Detecting Anatomical Abnormalities in 2D Medical Images

Abstract

The present invention relates to detecting anatomical abnormalities in medical images. A computer-implemented method 100 and system are disclosed that detects 2D contours 130 of anatomical features in medical images and compares these contours to predicted 2D contours 140 based on a 3D reference model to detect anatomical abnormalities 150. This approach may improve the accuracy of anatomical abnormality detection, thereby reducing time in medical facilities and potentially improving operator experience and patient outcomes.

Description

[Technical Field]

[0001] The present invention relates to a computer-implemented method, computer program product, and system configured for detecting anatomical abnormalities in 2D medical images. [Background technology]

[0002] Medical images are often used for diagnostic purposes. For example, medical images are examined by a human observer for anatomical abnormalities. Examples of such abnormalities include fractures and tumor tissue that may be identified in an x-ray, ultrasound, CT (computed tomography) or other medical image.

[0003] However, inspection of medical images by human observers can be time-consuming and prone to error, depending on the expertise, experience, time constraints, and fatigue of the human observer, especially in the challenging environment of a busy clinic and when anatomical abnormalities are not easily identifiable.

[0004] US Patent Application Publication No. 2022 / 0044041 discloses algorithms for detecting bone fractures, classifying the fractures, and providing guidance on how to treat the fractures. Summary of the Invention [Problem to be solved by the invention]

[0005] Among other things, it is an object of the present invention to provide computer-implemented anatomical abnormality detection in medical imaging. [Means for solving the problem]

[0006] The invention is defined by the independent claims. Advantageous embodiments are defined in the dependent claims.

[0007] One aspect of the present invention provides a computer-implemented method for detecting anatomical abnormalities in 2D medical images, the method comprising: - acquiring a 2D medical image; - detecting 2D contours of anatomical features; obtaining a 3D model representing an anatomical feature; - predicting 2D contours of anatomical features based on the 3D model; - detecting anatomical abnormalities based on the detected 2D contours and the predicted 2D contours; In this way, the 2D contours detected in the x-ray image can be compared to known reference contours obtained from the 3D model, such as to detect and identify differences that may be anatomical abnormalities. This approach can improve the accuracy of anatomical abnormality detection, thereby reducing time in medical facilities and potentially improving operator experience and patient outcomes.

[0008] The step of predicting a 2D contour of the anatomical feature based on the 3D model includes: - estimating the pose of an anatomical feature from a 2D medical image; - adjusting the 3D model to the estimated pose; generating a 2D projection from the pose-aligned 3D model; - predicting 2D contours of anatomical features from the 2D projections; In other words, based on the detected 2D contour, the observation position and posture of the imaged subject can be estimated. The 3D model can then be rotated and translated to match the position of the imaged anatomical feature relative to the virtual imaging system, and the 3D model can be adjusted by rotating its own joints, for example, by changing the flexion of the ankle joint, to match the posture of the imaged anatomical feature. This creates a healthy reference model with the same field of view and posture parameters as the imaged subject, which can be used to predict the 2D reference contour.

[0009] The step of detecting the 2D contour may further include a step of segmenting the 2D medical image. For example, individual pixels of the acquired medical image may be labeled as "bone / non-bone" to identify bones in the image and thereby detect the 2D contour. Alternatively, the step of detecting the 2D contour may include an end-to-end trained machine learning algorithm. In addition, the step of detecting the 2D contour may include identifying anatomical features. For example, identifying that the 2D contour is the contour of one or more bones of the ankle joint or wrist. This approach may be further used to optimize the workflow of any further processing or to provide guidance to a user or machine for further steps.

[0010] The method may further include classifying the anatomical abnormality. The step of classifying the anatomical abnormality may further be performed by a machine learning algorithm. In this way, the output of the computer-implemented method is not binary, e.g., whether an anatomical abnormality was detected or not, but provides the user with insight into the classification of the abnormality, e.g., whether it is severe or not, and / or whether it is a particular type of fracture and / or ligament rupture.

[0011] According to an embodiment of the present invention, the 3D model is -2D medical images, -anatomical features, and / or -User input is obtained based on

[0012] The 3D model is further - a computer-aided design model; a reference model constructed from 3D medical images of anatomical features in at least one pose; It may be one of the following.

