阅读说明：本技术 使用图像强度和解剖位置进行组织分类 (Tissue classification using image intensity and anatomical location ) 是由 C·比格尔 S·雷尼施 于 2019-01-09 设计创作，主要内容包括：本发明涉及用于图像分割的医学图像数据处理系统(101)。医学图像数据处理系统(101)包括机器学习框架,所述机器学习框架被训练为接收体素的解剖位置并提供组织类型分类。由医学图像数据处理系统(101)的处理器(130)对机器可执行指令的执行使处理器(130)控制医学图像数据处理系统(101)：-接收包括感兴趣解剖结构的医学图像数据(140),-使用基于模型的分割将解剖参考框架(302、402)拟合到医学图像数据(140),-使用机器学习框架对由医学图像数据(140)的体素表示的组织类型进行分类,其中,体素相对于解剖参考框架(302、402)的解剖位置被用作针对机器学习框架的输入。(The invention relates to a medical image data processing system (101) for image segmentation. A medical image data processing system (101) includes a machine learning framework trained to receive anatomical locations of voxels and provide tissue type classification. Execution of the machine executable instructions by the processor (130) of the medical image data processing system (101) causes the processor (130) to control the medical image data processing system (101): -receiving medical image data (140) comprising an anatomical structure of interest, -fitting an anatomical reference frame (302, 402) to the medical image data (140) using model-based segmentation, -classifying a tissue type represented by voxels of the medical image data (140) using a machine learning frame, wherein an anatomical position of a voxel with respect to the anatomical reference frame (302, 402) is used as an input to the machine learning frame.)

1. A medical image data processing system (101), the medical image data processing system (101) comprising:

a memory (136) storing machine executable instructions and a machine learning framework trained to receive as input an image intensity and an anatomical location of a voxel and in response provide as output a tissue type classification of the voxel,

a processor (130) for controlling the medical image data processing system (101), wherein execution of the machine executable instructions by the processor (130) causes the processor (130) to control the medical image data processing system (101) to:

receiving medical image data (140) comprising an anatomical structure of interest,

-fitting an anatomical frame of reference (302, 402) to the medical image data (140) using a model-based segmentation, wherein a model for the model-based segmentation comprises an anatomical reference structure (300) in a reference space defined by the anatomical frame of reference (302, 402) such that the fitting of the anatomical frame of reference provides an anatomical position relative to the anatomical reference structure,

-classifying tissue types represented by voxels of the medical image data (140) using the machine learning framework, wherein each voxel comprises an image intensity, and wherein an image intensity and the anatomical position of the voxel with respect to the anatomical reference structure (302, 402) are used as input to the machine learning framework.

2. A medical image data processing system (101), the medical image data processing system (101) comprising:

a memory (136) storing machine executable instructions and a machine learning framework trained to receive as input an image intensity and an anatomical location of a voxel and in response provide as output a tissue type classification of the voxel,

a processor (130) for controlling the medical image data processing system (101), wherein execution of the machine executable instructions by the processor (130) causes the processor (130) to control the medical image data processing system (101) to:

receiving medical image data (140) comprising an anatomical structure of interest,

fitting an anatomical frame of reference (302, 402) to the medical image data (140) using a model-based segmentation, wherein a model for the model-based segmentation comprises an anatomical reference structure (300) in a reference space defined by the anatomical frame of reference (302, 402),

-classifying tissue types represented by voxels of the medical image data (140) using the machine learning framework, wherein each voxel comprises an image intensity, and wherein the image intensity and the anatomical position of the voxel with respect to the anatomical frame of reference (302, 402) are used as input to the machine learning framework.

3. The medical image data processing system (101) according to any one of the preceding claims, wherein the medical image data (140) comprises magnetic resonance image data.

4. The medical image data processing system (101) according to claim 3, wherein execution of the machine executable instructions further causes the processor (130) to generate a pseudo CT image (166) using the magnetic resonance image data and a classification result.

