Automatic slice selection in medical imaging

Abstract

The present invention provides a medical imaging system (100, 300, 400, 700), comprising: a memory (110) for storing machine executable instructions (120) and a processor (106) for controlling the medical imaging system. Execution of the machine executable instructions causes the processor to: receive (200) three-dimensional medical image data (122) comprising a plurality of slices; receive (202) an imaging modality (124) of the three-dimensional medical image data; receive (204) an anatomical view classification (126) of the three-dimensional medical image data; select (206) a selected anomaly detection module (130) from a set of anomaly detection modules (128) using the imaging modality and the anatomical view classification, wherein at least a portion of the anomaly detection modules are convolutional neural networks trained to identify at least a portion of the plurality of slices as normal or abnormal; classify (208) the at least a portion of the plurality of slices as normal or abnormal using the anomaly detection module; and select (210) a set of selected slices (136) from the plurality of slices according to a predetermined selection criterion (134) if a predetermined number of the plurality of slices are classified as abnormal.

Description

Technical Field

[0001] The present invention relates to medical imaging, and in particular to the identification of abnormal slices. Background Art

[0002] Medical imaging techniques (magnetic resonance imaging (MRI), computed tomography (CT), positron emission tomography (PET), and other techniques) are becoming increasingly important in modern medical care. Various imaging modalities are being used more and more over time. Moreover, the amount of data collected by medical imaging scanners has also increased. In some hospitals, radiologists will spend most of their time visually scanning a large number of images after the imaging protocol has been completed. Summary of the invention

[0003] The invention provides a medical imaging system, a computer program product and a method in the independent claims. Embodiments are given in the dependent claims.

[0004] Embodiments of the present invention can provide means for reducing the workload when examining three-dimensional medical image data. Embodiments can do so by first receiving an imaging modality of an image that identifies the type of image. For example, the imaging modality can identify the three-dimensional medical image data as an MRI image, a CT image, or other medical imaging modality. Then receive an anatomical view classification. The anatomical view classification can, for example, identify the anatomical structures present in the three-dimensional medical image data. In another example, the anatomical view classification can identify the imaging protocol used. For example, the anatomical view classification can identify the specific type of MRI scan of the brain, shoulder, or other anatomical structure.

[0005] The imaging modality and anatomical view classification are then used to select a selected anomaly detection module from a set of anomaly detection modules. The set of anomaly detection modules are all convolutional neural networks trained for a specific imaging modality and a specific anatomical view classification. The selected anomaly detection module is then used to examine at least a portion of the slices of the three-dimensional medical image data to see if any are classified as abnormal. If any of the slices are abnormal, then predetermined selection criteria are used to select a set of selected slices from the three-dimensional image data. The set of selected slices can be displayed on a user interface or identified by a radiologist at a later examination, for example.

[0006] In one aspect, the present invention provides a medical imaging system comprising a memory for storing machine executable instructions. The medical imaging system also comprises a processor for controlling the medical imaging system. Execution of the machine executable instructions causes the processor to receive three-dimensional medical image data comprising a plurality of slices. The three-dimensional medical image data is data that can be used to draw a 3D volume or can be displayed as a plurality of two-dimensional slices of a three-dimensional volume. Execution of the machine executable instructions also causes the processor to receive an imaging modality of the three-dimensional medical image data. The imaging modality can be, for example, the type of imaging scanner or device used to acquire the three-dimensional medical image data.

[0007] Execution of the machine executable instructions also causes the processor to receive an anatomical view classification of the three-dimensional medical image data. The anatomical view classification may, for example, indicate the type of imaging procedure used to acquire the three-dimensional medical image data, or it may include a list or list of anatomical regions within the three-dimensional medical image data. Execution of the machine executable instructions also causes the processor to select a selected anomaly detection module from a set of anomaly detection modules using the imaging modality and the anatomical view classification. Each of the anomaly detection modules is a convolutional neural network trained to identify at least a portion of the multiple slices as normal or abnormal. In some specific examples, the at least a portion of the multiple slices is each of the multiple slices. Reference to "each of the multiple slices" may also be interpreted as at least a portion of the multiple slices.

[0008] By having a set of anomaly detection modules, the selected anomaly detection modules can be trained for a specific imaging modality and a specific anatomical view represented in the three-dimensional medical image data. Execution of the machine executable instructions also causes the processor to classify the at least a portion of the plurality of slices as normal or abnormal using the anomaly detection modules.

[0009] For example, each of the slices may be input into the convolutional neural network, and the output is that the slice is classified as normal or abnormal. Execution of the machine executable instructions further causes the processor to: if a predetermined number of the plurality of slices is classified as abnormal, select a set of selected slices from the plurality of slices according to a predetermined selection criterion. In some examples, the predetermined number of the plurality of slices is one. It should be understood herein that "the predetermined number of the plurality of slices" can be replaced with "any of the plurality of slices".

[0010] If the three-dimensional medical imaging data includes any slice identified as abnormal by the convolutional neural network, then a set of three-dimensional medical image data is selected as the set of selected slices.

[0011] This embodiment can be beneficial because the set of selected slices can be selected to enable a physician to identify one or more of the plurality of slices identified as abnormal. This can help radiologists or other technicians browse three-dimensional medical image data to find abnormalities.

[0012] In another embodiment, execution of the machine executable instructions further causes the processor to calculate the entropy of each of the plurality of slices. Execution of the machine executable instructions further causes the processor to calculate the mean squared intensity variation between adjacent slices for each of the plurality of slices. Execution of the machine executable instructions further causes the processor to: if the ratio of the entropy of the slice to the mean squared intensity variation between the adjacent slices is above a predetermined information content threshold, then add the selected slice of the plurality of slices to the set of selected slices. In this embodiment, slices with high entropy relative to the mean squared intensity variation are selected. This can be beneficial because, depending on how the predetermined information content threshold is selected, the image can contain more than a common amount of information. This can select slices with more details or information that can be beneficial to the operator's inspection.