[0013] In some examples, the method comprises: -Detected 2D contours and -Predicted 2D contours and - Steps that generate the difference between It further has:

[0014] Additionally, a difference map may be generated, said difference map containing information suitable for display on a display.

[0015] In some instances, the anatomical features include: - bone tissue, and -ligament tissue It is one of the following.

[0016] In some instances, the anatomical abnormality is - fractures, and -Ligament rupture It is one of the following.

[0017] In some examples, the step of classifying the anatomical abnormality comprises: -Weber classification, -Pauwels classification, -Smith and Colles classification, -Salter-Harris classification, and -AO / OTA classification Classify anatomic abnormalities according to one of the following:

[0018] In some examples, the medical image is an X-ray image.

[0019] Another aspect of the present invention provides a computer program product comprising instructions for enabling a processor to carry out an embodiment of the method described above.

[0020] Yet another aspect of the present invention provides a system for detecting anatomical abnormalities in 2D medical images, the system comprising: -Acquire 2D medical images, - Detects 2D contours of anatomical features, - Obtain a 3D model that represents your anatomical features, - predicting a 2D contour of an anatomical feature based on the 3D model, wherein predicting the 2D contour of the anatomical feature based on the 3D model comprises: - Estimating the pose of anatomical features from 2D medical images; - Adjusting the 3D model based on the estimated pose; generating a 2D projection from the pose-aligned 3D model; - predicting 2D contours of anatomical features from 2D projections; Including, Detecting anatomical abnormalities based on the detected and predicted 2D contours; The processor is configured to:

[0021] In one example, the system may further include an X-ray source and detector configured to capture an X-ray image.

[0022] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter. [Brief explanation of the drawings]

[0023] [Figure 1] 1 shows a flowchart according to one embodiment of the present invention. [Figure 2A] 1 shows a medical image according to an embodiment of the present invention. [Figure 2B] 1 illustrates detected contours of anatomical features in a medical image according to an embodiment of the present invention. [Figure 2C] 1 illustrates predicted contours of anatomical features in a medical image according to an embodiment of the present invention. [Figure 3] 4 illustrates pose and field of view parameters of an imaged object according to one embodiment of the present invention. [Figure 4] 1 illustrates a 3D model of an ankle joint according to an embodiment of the present invention. [Figure 5A] 1 illustrates ankle fracture classification according to an embodiment of the present invention. [Figure 5B] 1 illustrates classification of femoral neck fractures according to an embodiment of the present invention. [Figure 5C] 1 illustrates classification of fractures through the epiphysis or growth plate of a bone according to an embodiment of the present invention. [Figure 6] 1 illustrates a processor circuit according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0024] The present invention is described herein with reference to the drawings. It should be understood that the description and the specific examples provided, while indicating exemplary embodiments, are for purposes of illustration only and are not intended to limit the scope of the invention. It should also be understood that the drawings are merely schematic and are not drawn to scale. It should also be understood that the same reference numerals are used throughout the drawings to indicate the same or similar parts.

[0025] The present invention provides a computer-implemented method for analyzing and evaluating 2D (two-dimensional) medical images for the purpose of identifying anatomical abnormalities.

[0026] FIG. 1 shows an exemplary flow chart according to an embodiment of a method 100 according to the present invention, which includes the following steps:

[0027] In step 110, a 2D medical image 210 is acquired. The 2D medical image may be any type of medical image, including, but not limited to, for example, an X-ray image, a CT (Computed Tomography) image, an ultrasound image, an MR (Magnetic Resonance) image, or any other type of medical image.

[0028] An exemplary medical image 210 contemplated in the present invention is a 2D X-ray image of an ankle joint 211 shown in FIG. 2A.

[0029] In step 120, the 2D contours of the anatomical features are detected, which may involve, for example, segmentation of the 2D medical image or an end-to-end machine learning approach.

[0030] Generally, segmentation can be understood as the task of classifying an image into a label for each pixel. Alternatively, the classification can be for a larger image region as a whole, such as a contour or surface. The label represents the semantics of each pixel or image region. In this way, in a classification operation, for example, bone pixels, i.e., pixels with the label "bone," can be identified. In that case, the pixel is considered to represent bone tissue, and similarly for other tissue types. Thus, the footprint of an organ in an image can be identified by segmentation. In an alternative embodiment, segmentation may be performed using a neural network or a convolutional neural network, such as those disclosed in WO 2022 / 084074. Algorithms that include prior shape knowledge, such as those disclosed therein, are also contemplated by the present invention.