5. The medical image data processing system (101) according to any one of the preceding claims, wherein the fitting of the anatomical reference frames (302, 402) comprises deforming the respective frames together with the anatomical reference structure (300) to align the deformed anatomical reference structure (312) with the anatomical structure of interest.

6. The medical image data processing system (101) according to any one of the preceding claims, wherein the model for the model-based segmentation comprises the anatomical reference structure (300) in the form of a surface mesh (400) for the segmentation.

7. The medical image data processing system (101) according to claim 6, wherein the model comprises the anatomical frame of reference (302, 402) in the form of a spatial frame of reference of the surface mesh (400), and wherein the model-based segmentation comprises deforming the anatomical frame of reference together with the surface mesh (400).

8. The medical image data processing system (101) according to any one of the preceding claims, wherein at least two voxels having the same image intensity are assigned to different classes representing different tissue types based on different anatomical positions of the voxels relative to the anatomical frame of reference (302, 402).

9. The medical image data processing system (101) according to claim 8, wherein a first one of the different classes represents bone and a second one of the different classes represents air.

10. The medical image data processing system (101) according to any one of the preceding claims, wherein the receiving of the medical image data (140) comprises: sending a request for respective medical image data (140) to a database (125) comprising the medical image data, wherein the requested medical image data (140) is received from the database (125) in response to the request.

11. The medical image data processing system (101) according to any one of claims 3 to 10, wherein the medical image data processing system (101) further includes a magnetic resonance imaging system (100), and wherein the magnetic resonance imaging system (100) includes:

a main magnet (104) for generating a main magnetic field in an imaging region (108),

a magnetic field gradient system (110) for generating spatially dependent gradient magnetic fields within the imaging zone (108),

a radio frequency antenna system (114) configured for acquiring magnetic resonance data from the imaging zone (108),

wherein the memory (136) further stores pulse sequence commands (141), wherein the pulse sequence commands (141) are configured for controlling the magnetic resonance imaging system (100) to acquire the magnetic resonance data from the imaging zone (108),

wherein the receiving (140) of the medical image data comprises executing the machine executable instructions using pulse sequence commands (141) and acquiring the medical image data (140) in the form of magnetic resonance image data from the imaging zone (108) by the radio frequency antenna system (114).

12. A method for controlling a medical image data processing system (101), the medical image data processing system (101) comprising:

a memory (136) storing machine executable instructions and a machine learning framework trained to receive as input an image intensity and an anatomical location of a voxel and in response provide as output a tissue type classification of the voxel,

a processor (130) for controlling the medical image data processing system (101), wherein execution of the machine executable instructions by the processor (130) causes the processor (130) to control the medical image data processing system (101) to perform a method comprising:

receiving medical image data (140) comprising an anatomical structure of interest,

fitting an anatomical frame of reference (302, 402) to the medical image data (140) using a model-based segmentation, wherein a model for the model-based segmentation comprises an anatomical reference structure (300) in a reference space defined by the anatomical frame of reference (302, 402),

-classifying tissue types represented by voxels of the medical image data (140) using the machine learning framework, wherein each voxel comprises an image intensity, and wherein an anatomical position of the voxel with respect to the anatomical frame of reference (302, 402) is used as an input to the machine learning framework.

13. The method of claim 12, wherein the medical image data processing system (101) further comprises a magnetic resonance imaging system (100), and wherein the magnetic resonance imaging system (100) comprises:

a main magnet (104) for generating a main magnetic field in an imaging region (108),

a magnetic field gradient system (110) for generating spatially dependent gradient magnetic fields within the imaging zone (108),

a radio frequency antenna system (114) configured for acquiring magnetic resonance data from the imaging zone (108),

wherein the memory (136) further stores pulse sequence commands (141), wherein the pulse sequence commands (141) are configured for controlling the magnetic resonance imaging system (100) to acquire the magnetic resonance data from the imaging zone (108),

wherein the receiving (140) of the medical image data comprises executing the machine executable instructions using pulse sequence commands (141) and acquiring the medical image data (140) in the form of magnetic resonance image data from the imaging zone (108) by the radio frequency antenna system (114).

