, which acquires multiple numerical measurements of cortical bone thickness in a predefined trajectory and compares these measurements with cortical bone thickness at the same trajectory measured by an algorithm. To validate this testing method, multiple cases are annotated by three different annotators to calculate the mean thickness and perform a variance study. Annotation is performed by loading CT scans and O-arm scans from the test dataset in each of the 3D views (axial, sagittal, and coronal). The test set consists of three test groups (suitable range, pediatric, and metallic) designed to represent different patient populations.

[0057] The test method was validated based on the results, where the 90th percentile of the algorithm's thickness error was approximately 1.4 mm, and the 90th percentile of the mean difference was approximately 0.76 mm. These results demonstrate that ground truth annotation is an effective test method. Furthermore, the algorithm's mean error was -0.11 ± 0.02 mm. In addition, the average annotation time per vertebra was approximately one hour, compared to approximately seven hours per vertebra in previous segmentation annotation projects. Therefore, the new test method can reduce time, resources, and / or costs.

[0058] The embodiments of this disclosure provide technical solutions to one or more of the following problems: (1) reducing resources, time and / or costs associated with segmentation, (2) providing efficient testing methods for segmentation models, and / or (3) improving the accuracy of segmentation models, particularly cortical bone segmentation models.

[0059] First go to Figure 1This diagram illustrates a block diagram of a system 100 according to at least one embodiment of this disclosure. System 100 can be used to generate, use, verify, and / or validate one or more segmentation algorithms and / or implement one or more other aspects of the methods disclosed herein. System 100 includes a computing device 102, one or more imaging devices 112, a database 130, and / or a cloud or other network 134. Systems according to other embodiments of this disclosure may include more or fewer components than system 100. For example, system 100 may not include imaging devices 112, one or more components of computing device 102, database 130, and / or cloud 134.

[0060] Computing device 102 includes a processor 104, a memory 106, a communication interface 108, and a user interface 110. Other embodiments of the computing device according to this disclosure may include more or fewer components than computing device 102.

[0061] The processor 104 of computing device 102 can be any processor described herein or any similar processor. Processor 104 can be configured to execute instructions stored in memory 106, which can enable processor 104 to perform one or more computational steps using or based on data received from imaging device 112, database 130 and / or cloud 134.

[0062] Memory 106 may be or include RAM, DRAM, SDRAM, other solid-state memory, any memory described herein, or any other tangible non-transitory memory used to store computer-readable data and / or instructions. Memory 106 may store information or data that can be used to perform any step of, for example, the methods 200, 210, 300 and / or 400 described herein or any other method. Memory 106 may store, for example, instructions and / or machine learning models. For example, memory 106 may store the contents (e.g., instructions and / or machine learning models) that implement image processing 120, segmentation model 122, segmentation measurement model 124, and / or verification model 126 when executed by processor 104.

[0063] Image processing 120 enables processor 104 to process image data (received from, for example, imaging device 112 or any imaging device) of an image in order to, for example, identify information about one or more anatomical elements depicted in the image data. This information may include, for example, identification of hard and / or soft tissues, identification of the boundaries between hard and soft tissues, identification of cortical bone, identification of cancellous bone, etc. Image processing 120 may also use segmentation model 122, as described below. In some embodiments, image processing 120 may process image data for use in segmentation model 122.

[0064] Segmentation model 122 enables processor 104 to segment image data in order to identify anatomical elements in the image. More generally, segmentation performed by segmentation model 122 is the process of partitioning a digital image into multiple segments (sets of pixels, also called image objects). The goal of segmentation is to simplify and / or transform the representation of a complex image into something more meaningful and / or easier to analyze. Segmentation model 122 can enable processor 104 to identify the boundaries of anatomical elements by using, for example, feature recognition. For example, segmentation 122 can enable processor 104 to identify vertebrae in image data. In other instances, segmentation 122 can enable processor 104 to identify the boundaries of anatomical elements by determining the differences or contrasts between the colors or grayscale values ​​of image pixels.

[0065] The segmentation measurement model 124 enables the processor 104 to, for example, measure the thickness of an anatomical element depicted in the segmented image data (or obtain any other measurement of the anatomical element). The segmentation measurement model 124 receives the segmented image data and at least one numerical measurement as input. In some embodiments, the anatomical element is bone, and the segmentation measurement model 124 measures the cortical bone thickness of a vertebra (or multiple vertebrae) depicted in the image data. In some instances, the segmentation measurement model 124 may output one or more model measurements of the anatomical element to verify the accuracy of the segmentation model 122 or to validate the segmentation model, as in [the following context, likely related to a specific example or scenario]. Figures 2A to 2B and Figures 4 to 8B As described herein. In some embodiments, model measurements may be one or more measurements of cortical bone thickness acquired along one or more trajectories. In other embodiments, model measurements may be any measurement of any anatomical element. In still other embodiments, model measurements may be any measurement and may not be acquired along a predefined trajectory.

[0066] Verification model 126 enables processor 104 to verify the accuracy of segmentation model 122 or to validate the segmentation model. Verification model 126 can receive numerical measurements of anatomical elements (as will be described below). Figure 2B and Figure 4 (As described in the text) and receives model measurement results from segmentation measurement model 124. Validation model 126 can then verify the accuracy of segmentation model 122 or confirm the segmentation model based on the numerical measurement results and model measurement results. More specifically, the difference between the numerical measurement results and the model measurement results is determined, and this difference can be compared to one or more predetermined thresholds. If the difference is below one or more predetermined thresholds, segmentation model 122 can be validated.