[0031] From image segmentation, a segmented image is typically obtained, with each pixel and / or region having a specific label assigned to it. This label may be binary "bone / not bone" or multi-level "bone / ligament tissue / not bone / ...". According to one embodiment of the present invention, a portion within the medical image corresponding to an anatomical feature of interest, e.g., "bone" or more specifically "tibia / fibula / calcaneus / talus", may be selected to generate a 2D contour 212, as shown in FIG. 2B.

[0032] In step 130, a 3D (three-dimensional) model 330 representing the imaged object 211 or the identified anatomical features 213 is obtained. the medical image 210 acquired in step 110, the anatomical features identified in step 120, and -User input The 3D model may be selected from the list of 3D models based on at least one of:

[0033] For example, the image 210 acquired in step 110 may be annotated to indicate that the image is of an ankle joint. In another example, an identified anatomical feature or a combination of identified anatomical features may indicate that the image includes at least an ankle joint. For example, a DICOM tag may be associated with the anatomical feature being imaged, which may be used to identify a 3D model. In yet another example, a user of a computing system executing method 100 may select a 3D model from a list of models. The user selection may be from the entirety of all 3D models or from a reduced set of articulated 3D models that may be pre-selected based on the 2D medical image and / or identified anatomical features within the 2D medical image. In another example, the 3D model may be selected based on contextual information at the acquisition location, e.g., a hospital radiology department, such as user diagnostic request data and / or radiology standard operating procedures for the requested procedure. For example, prior to image acquisition, a user or operator may indicate a desire to take an ankle joint X-ray image, which may determine a relevant 3D model or create a pre-selection of relevant 3D models. The 3D model may be articulated to allow for different postures of the imaged subject; for example, the 3D model may be able to reproduce a particular flexion of the ankle joint.

[0034] The representative 3D model may be a computer-aided design (CAD) model created with any CAD software, or a reference model extracted from 3D imaging using known techniques. For example, a series of MR images of a healthy subject's ankle joint may be acquired at various flexion angles. One image may then be manually or automatically segmented according to known techniques. The segmented image may then be used to create a surface model, for example, using a triangular mesh. For each triangle in the mesh, features for model adaptation may then be trained, for example, using image intensity, and the resulting model may be used to depict the corresponding bone surface in other MR images. This procedure may generate a 3D model for each MR image. The pose in each image and / or the corresponding 3D model may be further determined, for example, by measuring the angle between the major axis of the tibia and the major axis of the calcaneus for flexion. Smooth interpolation of the surface model between the varying poses recorded in the MR images may allow the mesh constellation to be estimated at any pose between them.

[0035] A 3D model 430 contemplated by the present invention is shown in Figure 4. The 3D model may be articulated and modified to fit any pose encountered in 2D medical images. For example, throughout the image sequence of Figure 3, the calcaneus is shown rotated relative to the tibia from 80 degrees to 135 degrees of flexion.

[0036] In step 140, the 2D contours of the anatomical features in the medical image are predicted from the 3D model. The contours are calculated by the following steps: -D medical images, estimating field of view parameters and / or pose of the imaged subject; -2 Step 144 of correcting or adjusting the 3D model according to the estimated field of view parameters and / or pose; - creating 146 a 2D projection from the view and / or pose adjusted 3D model, the projection representing a 2D medical image 210; - a step 148 of generating 2D contours 214 of the anatomical features shown in the 2D projection of the 3D model; is predicted according to

[0037] The visual field parameters and postures referenced in step 142 are, for example, as shown in FIG. 3 for the ankle joint: - flexion angle δ between the tibia and calcaneus; - line of sight, including azimuthal and polar rotation; and - detector position, translation, rotation and scaling for X-ray images It may be any one of the above.

[0038] This list should not be construed as exhaustive, and other field of view and / or pose parameters are also contemplated by the present invention.