14. A computer program product for controlling a medical image data processing system (101) comprising machine executable instructions for execution by a processor (130) controlling the medical image processing system (101), wherein the medical image data processing system (101) comprises: a memory (136) storing a machine learning framework trained to receive as input an image intensity and an anatomical location of a voxel, and in response provide as output a tissue type classification of the voxel; and a processor (130) for controlling the medical image data processing system (101), wherein execution of the machine executable instructions by the processor (130) causes the processor (130) to control the medical image data processing system (101) to:

receiving medical image data (140) comprising an anatomical structure of interest,

fitting an anatomical frame of reference (302, 402) to the medical image data (140) using a model-based segmentation, wherein a model for the model-based segmentation comprises an anatomical reference structure (300) in a reference space defined by the anatomical frame of reference (302, 402),

-classifying a tissue type represented by voxels of the medical image data (140) using a machine learning framework, wherein each voxel comprises an image intensity, and wherein the image intensity and the anatomical position of the voxel with respect to the anatomical frame of reference (302, 402) are used as input to the machine learning framework.

Technical Field

The present invention relates to processing medical image data, in particular it relates to a method and apparatus for classifying tissue types using medical image data.

Background

For the use of medical image data, image segmentation and classification of voxels comprised by the image data is an important but challenging task in order to identify and analyze anatomical structures of interest.

Not only its boundary contour, but also its internal structure may be important in order to analyze the anatomical structure of interest. It may be particularly important to determine which tissue types are comprised by the respective anatomical structure. Model-based segmentation using shape models has proven to be robust and successful in segmenting organs from medical images with high accuracy. Using model-based segmentation, certain regions within the image are delineated and labeled. However, such model-based segmentation may not be able to determine the internal structural information of the segmented region, since only the boundaries are modeled. In other words, model-based segmentation only provides anatomical locations at organ boundaries, but is not able to provide the anatomical locations on the complete image volume required for 3D tissue classification.

Disclosure of Invention

The invention provides a medical image data processing system, a method of operating a medical image data processing system and a computer program product in the independent claims. Embodiments are given in the dependent claims.

As will be appreciated by one skilled in the art, aspects of the present invention may be embodied as an apparatus, method or computer program product. Accordingly, various aspects of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, micro-code, etc.) or an embodiment combining software and hardware aspects that may all generally be referred to herein as a "circuit," module "or" system. Furthermore, various aspects of the invention may take the form of a computer program product embodied in one or more computer-readable media having computer-executable code embodied thereon.

Any combination of one or more computer-readable media may be utilized. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. "computer-readable storage medium" as used herein encompasses any tangible storage medium that can store instructions executable by a processor of a computing device. The computer-readable storage medium may be referred to as a computer-readable non-transitory storage medium. The computer readable storage medium may also be referred to as a tangible computer readable medium. In some embodiments, the computer-readable storage medium may also be capable of storing data that may be accessed by a processor of a computing device. Examples of computer-readable storage media include, but are not limited to: a floppy disk, a magnetic hard drive, a solid state disk, flash memory, a USB thumb drive, Random Access Memory (RAM), Read Only Memory (ROM), an optical disk, a magneto-optical disk, and a register file for a processor. Examples of optical disks include Compact Disks (CDs) and Digital Versatile Disks (DVDs), e.g., CD-ROMs, CD-RWs, CD-R, DVD-ROMs, DVD-RWs, or DVD-R disks. The term computer readable storage medium also refers to various types of recording media that can be accessed by a computer device via a network or a communication link. For example, the data may be retrieved over a modem, the internet, or 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, wired, fiber optic cable, RF, etc., or any suitable combination of the foregoing.