[0013] In another embodiment, the anomaly detection module is configured to generate a feature vector for each of the plurality of slices. The calculation of the feature vector is any normal feature of a convolutional neural network. The execution of the machine executable instructions causes the processor to calculate the Euclidean distance between the feature vectors of adjacent slices for each of the plurality of slices. The execution of the machine executable instructions also causes the processor to calculate a likelihood measure for each of the plurality of slices by inputting the feature vector of each of the plurality of slices into a Gaussian mixture model. The Gaussian mixture model is trained to determine the probability of whether the feature vector is normal and abnormal. The execution of the machine executable instructions also causes the processor to: if the ratio of the likelihood measure to the Euclidean distance is greater than a predetermined abnormality measure threshold, then add a selected slice from the plurality of slices to the set of selected slices. This embodiment can be beneficial because a slice with a high probability of containing an abnormality as identified by a Gaussian mixture model can be selected.

[0014] In another embodiment, execution of the machine executable instructions further causes the processor to train the Gaussian mixture model using the feature vectors derived from the anomaly detection module.This embodiment may be beneficial because it may provide an effective and efficient way to train a Gaussian mixture model.

[0015] In another embodiment, execution of the machine executable instructions further causes the processor to determine the imaging modality of the three-dimensional medical image using an imaging modality classifier. For example, a neural network may be trained to receive a medical image as input and then output the type of imaging modality. In other embodiments, the imaging modality classifier may be code that looks at a header or metadata or a log file and uses that information to determine the imaging modality.

[0016] In another embodiment, the imaging modality classifier is configured to determine the imaging modality using any of: processing a report or log file using natural language processing, extracting the imaging modality from a DICOM header, and receiving the imaging modality from a health information system.

[0017] In another embodiment, execution of the machine executable instructions further causes the processor to determine an anatomical view classification using an anatomical classification module. The anatomical structure classification module is a trained convolutional neural network. In this embodiment, the trained convolutional neural network can be used to identify a specific anatomical view of the three-dimensional medical image data. This can be useful because it can then be used to pre-select the selected anomaly detection module.

[0018] In another embodiment, the imaging modality is any one of the following: magnetic resonance imaging, positron emission tomography, single photon emission tomography, ultrasound, X-ray, computed tomography, and endoscopy.

[0019] In another embodiment, execution of the machine executable instructions further causes the processor to receive a plurality of sets of three-dimensional medical image data.

[0020] In another embodiment, the medical imaging data includes a display. Execution of the machine executable instructions further causes the processor to display the set of selected slices on the display. This can be beneficial because it can save a medical professional or operator from browsing a large number of slices before finding an abnormality.

[0021] In another embodiment, the medical imaging system includes a display. Execution of the machine executable instructions further causes the processor to display a selected portion of the three-dimensional medical image data on the display. The selected portion includes the set of selected slices and slices of the three-dimensional medical image data between the set of selected slices. In this embodiment, the set of selected slices and the slices of the three-dimensional medical image data between them are displayed. This can be beneficial because it can provide more information for a medical professional or operator to find abnormalities.

[0022] In another embodiment, the medical imaging system includes a medical imaging scanner configured to acquire the three-dimensional medical image data. Execution of the machine executable instructions further causes the processor to control the medical imaging scanner to acquire the three-dimensional medical imaging data.

[0023] In another embodiment, the medical imaging scanner is any one of the following: a magnetic resonance imaging system, a positron emission tomography system, a single photon emission tomography system, an ultrasound imaging system, an X-ray system, a computed tomography system, and an endoscopy system.

[0024] In another aspect, the present invention provides a method for operating a medical imaging system. The method includes receiving three-dimensional medical image data including a plurality of slices. The method also includes receiving an imaging modality of the three-dimensional medical image data. The method also includes receiving an anatomical view classification of the three-dimensional medical image data. The method also includes selecting a selected anomaly detection module from a set of anomaly detection modules using the imaging modality and the anatomical view classification. Each of the anomaly detection modules is a convolutional neural network trained to identify each of the plurality of slices as normal or abnormal. The method also includes classifying each of the plurality of slices as normal or abnormal using the anomaly detection module. The method also includes: if any of the plurality of slices is classified as abnormal, selecting the set of selected slices from the plurality of slices according to a predetermined selection criterion.

[0025] In another aspect, the present invention provides a computer program product comprising machine executable instructions for execution by a processor controlling a medical imaging system. Execution of the machine executable instructions causes the processor to receive three-dimensional medical image data comprising a plurality of slices. Execution of the machine executable instructions also causes the processor to receive an imaging modality of the three-dimensional medical image data. Execution of the machine executable instructions also causes the processor to receive an anatomical view classification of the three-dimensional medical image data. Execution of the machine executable instructions also causes the processor to select a selected anomaly detection module from a set of anomaly detection modules using the imaging modality and the anatomical view classification. Each of the anomaly detection modules is a convolutional neural network trained to identify each of the plurality of slices as normal or abnormal.

[0026] Execution of the machine executable instructions further causes the processor to classify each of the plurality of slices as normal or abnormal using the anomaly detection module. Execution of the machine executable instructions further causes the processor to select the set of selected slices from the plurality of slices according to a predetermined selection criterion if any of the plurality of slices is classified as abnormal.

[0027] It should be understood that one or more of the foregoing embodiments of the present invention may be combined as long as the combined embodiments are not mutually exclusive.

[0028] Those skilled in the art will appreciate that aspects of the present invention may be implemented as an apparatus, method, or computer program product. Thus, aspects of the present invention may take the form of a complete hardware embodiment, a complete software embodiment (including firmware, resident software, microcode, etc.), or an embodiment combining software aspects and hardware aspects, all of which are generally referred to herein as "circuits," "modules," or "systems." In addition, aspects of the present invention may take the form of a computer program product embodied in one or more computer-readable media having a computer executable code embodied thereon.