[0067] In some embodiments, such content, if provided in the form of instructions, may be organized into one or more applications, modules, packages, layers, or engines. Alternatively or additionally, memory 106 may store other types of content or data (e.g., machine learning models, artificial neural networks, deep neural networks, etc.) that can be processed by processor 104 to implement the various methods and features described herein. Therefore, although the various contents of memory 106 may be described as instructions, it should be understood that the functionality described herein can be implemented using instructions, algorithms, and / or machine learning models. Data, algorithms, and / or instructions may enable processor 104 to manipulate data stored in memory 106, and / or data received from or via imaging device 112, database 130, and / or cloud 134.

[0068] The computing device 102 may also include a communication interface 108. The communication interface 108 may be used to receive image data or other information from external sources (such as imaging device 112, database 130, cloud 134, and / or any other system or component not part of system 100), and / or to transmit instructions, images, or other information to external systems or devices (e.g., another computing device 102, imaging device 112, database 130, cloud 134, and / or any other system or component not part of system 100). The communication interface 108 may include one or more wired interfaces (e.g., USB port, Ethernet port, FireWire port) and / or one or more wireless transceivers or interfaces (e.g., configured to send and / or receive information via one or more wireless communication protocols (e.g., 802.11a / b / g / n, Bluetooth, NFC, ZigBee, etc.). In some embodiments, the communication interface 108 may be used to enable the device 102 to communicate with one or more other processors 104 or computing devices 102, whether to reduce the time required to complete computationally intensive tasks or for any other reason.

[0069] The computing device 102 may also include one or more user interfaces 110. User interfaces 110 may be or include a keyboard, mouse, trackball, monitor, television, screen, touchscreen, and / or any other device for receiving information from and / or providing information to a user. User interfaces 110 may be used, for example, to receive user selections or other user input regarding any step of any method described herein. Although as foregoing, any required input for any step of any method described herein may be automatically generated by system 100 (e.g., by processor 104 or another component of system 100) or received by system 100 from a source external to system 100. In some embodiments, user interfaces 110 may be used to allow a surgeon or other user to modify instructions to be executed by processor 104 according to one or more embodiments of this disclosure, and / or modify or adjust settings of other information displayed on or corresponding to user interfaces 110.

[0070] Although user interface 110 is shown as part of computing device 102, in some embodiments, computing device 102 may utilize user interface 110 separately housed from one or more other components of computing device 102. In some embodiments, user interface 110 may be positioned close to one or more other components of computing device 102, while in other embodiments, user interface 110 may be positioned away from one or more other components of computing device 102.

[0071] Imaging device 112 is operable to image (e.g., bones, veins, tissues, etc.) and / or other aspects of a patient's anatomy to produce image data (e.g., image data depicting or corresponding to bones, veins, tissues, etc.). As used herein, "image data" means data generated or captured by imaging device 112, including data in machine-readable, graphical / visual, and any other form. In various examples, image data may include data corresponding to a patient's anatomical features or a portion thereof. Image data may be historical image data acquired from one or more patient categories (e.g., within a suitable range, pediatric, and / or containing metals). It should be understood that image data may be acquired for other patient categories, for any other category, and at any time (e.g., historical, preoperative, intraoperative, postoperative, etc.). The anatomical features depicted in the image data may be, for example, bones, such as one or more vertebrae. Figures 4 to 8B As discussed in detail, segmentation of vertebrae (and particularly the cortical bone of vertebrae) can be difficult because the cortical bone may be thin or appear thin in certain orientations. Therefore, it is desirable to verify the accuracy of or validate segmentation models used to segment the cortical bone of one or more vertebrae (or any anatomical element or feature).

[0072] In some embodiments, the first imaging device 112 may be used to acquire first image data (e.g., a first image) at a first time, and the second imaging device 112 may be used to acquire second image data (e.g., a second image) at a second time after the first time. The imaging device 112 may be capable of capturing 2D or 3D images to generate image data. The imaging device 112 may be or include, for example: an ultrasound scanner (which may include, for example, physically separate transducers and receivers, or a single ultrasound transceiver), an O-arm, a C-arm, a G-arm, or any other device utilizing X-ray-based imaging (e.g., a fluorescence microscope, a CT scanner, or other X-ray machine), a magnetic resonance imaging (MRI) scanner, an optical coherence tomography (OCT) scanner, an endoscope, a microscope, an optical camera, a thermal imaging camera (e.g., an infrared camera), a radar system (which may include, for example, a transmitter, a receiver, a processor, and one or more antennas), or any other imaging device 112 adapted to acquire images of a patient's anatomical features. Imaging device 112 may be entirely contained within a single housing, or may include transmitters / transmitters and receivers / detectors located in separate housings or otherwise physically separated.

[0073] In some embodiments, imaging device 112 may include more than one imaging device 112. For example, a first imaging device may provide first image data and / or a first image, and a second imaging device may provide second image data and / or a second image. In yet another embodiment, the same imaging device may be used to provide both the first image data and the second image data and / or any other image data described herein. Imaging device 112 may be operable to generate an image data stream. For example, imaging device 112 may be configured to operate with an open shutter or to continuously alternate between an open and closed shutter to capture continuous images. For the purposes of this disclosure, unless otherwise stated, if the image data represents two or more frames per second, the image data may be considered continuous and / or provided as an image data stream.