[0039] Pose and / or field of view parameters may be estimated in step 142 according to S. Kronke et al. 2022 "CNN-based pose estimation for assessing quality of ankle-joint X-ray images."

[0040] For example, a mesh 3D model of the ankle joint can be sampled at random flexion angles and gaze directions, and projections can be generated by identifying mesh triangles in the 3D model that are traversed by the X-ray beam in a direction approximately tangential to the selected gaze direction, projecting their edges onto a virtual detector, semantically grouping them, and discretizing them on the detector pixel matrix. The resulting set of discretized contours in the projections forms the input for a pose estimation network. Such a pose estimation network can be a standard recurrent network containing a fully connected layer and a rectified linear unit (ReLU) activation function with alternating stride 1 and stride 2 convolution blocks. A network trained according to the above principles can then be used to predict the pose of the detected 2D contour. For example, the 2D contour of the imaged ankle joint can be fed into a neural network, which, upon training, can provide the flexion angle and gaze direction of the imaged ankle joint.

[0041] The pose and / or field of view parameters estimated in step 142 may then be used as input to step 144 in modifying or adjusting the 3D model to reproduce the respective pose and / or field of view parameters. For example, the flexion angle of the 3D ankle model 430 may be changed as shown in Figure 4, where the angle between the tibia and the calcaneus is changed from 80 degrees to 135 degrees.

[0042] After aligning the 3D model to the estimated pose, a 2D projection is determined in step 146, for example, by identifying mesh triangles in the 3D model that are traversed by the x-ray beam approximately tangential to the selected line of sight, projecting their edges onto a virtual detector, semantically grouping them, and discretizing them onto a detector pixel matrix, thereby also directly providing the predicted 2D contour in step 148. Figure 2C shows an exemplary 2D contour 214 predicted in steps 140, 142-148 from the 3D model of the ankle.

[0043] In an optional step, the difference between the 2D contour generated based on the 2D medical image and the 2D contour generated based on the articulated 3D model may be determined. This difference may be calculated, for example, based on point-by-point or pixel-by-pixel subtraction. In some implementations, a difference map may be created, which may also be visualized on a user interface (not shown).

[0044] In step 150, anatomical abnormalities are detected based on the 2D contours detected in step 120 and predicted in step 140, respectively. Anatomical abnormalities may be detected, for example, by visual inspection of a difference map displayed on a user interface (not shown). In another exemplary embodiment, anatomical abnormalities may be detected automatically, for example, based on whether a difference or deviation between (i) a 2D contour generated based on a 2D medical image and (ii) a 2D contour generated based on an articulated 3D model exceeds a certain threshold. In yet another exemplary embodiment, a machine learning algorithm may be used to detect anatomical abnormalities. For example, 2D medical images may be annotated by a user or expert to identify anatomical abnormalities. These annotated images may be processed according to steps 120 and 140 to detect 2D contours and predict 2D contours, where the detected 2D contours, the predicted 2D contours, and / or the differences between the detected and predicted 2D contours may be added to the annotated images for use in the training process.

[0045] The identification of an anatomical abnormality may be binary, present or absent, or non-binary, in which case a probability of the anomaly being present may be provided. In an advantageous embodiment, the anatomical abnormality may also be classified in optional step 160 based on size, location, or any other form of classification to indicate a level of severity. That is, the anatomical abnormality may be classified into multiple categories. FIG. 5A, discussed in more detail below, illustrates classification of, for example, an ankle fracture, and FIG. 5B, discussed in more detail below, illustrates classification of, for example, a femoral neck fracture. According to one embodiment of the present disclosure, classification of the anatomical abnormality is performed using a trained machine learning algorithm. For example, a classification network may be trained using annotated data such as those shown in FIGS. 5A-5C and using the detected 2D contour 212 and the predicted 2D contour 214 as input.

[0046] For training a classification machine learning model, annotated medical images are provided. For example, an ankle x-ray image may have annotations indicating whether a fracture is present and the classification of the fracture, e.g., according to the Weber classification of FIG. 5A . The annotation may be performed by an expert, such as a clinician, radiologist, researcher, or any other qualified person. As an example, the annotated image may be an ankle x-ray image 210 with an associated label of no fracture. Another example may be an ankle x-ray image with a fracture and a corresponding label identifying the classification of the fracture, e.g., whether the fracture is a type A, B, or C fracture according to the Weber classification.