[0029] Any combination of one or more computer-readable media can be utilized. Computer-readable media can be computer-readable signal media or computer-readable storage media. "Computer-readable storage media" used herein includes any tangible storage media that can store instructions that can be run by a processor of a computing device. Computer-readable storage media can be referred to as computer-readable non-transient storage media. Computer-readable storage media can also be referred to as tangible computer-readable media. In some embodiments, computer-readable storage media can also store data that can be accessed by a processor of a computing device. Examples of computer-readable storage media include, but are not limited to, floppy disks, magnetic hard disk drives, solid-state hard disks, flash memories, USB thumb drives, random access memories (RAM), read-only memories (ROM), optical disks, magneto-optical disks, and register files of processors. Examples of optical disks include compact disks (CDs) and digital versatile disks (DVDs), such as CD-ROMs, CD-RWs, CD-Rs, DVD-ROMs, DVD-RWs, or DVD-R disks. The term computer-readable storage media also refers to various types of recording media that can be accessed by a computer device via a network or communication link. For example, data can be retrieved on a modem, on the Internet, or on a local area network. Computer executable code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0030] A computer readable signal medium may include a propagated data signal having computer executable code embodied therein (e.g., in baseband or as part of a carrier wave). Such a propagated signal may take any of the following forms, including, but not limited to, electromagnetic, optical, or any suitable combination thereof. A computer readable signal medium may be any computer readable medium that is not a computer readable storage medium and that is capable of conveying, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0031] "Computer memory" or "memory" is an example of a computer-readable storage medium. Computer memory is any memory that can be directly accessed by a processor. "Computer storage" or "storage device" is another example of a computer-readable storage medium. A computer storage device can be any non-volatile computer-readable storage medium. In some embodiments, a computer storage device can also be computer memory or vice versa.

[0032] As used herein, a "processor" includes an electronic component capable of running a program or machine executable instructions or computer executable code. References to computing devices including a "processor" should be interpreted as possibly including more than one processor or processing core. A processor may, for example, be a multi-core processor. A processor may also refer to a collection of processors within a single computer system or distributed among multiple computer systems. The term "computing device" should also be interpreted as possibly referring to a collection or network of multiple computing devices, each of which includes one or more processors. Computer executable code may be executed by multiple processors, which may be within the same computing device, or may even be distributed across multiple computing devices.

[0033] Computer executable code may include machine executable instructions or programs that cause a processor to perform an aspect of the present invention. Computer executable code for performing operations for various aspects of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​(e.g., Java, Smalltalk, C++, etc.) and conventional procedural programming languages ​​(e.g., "C" programming language or similar programming languages), and compiled into machine executable instructions. In some instances, the computer executable code may be in the form of a high-level language or in a precompiled form, and may be used in conjunction with an interpreter that generates machine executable instructions on the fly.

[0034] The computer executable code may run entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0035] Various aspects of the present invention are described with reference to the flowchart illustration and / or block diagram of the method, device (system) and computer program product according to an embodiment of the present invention. It should be understood that each frame or part of the frame of the flowchart, diagram and / or block diagram can be implemented by the computer program instruction in the form of computer executable code when appropriate. It should also be understood that when mutually non-exclusive, the frames in different flowcharts, diagrams and / or block diagrams can be combined. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer or other programmable data processing device to produce a machine, so that the instructions run by the processor of the computer or other programmable data processing device create a unit for implementing the function / action specified in one or more frames of the flowchart and / or block diagram.

[0036] These computer program instructions may also be stored in a computer-readable medium that can direct a computer, other programmable data processing apparatus, or other device to function in a particular manner so that the instructions stored in the computer-readable medium produce an article of manufacture that includes instructions for implementing the functions / actions specified in the flowchart and / or one or more block diagram blocks.

[0037] The computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operating steps to be performed on the computer, other programmable apparatus, or other devices, thereby producing a computer-implemented process, so that the instructions running on the computer or other programmable apparatus provide a process for implementing the functions / actions specified in the flowchart and / or one or more block diagram blocks.

[0038] "User interface" used herein is an interface that allows a user or operator to interact with a computer or computer system. "User interface" can also be referred to as "human-machine interface device". The user interface can provide information or data to the operator and / or receive information or data from the operator. The user interface can enable the input from the operator to be received by the computer, and can provide output to the user from the computer. In other words, the user interface can allow the operator to control or manipulate the computer, and the interface can allow the computer to indicate the effect of the control or manipulation of the operator. The display of data or information on a display or a graphical user interface is an example of providing information to the operator. Receiving data by keyboard, mouse, tracking ball, touch pad, pointing stick, graphic input board, joystick, game handle, game pad, webcam, head mounted device, foot pedal, wired gloves, remote controller and accelerometer is all examples of user interface components that enable receiving information or data from the operator.

[0039] As used herein, "hardware interface" encompasses an interface that enables a processor of a computer system to interact with and / or control an external computing device and / or apparatus. A hardware interface may allow a processor to send a control signal or instruction to an external computing device and / or apparatus. A hardware interface may also enable a processor to exchange data with an external computing device and / or apparatus. Examples of hardware interfaces include, but are not limited to, a universal serial bus, an IEEE 1394 port, a parallel port, an IEEE 1284 port, a serial port, an RS-232 port, an IEEE-488 port, a Bluetooth connection, a wireless LAN connection, a TCP / IP connection, an Ethernet connection, a control voltage interface, a MIDI interface, an analog input interface, and a digital input interface.

[0040] As used herein, a "display" or "display device" encompasses an output device or user interface suitable for displaying images or data. A display may output visual, auditory, and / or tactile data. Examples of displays include, but are not limited to, computer monitors, television screens, touch screens, tactile electronic displays, Braille screens, cathode ray tubes (CRTs), memory tubes, bi-stable displays, electronic paper, vectorscopes, flat panel displays, vacuum fluorescent displays (VFs), light emitting diode (LED) displays, electroluminescent displays (ELDs), plasma display panels (PDPs), liquid crystal displays (LCDs), organic light emitting diode displays (OLEDs), projectors, and head mounted displays.

[0041] Magnetic resonance (MR) data or magnetic resonance imaging data is defined herein as the measurement of radio frequency signals emitted by atomic spins recorded using the antenna of a magnetic resonance device during a magnetic resonance imaging scan. A magnetic resonance imaging (MRI) image or MR image is defined herein as a reconstructed two-dimensional or three-dimensional visualization of anatomical data contained within the magnetic resonance imaging data. This visualization can be performed using a computer. A magnetic resonance image is an example of medical image data or three-dimensional medical image data. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In the following, preferred embodiments of the present invention will be described by way of example only and with reference to the accompanying drawings, in which:

[0043] Figure 1 An example of a medical imaging system is illustrated;

[0044] Figure 2 The diagram shows the operation Figure 1 A flowchart of an example of a method of a medical imaging system;

[0045] Figure 3 Another example of a medical imaging system is illustrated;

[0046] Figure 4 Another example of a medical imaging system is illustrated;

[0047] Figure 5 An example of a neural network is shown;

[0048] Figure 6 illustrates yet another example of a neural network; and

[0049] Figure 7 Yet another example of a medical imaging system is illustrated.