[0074] Database 130 may store image data and / or image datasets. Database 130 may additionally or alternatively store, for example, segmentation model(s) 122, verification model(s) 126, and / or any other useful information. Database 130 may be configured to provide any such information directly or via cloud 134 to computing device 102, or to any other device of or outside system 100. In some embodiments, database 130 may be or include part of a hospital image storage system, such as a Picture Archiving and Communication System (PACS), a Health Information System (HIS), and / or another system for collecting, storing, managing, and / or transmitting electronic medical records including image data.

[0075] Cloud 134 can be or represents the Internet or any other wide area network. Computing device 102 can connect to cloud 134 via communication interface 108 using a wired connection, a wireless connection, or both. In some embodiments, computing device 102 can communicate with database 130 and / or external devices (e.g., computing devices) via cloud 134.

[0076] System 100 or similar systems may be used, for example, to implement one or more aspects of any of the methods 200, 210, 300 and / or 400 described herein. System 100 or similar systems may also be used for other purposes.

[0077] Go to Figure 2A An example of a model architecture 200 is shown that supports methods and systems for segmenting image data using one or more segmentation models (e.g., artificial intelligence (AI) based methods and / or systems).

[0078] Image data 202 of anatomical elements (e.g., one or more vertebrae) Figures 5A to 5B The example image data 202 shown can be used by a processor (e.g., processor 104) as input to segmentation model 122. Segmentation model 122 can output segmented image data 206 and / or (multiple) model measurement results 208. In some embodiments, image data 202 can be received from any component of an imaging device (e.g., imaging 112), any other imaging device, or a system (e.g., system 100). Image data 202 can be historical image data 202 acquired from one or more patient categories (e.g., within a suitable range, pediatric, and / or containing metals). In other instances, image data 202 can be acquired for other patient categories, for any other category, and at any time (e.g., historical, preoperative, intraoperative, postoperative, etc.). The anatomical features depicted in image data 202 can be, for example, bones, such as one or more vertebrae; however, it should be understood that in other embodiments, the anatomical features can be any anatomical feature (e.g., soft tissue, hard tissue, etc.).

[0079] As previously described, segmentation model 122 can identify anatomical elements and / or boundaries between one or more features of an anatomical element (e.g., hard tissue, soft tissue, cortical bone, cancellous bone, etc.). The segmentation model can achieve identification by, for example, determining the difference or contrast between the colors or grayscale levels of image pixels. For example, the boundary between hard and soft tissue can be identified as the contrast between brighter and darker pixels. Segmenting one or more anatomical elements from image data 202 when the image data includes a three-dimensional representation of a patient's anatomy can alternatively or additionally include identifying the boundaries of one or more anatomical elements and forming separate three-dimensional representations of the anatomical elements. In some embodiments, identifying boundaries may include a set of adjacent pixels that identify a boundary with contrast large enough to represent the boundary of an anatomical element or a feature depicted within the anatomical element. In other embodiments, feature recognition can be used to identify the boundaries of an anatomical element or a feature of an anatomical element. For example, feature recognition can be used to identify cortical bridges of vertebrae.

[0080] Historical image data can be used to train segmentation model 122. In other embodiments, image data 202 can be used to train segmentation model 122. In such embodiments, segmentation model 122 can be trained before image data 202 is input into segmentation model 122, or it can be trained in parallel with image data 202 being input into segmentation model 122.

[0081] As previously described, segmentation model 122 can output segmented image data 206, which can be used by segmentation measurement model 124.

[0082] As previously described, the segmentation measurement model 124 enables the processor 104 to, for example, measure the thickness of anatomical elements depicted in the segmented image data 206. The segmentation measurement model 124 receives the segmented image data 206 and at least one numerical measurement result 212 as input. In some instances, the segmentation measurement model 124 may also receive one or more trajectories. In some embodiments, the anatomical element is bone, and the segmentation measurement model 124 measures the cortical bone thickness of a vertebra (or multiple vertebrae) depicted in the segmented image data 206. In some instances, the segmentation measurement model 124 may output one or more model measurements 208 of the anatomical element to verify the accuracy of the segmentation model 122 or to validate the segmentation model, as will be explained below. Figure 2B and Figures 4 to 8BAs described herein. In some embodiments, model measurement result 208 may be one or more measurements of cortical bone thickness acquired along one or more trajectories. In other embodiments, model measurement result 208 may be any measurement of any anatomical element. In still other embodiments, model measurement result 208 may be any measurement and may not be acquired along a predefined trajectory.

[0083] Go to Figure 2B An example of model architecture 210 supporting methods and systems for validating one or more segmentation models 122 is shown.

[0084] Numerical measurements 212 and model measurements 208 of anatomical elements (e.g., one or more vertebrae) can be used by a processor (e.g., processor 104) as input to a validation model 126. The validation model 126 can output one or more validation results 218 that validate or confirm the accuracy of the segmentation model(s) 122. In some embodiments, numerical measurements 212 can be generated by manually measuring anatomical features depicted in image data 202 by one or more annotators. Figure 5D (Depicted in the text). In other words, the numerical measurement result 212 can be obtained by one or more people. In some embodiments, the measurement result can be the thickness of the cortical bone of the vertebra. In other embodiments, the measurement result can be any measurement result of any anatomical element or feature. In some embodiments, the numerical measurement result 212 is obtained from three annotators. In other embodiments, the numerical measurement result 212 is obtained from one annotator, two annotators, or more than two annotators. It can be along at least one trajectory 504 (in Figures 5A to 5C (As shown in the figure) Numerical measurement results 212 are obtained. At least one trajectory 504 may be, for example, orthogonal to the surface of the anatomical element. In some embodiments, at least one trajectory 504 is three trajectories and may be located in, for example, along axial views, sagittal views, and / or coronal views. In other embodiments, at least one trajectory 504 is one trajectory, two trajectories, or more than two trajectories. In some embodiments, at least one trajectory 504 is predefined; however, it should be understood that in other embodiments, at least one trajectory 504 may be randomized.