[0047] Similar to the images analyzed by method 100, the annotated images used for training may also be - Detects 2D contours of anatomical features, Predicting 2D contours of anatomical features from a healthy reference, e.g., from the 3D model obtained in step 130 To do so, steps 120 and 140 of method 100 are performed.

[0048] The detected 2D contours and predicted 2D contours may then form input to a machine learning model, and annotations from the training images may form output. As usual, training is performed using a subset of the training images and corresponding annotations, and the remaining training images and corresponding annotations are used to evaluate the trained model. Training is stopped when a predetermined criterion is reached, such as to ensure that the model is sufficiently trained but not overtrained. In some implementations, noise may also be added manually or automatically to the training data to avoid overfitting.

[0049] Although method 100 has been presented in a particular order, it will be understood by those skilled in the art that the order of the steps may be changed, additional steps may be added between, and / or steps may be removed.

[0050] It will further be understood that individual steps in method 100 may be performed in real time, e.g., during or after an examination procedure. For example, 2D medical images may be acquired in real time and analyzed according to method 100 during the image acquisition procedure, or 2D medical images may be acquired initially during an acquisition step and analyzed only upon request of an operator and / or specialist. Thus, method 100 may be performed within the image acquisition system or off-system in a separate computing unit. In another embodiment, method 100 may be performed in the cloud.

[0051] Predefined classifications as envisioned by the present invention are disclosed in Figures 5A to 5C. The classifications in these figures are merely exemplary and in no way limit the ideas of the present invention.

[0052] Figure 5A shows the Weber classification of the ankle. In Figure 5A, three different types of fractures are shown: Type A 511, Type B 512, and Type C 513, which can be described as follows according to Wikipedia (https: / / en.wikipedia.org / wiki / Danis%E2%80%93Weber_classification): - Type A 511 - Fracture of the fibula distal to the syndesmosis (the connection between the distal ends of the tibia and fibula). Typical features are: -Below ankle level -Intact tibiofibular syndesmosis -Intact deltoid ligament -Sometimes fractured medial malleolus - Usually stable; occasionally, however, open reduction and internal fixation (ORIF) is required, especially when the medial malleolus is fractured is. - Type B 512 - Fracture of the fibula at the level of the syndesmosis. Typical features: - Extends above and laterally to the fibula at the level of the ankle joint - Intact or only partially torn syndesmotic ligament, but without enlargement of the distal tibiofibular joint - Possible fracture of the medial malleolus or rupture of the deltoid ligament - Variable stability. - Type C 513 - Fracture of the fibula proximal to the syndesmosis. Typical features: -Above ankle level -Disrupted tibiofibular syndesmosis with extension of the distal tibiofibular joint -Presence of medial malleolus fracture or deltoid ligament injury -Unstable: Requires ORIF.

[0053] Type B512 and Type C513 fractures represent a degree of damage to the syndesmosis itself (which may not be directly visualized on an x-ray). They are inherently unstable and are more likely to require surgical repair to achieve a good outcome. Type A fractures are usually stable and can be treated with simple measures, such as a plaster cast. Therefore, accurate identification of ankle fractures by Weber classification may improve the outcome and / or recovery of the imaged subject.

[0054] Figure 5B shows the Pauwels classification of femoral neck fractures. These fractures are characterized by the fracture angle. Increasing angles result in more unstable fractures and increased shear stress at the fracture site. This shear leads to a higher rate of nonunion. According to Wikipedia (https: / / en.wikipedia.org / wiki / Pauwel%27s_angle), the Pauwels classification divides fractures into Type I 521, Type II 522, and Type III 523 fractures as follows: - Type I 521 - Fracture angle less than 30 degrees Type II 522 - Fracture angle of 30 to 50 degrees - Type III 523 - Fracture angle of 50° or more.