[0050] Reference numerals list

[0051] 100 Medical Imaging Systems

[0052] 102 Computer Systems

[0053] 104 Hardware / Network Interface

[0054] 106 processors

[0055] 108 User Interface

[0056] 110 Computer memory

[0057] 112 Graphical User Interface

[0058] 114 Slice Drawing

[0059] 116 Control device for flipping through slices

[0060] 120 machine executable instructions

[0061] 122 3D medical image data

[0062] 124 Imaging modalities

[0063] 126 Anatomical View Classification

[0064] 128 A set of anomaly detection modules

[0065] 130 Selected anomaly detection module

[0066] Classification of each of the 132 slices

[0067] 134 Pre-selection criteria

[0068] 136 A set of selected slices

[0069] 138 Imaging Modality Classifier

[0070] 140 Anatomical structure classification module

[0071] 200 receiving three-dimensional medical image data including a plurality of slices

[0072] 202 Imaging modality for receiving three-dimensional medical image data

[0073] 204 receiving anatomical view classification of three-dimensional medical image data

[0074] 206 Selecting a selected one from a set of anomaly detection modules using imaging modality and anatomical view classification

[0075] Anomaly detection module

[0076] 208 Classify each of the multiple slices as normal or abnormal using an anomaly detection module

[0077] 210 If any of the plurality of slices is classified as abnormal, select the slices from the plurality of slices according to a predetermined selection criterion.

[0078] Select a group of selected slices from the slices

[0079] 300 Medical Imaging Systems

[0080] 302 Magnetic Resonance Imaging System

[0081] 304 Magnet

[0082] 306 Magnet bore

[0083] 308 Imaging Area

[0084] 309 Area of ​​Interest

[0085] 310 Magnetic Field Gradient Coil

[0086] 312 Magnetic Field Gradient Coil Power Supply

[0087] 314 RF Coil

[0088] 316 Transceiver

[0089] 318 objects

[0090] 320 Object Support

[0091] 340 Pulse sequence command

[0092] 342 MRI data

[0093] 400 Medical Imaging Systems

[0094] 402 Image Modality Classification Model

[0095] 402 Imaging Anatomy Model

[0096] 402'CT anatomical structure classification module

[0097] 404 anomaly detection model

[0098] 404'CT brain abnormality model

[0099] 500 Input layer

[0100] 502 Hidden Layers

[0101] 504 Output layer

[0102] 506 Output

[0103] 600 feature maps

[0104] 602 hidden units

[0105] 604 Eigenvectors

[0106] 606 Output

[0107] 700 Medical Imaging Systems

[0108] 702CT

[0109] 704MR(MRI)

[0110] 706 Ultrasound

[0111] 708 X-ray

[0112] 710 Pathology

[0113] 712 Endoscope

[0114] 714 Brain

[0115] 716 Limbs

[0116] 718 Abdomen

[0117] 720 Lungs DETAILED DESCRIPTION

[0118] In these drawings, like numbered elements are equivalent elements or perform the same function. If the function is equivalent, an element that has been discussed previously will not necessarily be discussed in the following drawings.

[0119] Figure 1 An example 100 of a medical imaging system is illustrated. The medical imaging system is shown to include a computer system 102. The computer system 102 has a hardware and / or network interface 104 that can be used to control other components and / or communicate with other computer systems via a network. The computer system 102 is also shown to include a processor 106 that is connected to the hardware and / or network interface 104 and to a user interface 108 and a computer memory 110. The memory 110 can be any combination of memories accessible to the processor 106. This can include things like main memory, cache memory, and non-volatile memory, such as flash RAM, a hard drive, or other storage devices. In some examples, the memory 110 can be considered to be a non-transient computer readable medium.

[0120] The user interface 108 may include a display that renders a graphical user interface 112. The graphical user interface 112 may have an area for rendering a slice or two-dimensional image. There may also be controls 116 that allow an operator to scroll through and view the rendering 114.

[0121] Computer memory 110 is shown as containing machine executable instructions. Machine executable instructions 120 enable processor 106 to possibly control other components of medical imaging system 100 and / or also perform various computing and data processing tasks. Memory 110 is also shown as containing three-dimensional medical image data 122. The three-dimensional medical image data may have been acquired by an additional medical scanner, or it may have been received via, for example, network interface 104. Memory 110 is also shown as containing imaging modality 124 that identifies the imaging modality of the three-dimensional medical image data 122. Memory 110 is also shown as containing anatomical view classification 126 that classifies anatomical structures viewed or imaged in the three-dimensional medical image data 122.

[0122] The memory 110 is also shown as including a set of anomaly detection modules 128. The set of anomaly detection modules 128 is a set of convolutional neural networks that have been trained to identify anomalies for different imaging modalities and different anatomical views. The memory 110 is also shown as including a selected anomaly detection module 130, which is selected from the set of anomaly detection modules using the imaging modality 124 and the anatomical view classification 126.

[0123] The memory 110 is also shown as containing a classification of each of a plurality of slices 132 of the three-dimensional medical image data 122 performed using the selected abnormality detection module 130. The classification for each of the slices 132 classifies each of the slices as being normal or abnormal. If any of the slices are identified as being abnormal, then a set of predetermined selection criteria 134 is used to select a set of selected slices 136. The predetermined selection criteria 134 and the set of selected slices 136 are shown as being stored in the memory 110. The slices selected by the predetermined selection criteria 134 can then be used to select slices of the three-dimensional medical image data to display a rendering of the slices 114 on the graphical user interface 112.

[0124] The memory 110 is also shown as containing an optional imaging modality classifier 138 that can be used to provide the imaging modality 124. The memory 110 is also shown as containing an optional anatomy classification module 140 that can be used to provide the anatomical view classification 126.