[0085] In an embodiment where at least one trajectory 504 is randomized, at least one trajectory 504 may be generated using the following methods: (1) compute a vertebral segmentation mask using, for example, an AI algorithm in image data that is a CT scan or O-arm scan; (2) convert the vertebral segmentation mask to a floating-point data type and apply a spatial low-pass filter to the vertebral segmentation mask; (3) apply a filter such as a Sobel filter to the result of (2) to generate a gradient map; (4) generate an outer shell of the vertebral segmentation mask by applying a morphological erosion operator and a skeletonization operator; (5) sample multiple random locations from the outer shell of the vertebral segmentation mask (each vertebra) generated in (4); (6) for each of the multiple random locations, generate a normal vector (e.g., orthogonal to the vertebral surface) by sampling the corresponding gradient map location generated in (3); (7) utilize a vector (e.g., axial, coronal, and sagittal) that is nearly parallel to one of the annotation planes; (8) “coat” the resulting trajectory (vector) using an incrementing marker; and (9) The results of combining all trajectories are saved (e.g., in database 130, cloud 134, and / or storage 106) for further annotation purposes.

[0086] It should be understood that in some embodiments (e.g., in instances where a labeler is used to obtain numerical measurement results), measurement results may not be obtained along a predefined trajectory.

[0087] Model measurement results 214 are received from, for example, the segmented measurement model 124 described above. Model measurement results 208 can be obtained at the same predefined or randomized trajectory 504 as numerical measurement results 212. It should be understood that in some embodiments, model measurement results 214 and numerical measurement results 212 may not be obtained along a predefined trajectory.

[0088] As previously described, validation model 126 can verify the accuracy of segmentation model 122 or confirm the segmentation model based on numerical measurement results 212 and model measurement results 208. More specifically, the difference between numerical measurement results 212 and model measurement results 208 is determined, and this difference can be compared with one or more predetermined thresholds. If the difference is below one or more predetermined thresholds, segmentation model 122 can be validated. It should be understood that validation model 126 can use any other form of validation or confirmation.

[0089] Figure 3 Methods 300 that can be used, for example, to generate models (such as segmentation model 122 and / or image processing 120) are described.

[0090] Method 300 includes generating a model (step 304). This model may be segmentation model 122 and / or image processing 120. A processor (e.g., processor 104) may generate segmentation model 122 and / or image processing 120. Segmentation model 122 and image processing 120 may be generated to facilitate and achieve, for example, the identification of one or more anatomical elements and / or anatomical features (e.g., cortical bone) depicted in image data, and to verify the accuracy of segmentation model 122 or validate the segmentation model.

[0091] Method 300 also includes training a model (step 308). In some embodiments, historical image data from multiple patients is used to train segmentation model 122 and / or image processing 120. Historical data can be obtained from any patient. In some embodiments, historical data can be obtained from patients with similar statistical data.

[0092] Method 300 further includes storing the model (step 312). The segmentation model 122 and / or image processing 120 may be stored in memory (e.g., memory 106) and / or a database (e.g., database 130) for later use. In some embodiments, the segmentation model 122 and / or image processing 120 is stored in memory when it has been sufficiently trained. The segmentation model 122 and / or image processing 120 may be sufficiently trained when it produces an output that meets a predetermined threshold, which may be determined, for example, by a user, or automatically by a processor (e.g., processor 104).

[0093] This disclosure covers embodiments of method 300 that include one or more steps that are more or fewer than those described above and / or different from those described above.

[0094] Will be described together Figure 4 , Figures 5A to 5D , Figure 6 , Figure 7 and Figures 8A to 8B . Figure 4 A method 400 is described that can be used, for example, to validate one or more segmentation models; Figure 5A Images depicting anatomical elements marked with one or more trajectories 504; Figure 5B Images depicting the locations of anatomical elements with marked trajectories 504 and numerical measurements 212; Figure 5C These are example images of anatomical elements with marked trajectories 504 according to at least one embodiment of this disclosure; Figure 5D These are example images of anatomical elements having a marked trajectory 504 and a numerical measurement result 212 according to at least one embodiment of this disclosure; Figure 6It is a chart that shows the distribution of the mean differences in the obtained numerical measurement results; Figure 7 These are examples of evaluations of numerical measurement results and model measurement results; Figure 8A 8B is an example image of numerical and model measurement results obtained along the same trajectory; and 8B is another example image of numerical and model measurement results obtained along the same trajectory.

[0095] Method 400 (and / or one or more of its steps) may be implemented, for example, by at least one processor or otherwise executed. At least one processor may be the same as or similar to the processor(s) 104(s) of the computing device 102 described above. Processors other than any processor described herein may also be used to execute method 400. At least one processor may execute method 400 by executing elements stored in memory (e.g., memory 106). Elements stored in memory and executed by the processor may enable the processor to perform one or more steps of the function shown in method 400. One or more portions of method 400 may be executed by the processor executing any contents of memory (e.g., image processing 120, segmentation model 122, segmentation measurement model 124, and / or verification model 126).