[0055] Figure 5C shows the Salter-Harris classification of fractures involving the epiphysis or growth plate of a bone, specifically the zone of pseudocalcification. It is therefore a form of pediatric fracture, occurring in 15% of pediatric long bone fractures. There are nine types of Salter-Harris fractures: Types I 531 through V 535 (shown in Figure 5C), as described by Robert B. Salter and W. Robert Harris in 1963, and the rarer types VI through IX (not shown or discussed here), which were added later. According to Wikipedia (https: / / en.wikipedia.org / wiki / Salter%E2%80%93Harris_fracture), the first five types can be described as follows: - Type I 531 - Transverse fracture through the growth plate (also called "epiphysis"); -Type II 532 - Fractures through the growth plate and metaphysis, sparing the epiphysis: healing takes approximately 12 to 90 weeks or more in the spine; - Type III 533 - Fracture through the growth plate and epiphysis sparing the metaphysis; - Type IV 534 - fracture through all three elements of the bone: growth plate, metaphysis, epiphysis; -Type V 535-Compression fracture of the growth plate (resulting in a reduction in the perceived space between the epiphysis and metaphysis on x-ray).

[0056] 5A-5C are provided as a reference for illustrating possible classifications according to embodiments of the present invention. In particular, the use of predefined classes allows an operator to see not only that a fracture may have occurred, but also what type of fracture may have occurred, and thus may improve subsequent steps to improve the outcome and / or recovery of the imaged subject.

[0057] Another predefined classification may be the AO / OTA (Arbeitsgemeinschaft fur Osteosynthesefragen / Orthopaedic Trauma Association) classification, which defines fractures according to anatomical location, fracture type, and fracture subgroup. -Anatomical location follows the AO / OTA classification, described by a two-digit number, where the first digit indicates the major anatomical division (e.g., upper arm (humerus), lower arm, thigh, lower leg (tibia / fibula)), and the second digit indicates the specific part of the bone involved (e.g., proximal, middle, distal).

[0058] - Fracture type coded as type A, B or C, - Type A indicates a simple fracture; - Type B indicates a wedge fracture with more than two fragment buses until there is some contact between the main fragments; - Type C indicates a complex or comminuted fracture with potentially multiple separated fragments.

[0059] Fracture subgroups, indicated by a single digit, provide additional detail regarding the nature of the fracture, such as the fracture pattern or configuration (eg, is the fracture an oblique fracture, a transverse fracture, a spiral fracture, etc.).

[0060] While predetermined classifications for bone and / or ligament fractures are described above, other types of classifications are also contemplated by the present invention, including, but not limited to, cancer type and / or stage. For example, a segmented tumor may be automatically classified according to its size and / or location as a stage I, stage II, stage III, or stage IV tumor.

[0061] Method 100 may further include the optional step 170 of providing treatment advice based on the identified anatomical abnormality and / or the classification result of the identified anatomical abnormality. For example, the treatment advice may be immediate surgery required for a Weber type C ankle fracture, while rest and / or cast placement may be advised for hairline and / or stress fractures. The advice may be based on annotated data, for example, the classified image may also be annotated to provide treatment advice according to the above classification.

[0062] 6 is a schematic diagram of a processor circuit 600 according to an embodiment of the present disclosure. As shown, the processor circuit 600 may include a processor 606, a memory 603, and a communication module 608. These elements may communicate with each other directly or indirectly, for example, via one or more buses.

[0063] The processor 606 contemplated by this disclosure may include a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a controller, a field-programmable gate array (FPGA) device, another hardware device, a firmware device, or any combination thereof configured to perform the operations described herein. The processor 606 may also be implemented as a combination of computing devices, such as a DSP and a microprocessor, multiple microprocessors, one or other microprocessors in conjunction with a DSP core, or any other such configuration. The processor 606 may also implement various deep learning networks, which may include hardware or software implementations. The processor 606 may further include a preprocessor in either a hardware or software implementation.

[0064] The memory 603 contemplated by the present disclosure may be any suitable storage device, such as cache memory (e.g., cache memory of the processor 606), random access memory (RAM), magnetoresistive RAM (MRAM), read-only memory (ROM), field programmable gate array read-only memory (PROM), erasable field programmable gate array read-only memory (EPROM), electrically erasable field programmable gate array read-only memory (EEPROM), flash memory, solid-state memory devices, hard disk drives, other forms of volatile and non-volatile memory, or a combination of different types of memory. The memory may be distributed across multiple memory devices and / or located remotely relative to the processor circuitry. In one embodiment, the memory 603 may store instructions 605. The instructions 605 may include instructions that, when executed by the processor 606, cause the processor 606 to perform the operations described herein.