[0125] Figure 2 The diagram shows the operation Figure 1 A flowchart of a method for a medical instrument of a method for detecting a patient's condition. First, in step 200, three-dimensional medical imaging data including a plurality of slices is received. Next, in step 202, an imaging modality of the three-dimensional medical imaging data is received. The imaging modality may be an identification of the type of medical imaging scanner used to acquire the three-dimensional medical image data. Then, in step 204, an anatomical view classification of the three-dimensional medical image data is received. The anatomical view classification may be an identification of an anatomical structure imaged in the three-dimensional medical image data, or it may be an identification of the type of imaging protocol used during the acquisition of the three-dimensional medical image data. Next, in step 206, the selected abnormality detection module is selected from a set of abnormality detection modules using the imaging modality and the anatomical view classification. Each of the abnormality detection modules is a convolutional neural network trained to classify each of the plurality of slices as normal or abnormal. Then, in step 208, each of the plurality of slices is classified as normal or abnormal using the abnormality detection module. Finally, in step 210, if any of the plurality of slices is classified as abnormal, a set of selected slices is selected from the plurality of slices according to a predetermined selection criterion.

[0126] Figure 3Yet another example of a medical imaging system 300 is shown. The medical imaging system 300 is similar to the medical imaging system 100, except that the medical imaging system 300 additionally includes a magnetic resonance imaging system 302. The magnetic resonance imaging system 302 is intended to be a general example of a medical imaging scanner. The medical imaging system 300 can additionally include more than one additional medical imaging scanner, and can be, for example, one or more of a magnetic resonance imaging system, a positron emission tomography system, a single photon emission tomography system, an ultrasound imaging system, an X-ray system, a computed tomography system, and an endoscopy system.

[0127] Magnetic resonance imaging system 302 includes magnet 304. Magnet 304 is a superconducting cylindrical magnet having a bore 306 passing through it. Different types of magnets may also be used; for example, split cylindrical magnets and so-called open magnets may also be used. Split cylindrical magnets are similar to standard cylindrical magnets, except that the cryostat has been split into two parts to allow access to the isoplane of the magnet, which magnet may be used, for example, in conjunction with charged particle beam therapy. Open magnets have two magnet parts, one above the other, with a space large enough in between to accommodate a subject: the arrangement of the zones of the two parts is similar to the arrangement of Helmholtz coils. Open magnets are popular because the subject is less confined. Inside the cryostat of the cylindrical magnet there is a set of superconducting coils. Within the bore 306 of the cylindrical magnet 304 there is an imaging zone 308, where the magnetic field is strong and uniform enough to perform magnetic resonance imaging. A region of interest 309 within the imaging zone 308 is shown. Magnetic resonance data is typically acquired for a region of interest. Object 318 is shown supported by object support 320 such that at least a portion of object 318 is within imaging zone 308 and region of interest 309 .

[0128] Also within the bore 306 of the magnet is a set of magnetic field gradient coils 310 for acquiring preliminary magnetic resonance data to spatially encode magnetic spins within the imaging zone 308 of the magnet 304. The magnetic field gradient coils 310 are connected to a magnetic field gradient coil power supply 312. The magnetic field gradient coils 310 are intended to be representative. Typically, the magnetic field gradient coils 310 include three separate sets of coils for spatial encoding in three orthogonal spatial directions. The magnetic field gradient power supply provides current to the magnetic field gradient coils. The current provided to the magnetic field gradient coils 310 is controlled as a function of time and can be ramped or pulsed.

[0129] Adjacent to the imaging zone 308 is a radio frequency coil 314, which is used to manipulate the orientation of magnetic spins within the imaging zone 308 and to receive radio frequency transmissions from spins also within the imaging zone 308. The radio frequency antenna may contain multiple coil elements. The radio frequency antenna may also be referred to as a channel or antenna. The radio frequency coil 314 is connected to a radio frequency transceiver 316. The radio frequency coil 314 and the radio frequency transceiver 316 may be replaced by separate transmitting coils and receiving coils and separate transmitters and receivers. It should be understood that the radio frequency coil 314 and the radio frequency transceiver 316 are representative. The radio frequency coil 314 is intended to also represent a dedicated transmitting antenna and a dedicated receiving antenna. Similarly, the transceiver 316 may also represent a separate transmitter and receiver. The radio frequency coil 314 may also have multiple receiving / transmitting elements, and the radio frequency transceiver 316 may have multiple receiving / transmitting channels. For example, if a parallel imaging technique such as SENSE is performed, the radio frequency coil 314 will have multiple coil elements.

[0130] The memory 110 is shown as additionally containing pulse sequence commands 340. The pulse sequence commands 340 are instructions that can be used to control the magnetic resonance imaging system 302 to acquire magnetic resonance imaging data or data that can be converted into such instructions. The memory 110 is also shown as containing magnetic resonance imaging data 342 acquired by controlling the magnetic resonance imaging system 302 using the pulse sequence commands 340. Execution of the machine executable instructions 120 can be used to reconstruct the three-dimensional medical image data 122 from the magnetic resonance imaging data 342.

[0131] The increasing demands on radiologists to process multiple reports and provide their diagnoses within a limited time span has resulted in an overburdened healthcare system. In the present invention, we provide a solution that automatically identifies clinically significant slices and anatomical regions from scans and thereby identifies clinical changes in those scans that have information to make a diagnosis. Additionally, only slices containing clinically significant information are identified and registered with the region of interest, marked for further diagnostic decision making while reducing the time spent on visualization, thus improving throughput.

[0132] According to a recent study published in Academic Radiology, both CT and MRI utilization and the number of images being collected for those examinations have increased significantly in recent years. The authors believe that this could potentially lead to an increase in errors made by overworked radiologists. Robert J. McDonald, MD, and colleagues from the Department of Radiology at the Mayo Clinic in Rochester, Minnesota, conducted a study examining images of the total number of each CT and MRI examination performed at their institution between 1999 and 2010. Over this time span, more than 1.5 million CT and MRI examinations were performed by radiologists. From 1999 to 2010, CT utilization increased by 68%, MRI utilization increased by 85%, and total utilization increased by 75%.

[0133] The authors determined that more than 539 million images were collected from examinations during the study period. CT examinations increased from 82 images per examination in 1999 to 679 images per examination in 2010. MRI examinations increased from 164 images per examination in 1999 to 570 images per examination in 2010.

[0134] However, hiring additional radiologists to help with this increase in utilization and images was not enough. In 1999, the radiologists required to interpret CT scans interpreted 2.8 images per minute. In 2010, the same number was over 19 images per minute. Similarly, in 1999, the radiologists required to interpret MRI scans interpreted 3 images per minute, but in 2010, that number jumped to almost 12 images per minute.