[0096] Method 400 includes receiving image data depicting anatomical elements (step 404). The image data can be combined with image data 202 (in... Figure 2A and Figures 5A to 5D as well as Figures 8A to 8B The image data may be the same as or similar to that shown in the diagram. Image data may be received via a user interface (e.g., user interface 110) and / or a communication interface (e.g., communication interface 108) of a computing device (e.g., computing device 102), and may be stored in the memory (e.g., memory 106) of the computing device. Image data may also be received from an external database or image repository (e.g., a hospital image storage system, such as a Picture Archiving and Communication System (PACS), a Health Information System (HIS), and / or another system for collecting, storing, managing, and / or transmitting electronic medical records including image data) and / or via the Internet or another network. In other embodiments, image data may be received or obtained from an imaging device (e.g., imaging device 112), which may be any imaging device, such as an MRI scanner, CT scanner, O-arm scanner, any other X-ray-based imaging device, or ultrasound imaging device. Image data may also be generated by any other component of the system (e.g., system 100) and / or uploaded to any other component of the system. In some embodiments, image data may be received indirectly via any other component of the system or a node of a network to which the system is connected.

[0097] Image data can be 2D or 3D images, or sets of 2D and / or 3D images. Image data can depict a patient's anatomical structures or portions thereof. In some embodiments, image data can depict multiple anatomical elements associated with a patient's anatomical structure, including incidental anatomical elements (e.g., ribs or other anatomical objects to which surgery or surgical procedures will be performed) in addition to the target anatomical element (e.g., a vertebra or other anatomical object to which surgery or surgical procedures will be performed). Image data can include various features corresponding to the patient's anatomical structures and / or anatomical elements (and / or portions thereof), including gradients corresponding to the boundaries and / or contours of the depicted anatomical elements, different intensity levels corresponding to different surface textures of the depicted anatomical elements, combinations thereof, etc. Image data can depict any part or portion of a patient's anatomical structure and can include, but is not limited to, one or more vertebrae, ribs, lungs, soft tissues (e.g., skin, tendons, muscle fibers, etc.), vertebrae, etc.

[0098] Image data can be processed using image processing, such as image processing 120, to identify anatomical elements and / or to prepare the image data for segmentation by a segmentation model, such as segmentation model 122, as will be described below.

[0099] Method 400 further includes receiving at least one numerical measurement of an anatomical element (step 408). The at least one numerical measurement can be combined with (multiple) numerical measurements 212 (in... Figure 2B and Figure 5D (as shown in the image) are the same or similar, and can be received as input via a user interface, and can be stored in, for example, memory, a database (e.g., database 130), and / or the cloud (e.g., cloud 134). As previously described, anatomical elements 502 depicted in image data (which may or may not be segmented) can be manually measured by one or more annotators. Figures 5A to 5D Numerical measurement results are obtained by using anatomical features of anatomical elements such as (shown in the figure). In other embodiments, the measurement results can be any measurement result of any anatomical element or feature. In some embodiments, numerical measurement results are obtained from three annotators. In other embodiments, numerical measurement results are obtained from one annotator, two annotators, or more than two annotators. The measurement can be performed along at least one predetermined trajectory (e.g., trajectory 504). Figures 5A to 5C(As shown in the diagram) Numerical measurement results are obtained. In some embodiments, at least one trajectory is three trajectories and may be located in, for example, an axial view, a sagittal view, and / or a coronal view. In other embodiments, at least one trajectory is one trajectory, two trajectories, or more than two trajectories. In some embodiments, at least one trajectory is predefined; however, it should be understood that in other embodiments, at least one trajectory may be random. The numerical measurement results may correspond to, for example, measurements of the thickness of the cortical bone of a vertebra. It should be understood that in other embodiments, the numerical measurement results may correspond to any measurement result and / or any anatomical element or feature.

[0100] At least one numerical measurement can be used as ground truth data. In other words, at least one numerical measurement can be used to train or validate (multiple) segmentation models. This is in Figure 6 As shown in the mean measurement difference chart 600, the 90th percentile 602 of the measurement difference among the three annotators is 0.76 mm, indicating that the numerical measurements(s) from different annotators are relatively identical. Furthermore, the 90th percentile 602 of the annotator measurement difference is lower than the 90th percentile of the model thickness error (not shown), meaning that the numerical measurements(s) are more accurate than the model measurements. Therefore, in at least one embodiment of this disclosure, a set of numerical measurements from a single annotator can be used to validate the segmentation model.

[0101] It should be understood that in some embodiments, step 408 may occur after step 410, which is described below.

[0102] Method 400 also includes inputting image data into a segmentation model (step 410). The segmentation model may be the same as or similar to segmentation model 122. Image data may be received from, for example, step 410.