A computer readable signal medium may include a propagated data signal with computer executable code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. The computer readable signal medium may be any computer readable medium that: is not a computer-readable storage medium and is capable of communicating, propagating or transporting a program for use by or in connection with an instruction execution system, apparatus, or device.

"computer memory" or "memory" is an example of computer-readable storage media. Computer memory is any memory that is directly accessible by a processor. A "computer storage device" or "storage device" is another example of a computer-readable storage medium. The computer storage device is any non-volatile computer-readable storage medium. In some embodiments, the computer storage device may also be computer memory, or vice versa.

"processor," as used herein, encompasses an electronic component capable of executing a program or machine-executable instructions or computer-executable code. References to a computing device comprising "a processor" should be interpreted as being capable of containing more than one processor or processing core. The processor may be, for example, 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 read to be able to refer to a collection or network of computing devices each comprising one or more processors. The computer executable code may be executed by multiple processors, which may be within the same computing device or even distributed among multiple computing devices.

The computer executable code may include machine executable instructions or programs that cause the processor to perform aspects of the present invention. Computer executable code for performing operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. In some instances, the computer executable code may be in a high level language or in a pre-compiled form and used in conjunction with an interpreter that generates machine executable instructions when operated.

The computer executable code may execute 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 the remote computer or server. In the latter scenario, 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 the connection may be made to an external computer (for example, through the Internet using an Internet service provider).

Aspects of the present invention are described with reference to flowchart illustrations, pictorial illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each block or portion of the blocks of the flowcharts, illustrations and/or block diagrams, when applicable, can be implemented by computer program instructions in the form of computer-executable code. It will also be understood that combinations of blocks in different flow diagrams, illustrations, and/or block diagrams, when not mutually exclusive, may be combined. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks.

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 devices to function in a particular manner, such that the instructions stored in the computer-readable medium produce an article of manufacture including instructions which implement the function/act specified in the flowchart and/or block diagram block or blocks.

The computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks.

As used herein, a "user interface" is an interface that allows a user or operator to interact with a computer or computer system. The "user interface" may also be referred to as a "human interface device". The user interface may provide information or data to and/or receive information or data from an operator. The user interface may enable input from an operator to be received by the computer and may provide output from the computer to a user. In other words, the user interface may allow an operator to control or manipulate the computer, and the interface may allow the computer to indicate the effect of the operator's control or manipulation. The display of data or information on a display or graphical user interface is an example of providing information to an operator. The receipt of data by a keyboard, mouse, trackball, touchpad, pointing stick, tablet, joystick, gamepad, webcam, headphones, pedals, wired gloves, remote control, and accelerometer are all examples of user interface components that enable the receipt of information or data from an operator.

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. The hardware interface may allow the processor to send control signals or instructions to an external computing device and/or apparatus. The hardware interface may also enable the processor to exchange data with external computing devices 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 local area network connection, a TCP/IP connection, an Ethernet connection, a control voltage interface, a MIDI interface, an analog input interface, and a digital input interface.

"display" or "display device" as used herein encompasses an output device or user interface suitable for displaying images or data. The display may output visual, audio, 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, vector displays, 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.

Magnetic Resonance Imaging (MRI) data, also referred to as Magnetic Resonance (MR) data or magnetic resonance image data, is defined herein as measurements of radio frequency signals emitted by nuclear spins recorded during a magnetic resonance imaging scan using an antenna of a magnetic resonance apparatus. Magnetic resonance image data is an example of medical image data. A Magnetic Resonance Imaging (MRI) image or MR image is defined herein as a two-dimensional or three-dimensional visualization of the reconstruction of anatomical data comprised by the magnetic resonance imaging data, i.e. the MRI image is provided by an MRI dataset comprising representative selected MRI data. The visualization may be performed using a computer. The magnetic resonance imaging data may be provided using a representation of the corresponding data in k-space or image space. Using fourier transformation, magnetic resonance imaging data can be transformed from k-space to image space, and vice versa. In the following, the magnetic resonance image data may comprise a selection of MRI data in an image space representing a two-dimensional or three-dimensional anatomical structure, i.e. an MRI image.