[0135] Modern radiologists now have to interpret more examination images multiple times when compared to similar examinations performed 10-20 years ago. Although these advances in sensitivity and specificity are believed to translate into improved patient care, these increased imaging volumes place an increasing burden on practicing radiologists. As the workload continues to increase, there is a concern that the quality of health care delivered by radiologists will decline in the form of increased detection errors due to increased fatigue and stress. Therefore, we provide a technical solution that is able to view only clinically relevant scans from all scans obtained so that the radiologist is not overburdened and he can spend less time making diagnostic decisions and is also more accurate.

[0136] A paradigm may have one or more of the following advantages:

[0137] 1. Instead of the radiologist having to go through all the scan slices of the object, our solution helps identify only the clinically significant specific slices that need to be shown to the radiologist first instead of the entire scan volume

[0138] 2. Helps reduce the time spent by radiologists analyzing all scan slices

[0139] 3. The system also provides suggestions for other clinical findings (if any) based on previous history and current examinations in addition to the requested examination, thereby reducing scan duplication and improving diagnostic quality

[0140] 4. The system also stitches only slices with clinically significant information into the visualization volume.

[0141] Examples can have one or more of the following characteristics:

[0142] 1. Imaging modality type classification module (imaging modality classifier) ​​for determining the imaging modality

[0143] 2. Determine the anatomical structure classification module for anatomical view classification

[0144] 3. Anomaly detection module that classifies slices as normal or abnormal

[0145] 4. Abnormal slice selection module (predetermined selection criteria)

[0146] 5. Coordinator module (machine executable instructions)

[0147] Figure 4 A high-level functional diagram of yet another example of a medical imaging system 400 is shown. Figure 4 In the present invention, there is a coordinator module 120 that controls the processes and operations of the medical imaging system 400. The coordinator module 120 can be equivalent to machine executable instructions 120. The coordinator module 120 receives an image study volume 122. The image study volume 122 can be processed by an imaging modality classification module 138, an anatomical structure classification module 140, and an anomaly detection module 130. The imaging modality classification module 138 can use an imaging modality classification model 402. The anatomical structure classification module 140 can use an imaging anatomical structure model 404. The anomaly detection module 130 can use an anomaly detection model 406. If there are any abnormalities in the slice, the coordinator module can use the abnormal slice selection module 134. The abnormal slice selection module 134 can be equivalent to the abnormal slice selection module 130. Figure 1 and Figure 3 The example in uses a predetermined selection criterion 134. The output of the abnormal slice selection module 134 is a selected slice sequence 136 that is equivalent to the set of selected slices.

[0148] The imaging modality classification module can be implemented, for example, by detecting the DICOM header or by processing the report using NLP. Moreover, we do consider the information from RIS and HIS to obtain the modality type and the associated study.

[0149] The anatomy classification module can be used to identify slices of relevant anatomy present in the scan. For example, in an MRI brain image, this module will identify the anatomy as the brain. We use the same convolutional neural network as in the previous module, with the only difference being that the last softmax layer has dimensions such as the number of each imaging modality. Figure 5 Depicted CNN architecture.

[0150] Figure 5 An idealized convolutional neural network that can be used to implement an anatomy classification module and / or in some cases an imaging modality classifier 138 is illustrated. There is an input layer 500 for input images and then multiple hidden layers 502. There is then an output layer 504 that provides outputs 506. In the case of the anatomy classification module 140, there is one output 506 for each possible image mode. For example, an output may indicate a specific image of a knee, brain, or other anatomy. In the case of the imaging modality classifier 138, the imaging modality may be indicated by multiple outputs 506. For example, there may be one output indicating a magnetic resonance image and another output indicating another type of image, such as a CT image.

[0151] The anomaly detection module can be used to identify if there are any anomalies in the medical images of the anatomical structure of interest, which is identified using step 2. For example, in an MRI brain study, there may be normal studies and abnormal studies like tumors, hemorrhages, etc. We use a convolutional neural network as in the previous module, where the only difference is that the last softmax layer has two nodes since it is a two-class problem (normal vs. abnormal images). The following Figure 6 Depicts the architecture.

[0152] Figure 6 The general implementation of a convolutional neural network that can be used to implement the selected anomaly detection module 130 is illustrated. The neural network 130 again has an input, and then there are multiple future images 600. This is then fed to hidden unit 602 and hidden unit 604. Hidden unit 604 can be thought of as a feature vector. Feature vector 604 is then used to provide output 606. In this example, one output will indicate that the image is normal, and the other output will be used to indicate whether the image is abnormal.

[0153] The slice selection module may be used to identify clinically significant slices from the entire volume of the image using one or more of the following method steps:

[0154] 1. A two-class GMM hybrid model is trained on features derived from anomaly detection CNN.

[0155] 2. The Euclidean distance between feature vectors of consecutive slices (ED), the intensity variation between slices by mean square error (MSE), the entropy of the slice (ENT) and the likelihood measure (LLM) of the individual slice with respect to normal / abnormal anatomy are calculated. All measures are calculated as follows:

[0156]

[0157] Among them, FV i-1 and FV i are the eigenvectors of the 'i-1'th and 'i'th slices, and 'D' is the dimension of the eigenvector.

[0158]

[0159] Among them, I i-1 and I i are the intensity values ​​of two consecutive frames.

[0160]

[0161] Among them, p o (.) is the probability density function for the intensity value 'O', and 'Ω' is the set of all possible grayscale values ​​in 'I'.

[0162] LLR=log[p(FV|S A )]-log[p(FV|S N )]

[0163] Among them, 'S N '、'S A ' is the Gaussian mixture model of the normal and abnormal cases derived from step 1, and 'F V ' is the eigenvector.

[0164] The possible first set of slices is selected based on the ratio of entropy to MSE. Since MSE is inversely related to image similarity and entropy, the information is directly related to the content of the image. That is,

[0165]

[0166] This set of slices is selected based on a predefined threshold, i.e., SL1>Th1. (b) The second set of slices is selected based on the ratio of LLR to ED. Since ED is inversely related to the similarity of the image and LLR, the information is directly related to the content of the image (normal / abnormal). That is,

[0167]

[0168] The set of slices is selected based on predefined thresholds, ie, SL2>0 and SL2>Th2.

[0169] The normalizations shown above are advantageous because they can prevent pathology and increase stability. Normalization is performed to capture the inherent relationship between ENT and MSE, LLR and ED. The resulting feature vectors provide better distinction between significant and non-significant slices.