[0103] As previously described, a processor can use a segmentation model to identify boundaries between anatomical elements and / or one or more features of an anatomical element (e.g., hard tissue, soft tissue, cortical bone, cancellous bone, etc.). The segmentation model can achieve identification by, for example, determining the differences or contrasts in color or grayscale between image pixels. For example, the boundary between hard and soft tissue can be identified as the contrast between brighter and darker pixels. When image data includes a three-dimensional representation of a patient's anatomy, segmenting one or more anatomical elements from the image data may alternatively or additionally include identifying the boundaries of one or more anatomical elements and forming separate three-dimensional representations of the anatomical elements. In some embodiments, identifying boundaries may include identifying a set of adjacent pixels whose contrast is large enough to represent the boundary of an anatomical element or a feature depicted within the anatomical element. In other embodiments, feature recognition may be used to identify the boundaries of an anatomical element or a feature of an anatomical element. For example, feature recognition can be used to identify cortical bridges of vertebrae. The segmentation model outputs one or more segmented images (whether multiple 2D images or multiple 3D images).

[0104] Method 400 further includes inputting the segmented image into a segmentation measurement model (step 411). Numerical measurement results obtained in step 408 may also be input into the segmentation measurement model. The segmentation measurement model may be the same as or similar to segmentation measurement model 124. In embodiments where numerical measurement results are obtained along a predetermined trajectory, the predetermined trajectory may also be input into the segmentation measurement model. The trajectory may be the same as the trajectory along which at least one numerical measurement result is obtained in step 408. It should be understood that in some embodiments, the model measurement results and the numerical measurement results may not be obtained along a predefined trajectory. The segmentation model may automatically obtain model measurement results corresponding to the numerical measurement results(s). In other words, the segmentation model may obtain the same type of measurement result at the same trajectory as the numerical measurement result (in embodiments where measurement results are obtained along one or more trajectories).

[0105] Method 400 also includes receiving at least one model measurement result of an anatomical element (step 412). For example, at least one model measurement result may be received from a segmentation measurement model in step 411.

[0106] Method 400 further includes averaging at least one numerical measurement result (step 416). The at least one numerical measurement result may be averaged by, for example, a processor or any other processor. At least one numerical measurement result can be combined by averaging the starting coordinates and averaging the ending coordinates of the at least one numerical measurement result. In some embodiments, the coordinates of the at least one numerical measurement result may be examined such that corresponding starting and ending points are close to each other, and may be flipped if necessary. The (multiple) averaged numerical measurement results may be used to compare with model measurement results and / or validate the segmentation model, as described below.

[0107] It should be understood that method 400 may exclude step 416. In such an embodiment, a set of numerical measurements from one annotator can be used for comparison and / or verification. In other embodiments, each set of numerical measurements from different annotators can be used, and each can be compared with the model measurement results and / or used for verification.

[0108] Method 400 further includes comparing at least one numerical measurement result with at least one model measurement result (step 420). The at least one numerical measurement result can be compared with at least one model measurement result. As described above, in some embodiments where the at least one numerical measurement result includes multiple sets of numerical measurement results from different annotators, the average of the multiple sets of numerical measurement results can be compared with at least one model measurement result. In other embodiments, a set of numerical measurement results obtained from one annotator can be compared with at least one model measurement result. Further, each set of numerical measurement results in the multiple sets can be compared individually with at least one model measurement result. For illustrative purposes, Figure 8A and Figure 8B At least one numerical measurement 800 of the thickness of cortical bone 804 and at least one model measurement 802 superimposed on the at least one numerical measurement 800 are shown.

[0109] Method 400 further includes determining an error based on the difference between at least one numerical measurement and at least one model measurement (step 424). The difference between at least one numerical measurement (whether as an average or as a single set of numerical measurements) and at least one model measurement can be determined by a processor. In some embodiments, the difference is determined by subtracting at least one numerical measurement from at least one model measurement (or vice versa).

[0110] In some embodiments, such as Figure 7As shown, the error can be determined by creating vector v1 700 from the first measurement point m1 702 and the point a1 706 found at the end of the cortex 704. Then, the same operation is performed in the opposite direction, where vector v2 708 is created from the second measurement point m2 710 and the point a2 712 found on the cortex 704. The sampled values ​​are then median filtered to reduce isolated voxels and noise, resulting in a measurement accuracy of approximately 0.1 voxels. The model measurements are then taken as the Euclidean distance between points a1 706 and a2 712, indicating the cortical thickness along the labeled trajectory. The error between the model measurements (a1 706 and a2 712) and the numerical measurements (m1 702 and m2 710) is calculated using, for example, the following equation: .

[0111] Method 400 further includes validating one or more segmentation models (step 428). The processor can use a validating model (e.g., validating model 126) to validate one or more segmentation models to verify the accuracy of the segmentation models or to validate the segmentation models. The validating model can receive numerical measurements and model measurements, and validate the accuracy of the segmentation models or to validate the segmentation models based on the numerical measurements and model measurements. More specifically, the difference or error between the numerical measurements and model measurements measured in step 424 is compared to one or more predetermined thresholds. The predetermined thresholds can be associated with the maximum permissible difference for each expected difference. In some embodiments, the predetermined thresholds can be automatically determined using artificial intelligence and training data (e.g., historical cases). In other embodiments, the predetermined thresholds can be or include or are based on annotator input or any other user input received via a user interface. In yet another embodiment, the predetermined thresholds can be automatically determined using artificial intelligence and can subsequently be reviewed and approved (or modified) by annotators or other users. In examples where the predetermined thresholds include multiple predetermined thresholds, an alert can be issued to the user (e.g., a notification can be generated) when each expected predetermined threshold is met or exceeded.

[0112] If the difference is below one or more predetermined thresholds, the segmentation model can be validated (and a notification can be generated accordingly). It should be understood that the model can be validated using any other form of validation or verification.