[0065] The instructions 605 may also be referred to as code. The terms “instructions” and “code” should be interpreted broadly to include any type of computer-readable statement. For example, the terms “instructions” and “code” may refer to one or more programs, routines, subroutines, functions, procedures, etc., and “instructions” and “code” may include a single computer-readable statement or many computer-readable statements. The instructions 205 may be in the form of an executable computer program or script. For example, the routines, subroutines, and / or functions may be defined in a programming language, including, but not limited to, C, C++, C#, Pascal, BASIC, API calls, HTML, XHTML, XML, ASP script, JavaScript, FORTRAN, COBOL, Perl, Java, ADA, .NET, etc. The instructions may also include instructions for machine learning and / or deep learning.

[0066] The communications module 608 may include any electronic and / or logical circuitry for facilitating direct or indirect communication of data between the processor circuit 600 and, for example, an external display (not shown) and / or an imaging device or system, such as an X-ray imaging system. In this regard, the communications module 608 may be an input / output (I / O) device. Communications may be performed via any suitable means. For example, the communications means may be a wired link, such as a Universal Serial Bus (USB) link or an Ethernet link. Alternatively, the communications means may be a wireless link, such as an Ultra-Wideband (UWB) link, an Institute of Electrical and Electronics Engineers (IEEE) 802.11 Wi-Fi link, or a Bluetooth link.

[0067] It should be noted that the above-described embodiments illustrate rather than limit the present invention, and that those skilled in the art can design many alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps other than those listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The implementation may be by means of hardware comprising several distinct elements and / or by a suitably programmed processor. In a device claim enumerating several means, several of these means may be embodied by one and the same item of hardware. Measures recited in mutually different dependent claims may be advantageously used in combination.

Claims

1. 1. A computer-implemented method for detecting anatomical abnormalities in 2D medical images, comprising: acquiring the 2D medical image; Detecting 2D contours of anatomical features; obtaining a 3D model representing the anatomical feature; predicting a 2D contour of the anatomical feature based on the 3D model, the predicting the 2D contour of the anatomical feature based on the 3D model comprising: estimating the pose of the anatomical features from the 2D medical images; adjusting the 3D model based on the estimated pose; generating a 2D projection from the pose-aligned 3D model; and predicting a 2D contour of the anatomical feature from the 2D projection; and detecting the anatomical abnormality based on the detected 2D contour and the predicted 2D contour; A method having the following.

2. The step of detecting the 2D contour comprises: segmenting the 2D medical image; The method of claim 1 , comprising:

3. classifying the anatomical abnormality; 3. The method of claim 1, further comprising:

4. The method of claim 3 , wherein the step of classifying the anatomical abnormality is performed by a machine learning algorithm.

5. The 3D model is the 2D medical image; said anatomical features, and / or user input, The method according to claim 1 , wherein the value is obtained based on:

6. The 3D model is Computer-aided design models, and a reference model constructed from 3D medical images of the anatomical feature in at least one pose; 6. The method according to claim 1, wherein the method is one of:

7. the detected 2D contour; the predicted 2D contour line; generating a difference between The method of any one of claims 1 to 6, further comprising:

8. The anatomical feature is: bone tissue, and ligament tissue, 8. The method according to claim 1, wherein the method is one of:

9. The anatomical abnormality is fractures, and ligament rupture, 9. The method according to claim 1, wherein the method is one of:

10. The step of classifying the anatomical abnormality comprises: Weber classification, Pauwels classification, Smith and Colles classification, Salter-Harris classification, and AO / OTA classification, 5. The method of claim 3, wherein the anatomical abnormality is classified based on one of:

11. generating a treatment recommendation based on the detected anatomical abnormality; 11. The method of claim 1, further comprising:

12. The method of claim 1 , wherein the medical image is an X-ray image.

13. A computer program comprising instructions for enabling a processor to carry out the method according to any one of claims 1 to 12.

14. A system for detecting anatomical abnormalities in 2D medical images, the system comprising a processor configured to perform the method of any one of claims 1 to 13.

15. The system of claim 14 further comprising an x-ray source and an x-ray detector.

Images