[0170] Finally, the user is provided with one or two sets of slices for visualization, because the first set of slices is selected based on information content only, while the second set of slices is selected to be more relevant to the abnormal content.

[0171] The coordinator module can be used for dynamic deployment of training modules and models. The coordinator module can automate the process of selecting volumes from medical studies. In the training phase, the coordinator part takes as input a collection of volumes of various modalities and associated labels about modalities, anatomical structures, and abnormalities. The coordinator component then sets the initial configuration of the various modules during the training process. In the testing phase, the coordinator component will take as input a medical study volume or a collection of medical study volumes, and then apply the appropriate model at each module to generate clinically significant volumes.

[0172] In some examples, implementation may occur in two phases: A) a training phase and B) a deployment phase

[0173] A) Training phase:

[0174] In the training phase, the input to the system is a collection of medical study volumes and associated labels like modality, anatomical structures and abnormalities. These medical study volumes are acquired from various imaging modalities like CT, MR, etc.

[0175] The coordinator component takes as input a collection of medical study volumes and associated labels like modality, anatomical structures, and abnormalities. The imaging volumes are classified into various imaging modality categories using the modality information represented as ground truth labels. Similarly for each modality category, the imaging volumes are further classified into corresponding anatomical structures. Slices of the volumes are further classified into normal and abnormal categories.

[0176] Once all training phases are completed, a hierarchy of classification models is generated and the coordinator component stores these models in the appropriate model database.

[0177] B) Deployment phase:

[0178] In the deployment phase, the goal is to generate clinically relevant slices for the imaging volume. The coordinator component takes the imaging study volume as input and sends the slices to the imaging modality classification module. It then classifies the imaging volume into a modality. The coordinator component then selects the appropriate anatomy classification model based on the predicted modality from the modality classification module and instantiates the anatomy classification module.

[0179] The anatomy classification module predicts the anatomy. The coordinator then instantiates an anomaly module with an appropriate anomaly detection model based on the predicted anatomy and modality from the previous stage for generating abnormal slices. The slice selection module then selects the appropriate slice for the study. Please refer to the following schematic diagram that describes this process for CT modalities and brain anatomy.

[0180] Figure 7 Yet another example 700 of a medical imaging system is illustrated. Figure 7 The depiction in is functional and similar to Figure 4 4. In this example, the medical imaging system 400 receives the imaging study volume 122 again. This is then first passed to the imaging modality classification module 138 which uses the imaging modality classification model 402. In this example, the imaging modality classification module is able to identify a number of different imaging modalities such as CT, MR, ultrasound 706, X-ray 708, pathology 710, and endoscopy 712. In this particular example, the imaging study volume is a CT image, so the imaging study volume is then passed to the CT anatomy classification module 404'. The CT anatomy classification module 404' uses the CT anatomy classification model 402'. In this example, the CT anatomy classification module 404' identifies the imaging study volume 122 as the volume of the brain 714. In this case, there are also multiple other possible examples, which can be the brain 714, the limbs 716, the abdomen 718, or the lungs 720. When it is the brain 714, it is then passed to the specific abnormality detection module 130' for the brain.

[0181] The abnormality detection module 130' uses the CT brain abnormality model 404'. The abnormality detection module 130' identifies one or more slices as abnormal, so they are then passed to the abnormal slice selection module 134 to provide the set of selected slices 136. In this example, they are a stitched sequence of slices. In the stitched sequence of slices, there are abnormal slices that are selected to be displayed to the operator or physician and then the intermediate slices between those slices so that the physician or operator has a complete volume to examine.

[0182] While the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive; the invention is not limited to the disclosed embodiments.

[0183] Those skilled in the art can understand and implement other variations of the disclosed embodiments when practicing the claimed invention by studying the drawings, the specification and the claims. In the claims, the word "comprising" does not exclude other elements or steps, and the words "one" or "an" do not exclude multiple. A single processor or other unit can implement the functions of several items recorded in the claims. Although certain measures are recorded in mutually different dependent claims, this does not indicate that a combination of these measures cannot be used to advantage. The computer program may be stored / distributed on a suitable medium, for example, an optical storage medium or solid-state medium supplied together with other hardware or as part of other hardware, but may also be distributed in other forms, for example, via the Internet or other wired or wireless telecommunications systems. Any reference numerals in the claims should not be interpreted as limiting the scope.

Claims

1. A medical imaging system (100, 300, 400, 700), comprising: a memory (110) for storing machine executable instructions (120); A processor (106) for controlling the medical imaging system, wherein execution of the machine executable instructions causes the processor to: Receiving (200) three-dimensional medical image data (122) including a plurality of slices; Receiving (202) an imaging modality (124) of the three-dimensional medical image data; receiving (204) an anatomical view classification (126) of the three-dimensional medical image data; selecting (206) the selected anomaly detection module (130) from a set of anomaly detection modules (128) using the imaging modality and the anatomical view classification, wherein each of the anomaly detection modules is a convolutional neural network trained to identify at least a portion of the plurality of slices as normal or abnormal; classifying (208) the at least a portion of the plurality of slices as normal or abnormal using the selected anomaly detection module; and selecting (210) a set of selected slices (136) from the plurality of slices according to predetermined selection criteria (134) if a predetermined number of the plurality of slices are classified as abnormal; (1) wherein the execution of the machine executable instructions further causes the processor to: calculating entropy of the at least a portion of the plurality of slices; calculating a mean squared difference intensity variation between adjacent slices for the at least a portion of the plurality of slices; and adding the selected slice of the plurality of slices to the set of selected slices if the ratio of the entropy of the slice to the mean square error intensity variation between adjacent slices is above a predetermined information content threshold; and / or (2) wherein the anomaly detection module is configured to generate a feature vector (604) for at least a portion of the plurality of slices, wherein execution of the machine executable instructions causes the processor to: calculating, for said at least a portion of said plurality of slices, a Euclidean distance between said feature vectors of adjacent slices; calculating a likelihood measure for the at least a portion of the plurality of slices by inputting the feature vectors for the at least a portion of the plurality of slices into a Gaussian mixture model, wherein the Gaussian mixture model is trained to determine probabilities of whether the feature vectors are normal and abnormal; If a ratio of the likelihood measure to the Euclidean distance is greater than a predetermined anomaly measure threshold, the selected slice of the plurality of slices is added to the set of selected slices.