[0113] This disclosure covers embodiments of method 400 that include one or more steps that are more or fewer than those described above and / or different from those described above.

[0114] The validation of the segmentation model as described beneficially enables the testing and validation of multiple segmentation models. As more segmentation models are developed and / or further optimized, validation is beneficial in determining whether a segmentation model is accurate and / or usable, for example, for measuring cortical bone thickness. Cortical bone thickness measurements are useful for bone mineral density assessment, force feedback prediction for robotic osteotomy, and / or for data used in automated surgical planning.

[0115] As noted above, this disclosure covers companies with fewer than [number missing] [details missing]. Figures 2A to 2B , Figure 3 and Figure 4 (and the corresponding descriptions of methods 200, 210, 300, and 400) all steps identified in the methods, and including those other than Figures 2A to 2B , Figure 3 and Figure 4 (And corresponding descriptions of methods 200, 210, 300, and 400) methods that include additional steps beyond those identified in the methods described herein. This disclosure also covers methods that include one or more steps from one method described herein and one or more steps from another method described herein. Any relevance described herein may be or includes registration or any other relevance.

[0116] The foregoing is not intended to limit this disclosure to one or more of the forms disclosed herein. For example, in the foregoing specific embodiments, various features of this disclosure are grouped together in one or more aspects, embodiments, and / or configurations for the purpose of fluent expression. Features of aspects, embodiments, and / or configurations of this disclosure may be combined in alternative aspects, embodiments, and / or configurations other than those discussed above. The method of this disclosure is not to be construed as reflecting an intention that the claims would require more features than expressly stated in each claim. Rather, as reflected in the claims, the inventive aspect possesses fewer features than all the features possessed by a single foregoing disclosure aspect, embodiment, and / or configuration. Therefore, the claims are thus incorporated into the specific embodiments, wherein each claim itself may serve as a separate preferred embodiment of this disclosure.

[0117] Furthermore, while the foregoing has included descriptions of one or more aspects, embodiments, and / or configurations, as well as certain variations and modifications, other variations, combinations, and modifications are also within the scope of this disclosure, for example, those that, upon understanding this disclosure, would be within the skill and knowledge of someone skilled in the art. The foregoing is intended to obtain rights to alternative aspects, embodiments, and / or configurations, including those within the permitted scope (including alternative, interchangeable, and / or equivalent structures, functions, scopes, or steps of those claimed) (whether such alternative, interchangeable, and / or equivalent structures, functions, scopes, or steps are disclosed herein), and is not intended to publicly offer any patentable subject matter.

[0118] The following provides a set of examples:

[0119] Example 1. A method for validating one or more segmentation models, the method comprising: receiving image data depicting anatomical elements (202); inputting the image data into the one or more segmentation models (122), the one or more segmentation models being configured to segment the image data and output segmented image data (206); receiving at least one numerical measurement result (212) of the anatomical elements; receiving at least one model measurement result (208) of the anatomical elements, wherein the at least one model measurement result is obtained by a segmentation measurement model (124) using the segmented image data; comparing the at least one numerical measurement result with the at least one model measurement result; determining an error based on the difference between the at least one numerical measurement result and the at least one model measurement result; and validating the one or more segmentation models as a result of determining that the error is less than a predetermined error threshold.

[0120] Example 2. The method as described in Example 1, wherein the at least one numerical measurement result includes at least one cortical bone thickness measurement result.

[0121] Example 3. The method as described in Example 1 or 2, wherein the at least one numerical measurement result and the at least one model measurement result are each acquired along one or more trajectories.

[0122] Example 4. The method as described in any one of Examples 1 to 3, wherein the one or more trajectories are at least one of a predefined trajectory or a random trajectory.

[0123] Example 5. The method as described in any one of Examples 1 to 4, wherein the at least one numerical measurement result is obtained by one or more users.

[0124] Example 6. The method as described in Example 5, wherein the one or more users include at least three users.

[0125] Example 7. The method of any one of Examples 1 to 6, wherein the at least one numerical measurement result and the at least one model measurement result are measured in one or more of an axial view, a sagittal view, and a coronal view.

[0126] Example 8. The method of any one of Examples 1 to 7, wherein the at least one numerical measurement result and the at least one model measurement result are obtained for one or more test groups, including pediatric and / or metal-containing test groups.

[0127] Example 9. The method of any one of Examples 1 to 8, wherein the image data is received from at least one of a CT imaging device or an O-arm imaging device.

[0128] Example 10. The method of any one of Examples 1 to 9 further includes: averaging the at least one numerical measurement result prior to the comparison step.

[0129] Example 11. The method of any one of Examples 1 to 10, further comprising: inputting the segmented image data, the at least one numerical measurement result, and one or more trajectories into the segmentation measurement model, wherein the segmentation measurement model measures the at least one numerical measurement result in the segmented image.

[0130] Example 12. A system for validating one or more segmentation models, the system comprising: an imaging device (112); a processor (104); and a memory (106) storing data for processing by the processor, the data, when processed, causing the processor to: receive image data (202) depicting anatomical elements from the imaging device; input the image data into the one or more segmentation models (122), the one or more segmentation models being configured to segment the image data and output segmented image data (206); receive at least one numerical measurement result (212) of the anatomical elements; receive at least one model measurement result (208) of the anatomical elements, wherein the at least one model measurement result is obtained by a segmentation measurement model (124); compare the at least one model measurement result with the at least one model measurement result; determine an error based on the difference between the at least one numerical measurement result and the at least one model measurement result; and validate the one or more segmentation models as a result of determining that the error is less than a predetermined error threshold.