An "anatomical structure" is any anatomical structure of an object, such as e.g. a human or an animal. The structure may comprise or be part of an organ, such as the liver or the brain, or it may comprise an anatomical region, such as the spinal region, the knee, the shoulder, etc.

"segmentation" refers to the subdivision of digital image data into a plurality of segments, e.g., sets of pixels or voxels, to identify one or more segments that represent a structure of interest that may be further analyzed. Segmentation may be used to identify and locate structures, particularly their boundaries in the image data. For example, lines, curves, planes, surfaces, etc. may be identified. A label may be assigned to each pixel or voxel in the image such that pixels/voxels with the same label share certain characteristics and may be highlighted to indicate a contour, region, or volume containing the structure of interest. When applied to three-dimensional image data, such as a stack of two-dimensional images, the resulting contours after image segmentation may be used to create a three-dimensional reconstruction of the shape of a structure of interest, such as an anatomical structure, by means of an interpolation algorithm, such as for example a marching cube. The three-dimensional profile may be provided by a surface mesh, for example. The mesh may comprise a plurality of flat two-dimensional polygons, such as triangles.

"model-based segmentation" refers to segmentation using a shape model, for example in the form of a surface mesh, which is deformed to align with and identify contours of the boundaries of the anatomical structure of interest.

The term "machine learning" (ML) refers to a computer algorithm for extracting useful information from a training data set by building a probabilistic framework, referred to as a machine learning framework (also referred to as a machine learning model), in an automated fashion. Machine learning may be performed using one or more learning algorithms, such as linear regression, K-means, classification algorithms, reinforcement algorithms, and the like. The "machine learning framework" may be, for example, an equation or a set of rules that enables prediction of unmeasured values (e.g., which label corresponds to a given token) from other known values and/or prediction or selection of actions to maximize future rewards. According to one embodiment, the machine learning framework is a deep learning framework.

For example, the learning algorithm may include a classification and/or reinforcement algorithm. The reinforcement algorithm may, for example, be configured to learn one or more policies or rules for determining a next set of parameters (action) based on the current set of parameters and/or a previously used set of parameters. For example, starting from a current set of parameters and/or a previous set of acquisition parameters, the machine learning framework may follow the policy until a desired set of acquisition parameters is reached. The policy represents the decision making process of the model at each step, e.g. it defines which parameter to change and how to change it, it adds a new parameter to the parameter set, etc. The selection of actions may be optimized by learning based on known landmarks marked on the input image.

"image intensity" refers to the signal intensity comprised by an individual voxel of the medical image data.

An "anatomical frame of reference" refers to a frame of reference, such as a coordinate system, that defines a position relative to an anatomical reference structure. The anatomical reference frame defines a reference space in which the anatomical reference structure is located. When the spatial form of the anatomical reference structure is deformed, the anatomical reference frame is deformed accordingly.

In one aspect, the invention relates to a medical image data processing system. The medical image data processing system includes a memory storing machine executable instructions and a machine learning model. The machine learning model is trained to receive as input the image intensity and anatomical location of the voxel and in response provide as output a tissue type classification of the voxel. The medical image data processing system further comprises a processor for controlling the medical image data processing system. Execution of the machine executable instructions by the processor causes the processor to control a medical image data processing system to receive medical image data comprising an anatomical structure of interest. An anatomical reference frame is fitted to the medical image data using a model-based segmentation, wherein the model for the model-based segmentation comprises an anatomical reference structure in a reference space defined by the anatomical reference frame. A machine learning framework is used to classify tissue types represented by voxels of medical image data. Each voxel comprises an image intensity. The image intensity of the voxel and the anatomical position of the voxel relative to an anatomical frame of reference are used as inputs to a machine learning framework.