2. The medical imaging system according to claim 1, wherein: Execution of the machine executable instructions further causes the processor to train the Gaussian mixture model using the feature vector (604) derived from the anomaly detection module.

3. The medical imaging system according to any one of claims 1 to 2, wherein: Execution of the machine executable instructions further causes the processor to utilize an imaging modality classifier (138) to determine the imaging modality of the three-dimensional medical image data.

4. The medical imaging system according to claim 3, wherein: The imaging modality classifier is configured to determine the imaging modality using any of: processing a report or log file using natural language processing, extracting the imaging modality from a DICOM header, and receiving the imaging modality from a health information system.

5. The medical imaging system according to any one of claims 1 to 2, wherein: Execution of the machine executable instructions further causes the processor to determine the anatomical view classification using an anatomy classification module (140), wherein the anatomy classification module is a trained convolutional neural network.

6. The medical imaging system according to any one of claims 1 to 2, wherein: The imaging modality is any one of the following: magnetic resonance imaging (704), computed tomography (702), positron emission tomography, ultrasound (706), X-ray (708), and endoscopy (712).

7. The medical imaging system according to any one of claims 1 to 2, wherein: Execution of the machine executable instructions further causes the processor to: receiving a plurality of sets of three-dimensional medical image data, and The set of selected slices is selected for the at least a portion of the plurality of sets of three-dimensional medical image data.

8. The medical imaging system according to any one of claims 1 to 2, wherein: The medical imaging system includes a display (112), and wherein execution of the machine-executable instructions further causes the processor to display the set of selected slices on the display.

9. The medical imaging system according to any one of claims 1 to 2, wherein: The medical imaging system includes a display (112), wherein execution of the machine executable instructions further causes the processor to display a selected portion of the three-dimensional medical image data on the display, wherein the selected portion includes the set of selected slices and slices of the three-dimensional medical image data between the set of selected slices.

10. The medical imaging system according to any one of claims 1 to 2, wherein: The medical imaging system includes a medical imaging scanner (302) configured to acquire the three-dimensional medical image data, wherein execution of the machine-executable instructions further causes the processor to control the medical imaging scanner to acquire the three-dimensional medical image data.

11. The medical imaging system according to claim 10, wherein: The medical imaging scanner is any one of the following: a magnetic resonance imaging system (302), a positron emission tomography system, a single photon emission tomography system, an ultrasound imaging system, an X-ray system, a computed tomography system, and an endoscopy system.

12. A method for operating a medical imaging system (100, 300, 400, 700), wherein: The method comprises: Receiving (200) three-dimensional medical image data (122) including a plurality of slices; Receiving (202) an imaging modality (124) of the three-dimensional medical image data; receiving (204) an anatomical view classification (126) of the three-dimensional medical image data; selecting (206) the selected anomaly detection module (130) from a set of anomaly detection modules (128) using the imaging modality and the anatomical view classification, wherein each of the anomaly detection modules is a convolutional neural network trained to identify at least a portion of the plurality of slices as normal or abnormal; classifying (208) the at least a portion of the plurality of slices as normal or abnormal using the selected anomaly detection module; and selecting (210) a set of selected slices (136) from the plurality of slices according to predetermined selection criteria (134) if a predetermined number of the plurality of slices are classified as abnormal; (1) The method further comprises: calculating entropy of the at least a portion of the plurality of slices; calculating a mean squared difference intensity variation between adjacent slices for the at least a portion of the plurality of slices; and adding the selected slice of the plurality of slices to the set of selected slices if the ratio of the entropy of the slice to the mean square error intensity variation between adjacent slices is above a predetermined information content threshold; and / or (2) wherein the anomaly detection module is configured to generate a feature vector (604) of at least a portion of the plurality of slices, wherein the method further comprises: calculating, for said at least a portion of said plurality of slices, a Euclidean distance between said feature vectors of adjacent slices; calculating a likelihood measure for the at least a portion of the plurality of slices by inputting the feature vectors for the at least a portion of the plurality of slices into a Gaussian mixture model, wherein the Gaussian mixture model is trained to determine probabilities of whether the feature vectors are normal and abnormal; If a ratio of the likelihood measure to the Euclidean distance is greater than a predetermined anomaly measure threshold, the selected slice of the plurality of slices is added to the set of selected slices.

13. A computer program product comprising machine executable instructions for execution by a processor controlling a medical imaging system (100, 300, 400, 700), wherein: Execution of the machine executable instructions causes the processor to: Receiving (200) three-dimensional medical image data (122) including a plurality of slices; Receiving (202) an imaging modality (124) of the three-dimensional medical image data; receiving (204) an anatomical view classification (126) of the three-dimensional medical image data; selecting (206) the selected anomaly detection module (130) from a set of anomaly detection modules (128) using the imaging modality and the anatomical view classification, wherein each of the anomaly detection modules is a convolutional neural network trained to identify at least a portion of the plurality of slices as normal or abnormal; classifying (208) the at least a portion of the plurality of slices as normal or abnormal using the selected anomaly detection module; and selecting (210) a set of selected slices (136) from the plurality of slices according to predetermined selection criteria (134) if a predetermined number of the plurality of slices are classified as abnormal; (1) wherein the execution of the machine executable instructions further causes the processor to: calculating entropy of the at least a portion of the plurality of slices; calculating a mean squared difference intensity variation between adjacent slices for the at least a portion of the plurality of slices; and adding the selected slice of the plurality of slices to the set of selected slices if the ratio of the entropy of the slice to the mean square error intensity variation between adjacent slices is above a predetermined information content threshold; and / or (2) wherein the anomaly detection module is configured to generate a feature vector (604) for at least a portion of the plurality of slices, wherein execution of the machine executable instructions further causes the processor to: calculating, for said at least a portion of said plurality of slices, a Euclidean distance between said feature vectors of adjacent slices; calculating a likelihood measure for the at least a portion of the plurality of slices by inputting the feature vectors for the at least a portion of the plurality of slices into a Gaussian mixture model, wherein the Gaussian mixture model is trained to determine probabilities of whether the feature vectors are normal and abnormal; If a ratio of the likelihood measure to the Euclidean distance is greater than a predetermined anomaly measure threshold, the selected slice of the plurality of slices is added to the set of selected slices.

Images