[0131] Example 13. The system as described in Example 12, wherein the imaging device includes at least one of a CT imaging device or an O-arm imaging device.

[0132] Example 14. A system as described in Example 12 or 13, wherein the at least one model measurement result is received from the segmentation measurement model, the segmentation measurement model being configured to receive segmented image data and the at least one numerical measurement result, and output the at least one model measurement result.

[0133] Example 15. The system of any one of Examples 12 to 14, wherein the at least one numerical measurement includes at least one cortical bone thickness measurement.

[0134] Example 16. The system as described in any one of Examples 12 to 15, wherein the at least one numerical measurement result and the at least one model measurement result are each acquired along one or more trajectories.

[0135] Example 17. The system as described in any one of Examples 12 to 16, wherein the at least one numerical measurement result is obtained by one or more users.

[0136] Example 18. The system of any one of Examples 12 to 17, wherein the at least one numerical measurement result and the at least one model measurement result are measured in one or more of an axial view, a sagittal view, and a coronal view.

[0137] Example 19. A system as described in any one of Examples 12 to 19, wherein the memory stores additional data for the processor to process, the data, when processed, causing the processor to: average the at least one numerical measurement result prior to the comparison step.

[0138] Example 20. A method for validating one or more segmentation models, the method comprising: receiving image data (202) depicting anatomical elements from a database (130); inputting the image data into one or more segmentation models (122) configured to segment the image data and output segmented image data (206); receiving at least one numerical measurement result (212) of the anatomical elements; inputting the segmented image data, the at least one numerical measurement result, and one or more trajectories (504) into a segmentation measurement model (124) configured to measure the at least one numerical measurement result from the segmented image data; receiving at least one model measurement result (124) of the anatomical elements from the segmentation measurement model; averaging the at least one numerical measurement result; comparing the average numerical measurement result with the at least one model measurement result; determining an error based on the difference between the average numerical measurement result and the at least one model measurement result; and validating the one or more segmentation models as a result of determining that the error is less than a predetermined error threshold.

Claims

1. A method for validating one or more segmentation models, the method comprising: Receive image data depicting anatomical elements (202); The image data is input into one or more segmentation models (122), which are configured to segment the image data and output segmented image data (206). Receive at least one numerical measurement result of the anatomical element (212); Receive at least one model measurement result (208) of the anatomical element, wherein the at least one model measurement result is obtained by segmentation measurement model (124) using the segmented image data; Compare the at least one numerical measurement result with the at least one model measurement result; The error is determined based on the difference between the at least one numerical measurement result and the at least one model measurement result; and As a result of determining that the error is less than a predetermined error threshold, the one or more segmentation models are validated.

2. The method as described in claim 1, wherein, The at least one numerical measurement result includes at least one cortical bone thickness measurement result.

3. The method as described in claim 1 or 2, wherein, The at least one numerical measurement result and the at least one model measurement result are each obtained along one or more trajectories.

4. The method according to any one of claims 1 to 3, wherein, The one or more trajectories are at least one of predefined trajectories or random trajectories.

5. The method according to any one of claims 1 to 4, wherein, The at least one numerical measurement result is obtained by one or more users.

6. The method of claim 5, wherein, The one or more users include at least three users.

7. The method according to any one of claims 1 to 6, wherein, The at least one numerical measurement result and the at least one model measurement result are obtained in one or more of the axial view, sagittal view and coronal view.

8. The method according to any one of claims 1 to 7, wherein, The at least one numerical measurement result and the at least one model measurement result were obtained for one or more test groups within the applicable scope, pediatric and / or containing metals.

9. The method according to any one of claims 1 to 8, wherein, The image data is received from at least one of a CT imaging device or an O-arm imaging device.

10. The method of any one of claims 1 to 9, further comprising: The at least one numerical measurement result is averaged before the comparison step.

11. The method of any one of claims 1 to 10, further comprising: The segmented image data, the at least one numerical measurement result, and one or more trajectories are input into the segmentation measurement model, wherein the segmentation measurement model measures the at least one numerical measurement result in the segmented image.

12. A system for validating one or more segmentation models, the system comprising: Imaging equipment (112); Processor (104); as well as Memory (106), which stores data for processing by the processor, wherein the data, when processed, causes the processor to: Receive image data (202) depicting anatomical elements from the imaging device; The image data is input into one or more segmentation models (122), which are configured to segment the image data and output segmented image data (206). Receive at least one numerical measurement result of the anatomical element (212); Receive at least one model measurement result (208) of the anatomical element, wherein the at least one model measurement result is obtained by segmentation measurement model (124); Compare the measurement results of the at least one model with the measurement results of the at least one model; The error is determined based on the difference between the at least one numerical measurement result and the at least one model measurement result; and As a result of determining that the error is less than a predetermined error threshold, the one or more segmentation models are validated.

13. The system of claim 12, wherein, The imaging device includes at least one of a CT imaging device or an O-arm imaging device.

14. The system as claimed in claim 12 or 13, wherein, The at least one model measurement result is received from the segmentation measurement model, which is configured to receive segmented image data and the at least one numerical measurement result, and output the at least one model measurement result.

15. The system according to any one of claims 12 to 14, wherein, The at least one numerical measurement result includes at least one cortical bone thickness measurement result.

Images