WO2009034499A2

WO2009034499A2 - Flexible 'plug-and-play' medical image segmentation

Info

Publication number: WO2009034499A2
Application number: PCT/IB2008/053554
Authority: WO
Inventors: Jochen Peters; Olivier Ecabert; Hauke Schramm; Juergen Weese
Original assignee: Philips Intellectual Property & Standards Gmbh
Priority date: 2007-09-13
Filing date: 2008-09-03
Publication date: 2009-03-19
Also published as: WO2009034499A3

Abstract

The invention relates to a system (100) for segmentation of image data describing an object comprising a first and second object component, the system comprising: a selection unit (120) for selecting a first component model of a plurality of first component models for adapting to the image data in order to delineate the first object component and for selecting a second component model of a plurality of second component models for adapting to the image data in order to delineate the second object component, a connection unit (130) for connecting the first and second component model, an initialization unit (140) for initializing the first and second component model in the image data volume, and an adaptation unit (150) for adapting the first and second component model to the image data, thereby segmenting the image data. Thus, the system (100) of the invention offers a good coverage of object variability, allowing many object models constructed from component models, while each component model is optimized to describe the modeled object component. Consequently, selecting an optimal first and second component model and an optimal way of connecting them improves the flexibility, robustness, and accuracy of the image data segmentation.

Description

Flexible "plug-and-play" medical image segmentation

FIELD OF THE INVENTION:

The invention relates to the field of image segmentation and more specifically to the field of image segmentation based on deformable models.

BACKGROUND OF THE INVENTION

Deformable models are often successfully used for image segmentation, e.g., for segmentation of medical images. The robustness and accuracy of the segmentation depends on how well a deformable model describes a modeled object. On the other hand, to cope with individual variability of the modeled object, e.g., with different shapes and sizes of an anatomical organ in different patients, the deformable model should be easily deformable. These two requirements are in conflict with each other: a deformable model optimized for describing an object tends to be less flexible in describing individual variability while an easily deformable model is more easily attracted to wrong image features. One way to cope with the problem is to use more models, e.g., a model bank, wherein an object may be described by a plurality of models, each model having different features such as shape, size, and elasticity, typical of a population of objects. For example, a method of constructing a bank of deformable models based on triangular meshes is described in "Automated 3-D PDM Construction From Segmented Images Using Deformable Models", Michael R. Kaus et al, IEEE TRANSACTIONS ON MEDICAL IMAGING, Volume. 22, No. 8, AUGUST 2003, Pages 1005-1013. The use of deformable models based on triangular meshes is described in "Shape-constrained deformable models for 3D medical image segmentation", J. Weese et al., Proc. IPMI, pages 380-387, 2001, hereinafter referred to as Ref. 1. Further, each triangle of the mesh may be assigned a feature function for detecting a feature in the image data to attract said triangle, which feature function is optimal for some objects of a population of objects. Optimizing feature functions of deformable models based on triangular meshes is described in "Feature optimization via simulated search for model-based heart segmentation", Jochen Peters et al., CARS 2005: Computer Assisted Radiology and Surgery, International Congress Series, Volume 1281, May 2005, Pages 33-38. A problem related to this approach is that it may require a large number of deformable models to be created, which is a tedious task.

Alternatively, in a statistical model approach, a few training models describing typical geometries of the modeled objects from a population of objects are used to span a 3N- dimensional deformable model vector space. Each model comprises a point distribution of N points in a 3D space of the image data, the points describing the object surface, for example.

The adapted model is a linear combination of basis vectors of this space. In the adaptation process, the coefficients of this linear combination are optimized. The basis vectors are computed using the principal component analysis of the point distributions of the few training models, as described in "Generalization of point based 3D statistical shape models for anatomical objects", Christian Lorenz and Nils Krahnstόver, Computer Vision and Image

Understanding, Volume 77, 2000, Pages 175-191, for example.

Many further improvements of segmentation techniques using deformable models have been described in the literature. Examples of these improvements include, for example, a progressive adaptation technique described in WO 2006/137013 A2 entitled

"Progressive model-based adaptation". However, despite all the efforts and progress, the need for a further improvement of image data segmentation still exists.

SUMMARY OF THE INVENTION It would be advantageous to have a system for segmenting image data using a deformable model, which would further improve the deformable model capability to cope with individual variations of an object without compromising the robustness and accuracy of the segmentation.

To better address this issue, in an aspect of the invention, a system for segmentation of image data describing an object comprising a first and second object component, is provided, the system comprising: a selection unit for selecting a first component model of a plurality of first component models for adapting to the image data in order to delineate the first object component and for selecting a second component model of a plurality of second component models for adapting to the image data in order to delineate the second object component; a connection unit for connecting the first and second component model; an initialization unit for initializing the first and second component model in the image data volume; and an adaptation unit for adapting the first and second component model to the image data, thereby segmenting the image data.