Automatic image segmentation may be desirable in various medical imaging applications. For example, model-based segmentation using shape models may provide a robust and accurate method to segment an anatomical structure of interest. However, since only the boundaries are modeled, this approach may not be able to determine the internal structure contained by the segmented region. Thus, when highly accurate voxel-by-voxel volume classification is desired, model-based segmentation may fail. On the other hand, ML methods that attempt to classify each image voxel may have the ability to deliver volumetric classification results with relatively little training effort. In other words, the ML method may be able to take into account internal structures that are ignored by the model-based segmentation. However, the ML method generally relies on image features that pass through the voxels to be classified. Such internal structures may include different tissue types showing similar or even the same image characteristics (such as, for example, intensity). ML methods based on image features of the structure to be analyzed as input may fail when different anatomical structures show similar image features. Thus, such anatomical structures cannot be distinguished into separate classification categories. Known ML methods may lack the regularization properties of the shape model and may require significant post-processing.

Drawings

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

figure 1 illustrates an example of a medical image data processing system,

figure 2 illustrates an example of a magnetic resonance imaging system;

FIG. 3 illustrates an example of a method of operating a medical image data processing system;

figure 4 illustrates an example of a method of operating a magnetic resonance imaging system;

FIG. 5 illustrates an example of generating a pseudo CT image;

FIG. 6 is an example of a fit;

FIG. 7 is an example of a fit; and is

Fig. 8 illustrates an example of generating a pseudo CT image.

List of reference numerals

100 magnetic resonance imaging system

101 medical image data processing system

104 main magnet

106 magnet bore

108 imaging zone

110 magnetic field gradient coil

112 magnetic field gradient coil power supply

114 radio frequency coil

115 transceiver

118 object

120 object support

122 actuator

125 database

126 computer

128 hardware interface

130 processor

132 user interface

134 computer storage device

136 computer memory

140 medical image data

141 pulse sequence commands

142 control module

143 control module

144 segmentation module

146 fitting module

148 machine learning module

150 transformation module

160 segmented results

162 fitting results

164 results of machine learning

166 pseudo CT image

300 anatomical reference structure

302 anatomical frame of reference

310 deformed anatomical reference structure

312 deformed anatomical frame of reference

400 surface mesh

402 anatomical frame of reference

412 deformed anatomical frame of reference

500 MRI images

502 MRI image

504 MRI images

510 pseudo CT image

512 pseudo CT image

514 pseudo CT image

Embodiments propose to additionally consider the spatial relationship of the underlying anatomy in order to be able to distinguish tissue types showing similar or even identical image features. Thus, the common model-based segmentation method is extended by Spatial Encoding (SE), which provides a volumetric coordinate system with respect to the anatomical frame of reference, i.e. spatial information for each voxel.

According to embodiments, a combination of ML and SE may be used to overcome two disadvantages of the methodology. The shape model may be applied to implement an anatomical frame of reference of the medical image, i.e. a spatial frame of reference providing spatial coordinates relative to an anatomical reference structure represented by the shape model. The anatomical position relative to the anatomical frame of reference serves as an additional input to the ML in addition to image features such as image intensity. Embodiments may allow for accurate classification of target structures on a voxel-by-voxel basis while effectively distinguishing even different types of tissue with similar image features. In case of similar and/or identical identities, the anatomical locations may be used for differentiation. Thus, even internal structures may be applied to the segmented image area.

According to an embodiment, only these voxels may be classified by ML using their anatomical locations for which it is known that the image features provided by them are ambiguous with respect to the tissue type represented. Considering, for example, an MR image, voxels representing air have similar image characteristics as voxels representing bone. Both are dark, i.e. comprise low intensity.

According to an embodiment, only such voxels may be classified by ML using their anatomical location, which comprises an image intensity above a predefined threshold, below a predefined threshold and/or within a range defined by a lower threshold and an upper threshold. According to an embodiment, all voxels are classified by ML using their anatomical location except for image features like image intensity.