The component models of the first plurality of component models describe the variability of the first object component. The component models of the second plurality of component models describe the variability of the second object component. The number of compound models for modeling the first and second object component is equal to at least the product of the number of component models of the first and second plurality of component models. Further variations of the compound model are possible by using different ways of connecting the first and second component model. The relative position and orientation of the connected first and second component model is based at least on an interaction between the first and second component model. Thus, the system of the invention offers a good coverage of object variability, allowing many object models constructed from component models, while each component model is optimized to describe the modeled object component. Consequently, selecting an optimal first and second component model and an optimal way of connecting them improves the robustness and accuracy of the image data segmentation.

It is a further advantage of the system of the invention that a large number of compound models for modeling the first and second object component can be obtained with a relatively small number of component models optimized for modeling, respectively, the first and second object component, the total number of compound models being equal to the sum of the number of component models of the first and second plurality of component models.

It is a further advantage of the system of the invention that the system may be arranged to label components of the modeled object. Each component model may comprise a component label and the system may be arranged to display the component label of each adapted component model. It is a further advantage of the system of the invention that different segmentation strategies may be used. For example, the order of initializing, connecting and adapting the component models may be optimally determined by a user of the system.

In an embodiment, the system further comprises a component model bank comprising the plurality of the first and second component models. In an embodiment of the system, the connection unit is arranged to match an interface of the first component model with an interface of the second component model. The matching may be based on a measure of similarity of the first and second component model interface. For example, if the first and second component model comprises a first and second polygonal mesh, respectively, the interface similarity measure may be based on the geometry of the distribution of vertices comprised in the first and second component model interface. Matching interfaces may allow automating the selection of component models. Further, matching interfaces may improve the initialization and/or adaptation of the first and/or second component model. In an embodiment of the system, the connection unit is arranged to match a first subset of vertices of a first polygonal mesh comprised in the first component model with a second subset of vertices of a second polygonal mesh comprised in the second component model. Many implementations of deformable models are based on polygonal meshes. The second subset of vertices of the second polygonal mesh may comprise interface vertices to be merged with or connected to vertices of the first subset of the first polygonal mesh also comprising interface vertices. For example, the connection unit may be arranged to find, for each vertex of the second subset, a vertex of the first polygonal mesh to be merged with or connected to the second subset vertex, such as the nearest vertex of the first polygonal mesh. Thus, the first subset of vertices is determined by the connection unit. The skilled person will know how to implement other useful methods of connecting vertices of the first and second subset.

In an embodiment of the system, the connection unit is arranged to create a virtual component model for connecting the first and second component model. This may be especially useful when the first and second object components are connected with each other via a structure which is not comprised in the first and second component model, like the femur and the tibia connected via the cartilage.

In an embodiment of the system, the first component model is adapted to the image data, using a first adaptation method, and the second component model is adapted to the image data, using a second adaptation method. The first adaptation method may be suitable for the first component model that should be, e.g., flexible, and the second adaptation method may be suitable for the second component model that should be, e.g., rigid.

In a further aspect of the invention, a method of segmentation of image data describing an object comprising a first and second object component, is provided, the method comprising: - a selection step for selecting a first component model of a plurality of first component models for adapting to the image data in order to delineate the first object component and for selecting a second component model of a plurality of second component models for adapting to the image data in order to delineate the second object component; a connection step for connecting the first and second component model; an initialization step for initializing the first and second component model in the image data volume; and an adaptation step for adapting the first and second component model to the image data, thereby segmenting the image data, wherein the first and second component model are interacting with each other.

In an implementation of the method, the selection, connection, initialization, and adaptation steps are carried out in the following order: initializing the first component model; adapting the first component model; - selecting the second component model; connecting the second component model to the first component; initializing the second component model; and adapting the second component model.

This flow of the method is especially useful when the geometry and location of the second object component depends on the geometry and location of the first object component. In this case, the initialization and adaptation of the second component model may be improved by using the results of the adaptation of the first component model. Adapting the second component model may also involve adapting the first component model. Alternatively, the first component model may be frozen and only a part of the second component model may be adapted.

In a further aspect of the invention, a computer program product to be loaded by a computer arrangement is provided, the computer program product comprising instructions for segmentation of image data describing an object comprising a first and second object component, the computer arrangement comprising a processing unit and a memory, the computer program product, after being loaded, providing said processing unit with the capability to carry out the tasks of: selecting a first component model of a plurality of first component models for adapting to the image data in order to delineate the first object component and for selecting a second component model of a plurality of second component models for adapting to the image data in order to delineate the second object component; connecting the first and second component model; initializing the first and second component model in the image data volume; and adapting the first and second component model to the image data, thereby segmenting the image data.

In a further aspect of the invention, the system according to the invention is comprised in an image acquisition apparatus. In a further aspect of the invention, the system according to the invention is comprised in a workstation.

It will be appreciated by those skilled in the art that two or more of the above- mentioned embodiments, implementations, and/or aspects of the invention may be combined in any way deemed useful. Modifications and variations of the image acquisition apparatus, of the workstation, of the method, and/or of the computer program product, which correspond to the described modifications and variations of the system, can be carried out by a skilled person on the basis of the present description.

The skilled person will appreciate that the method may be applied to multidimensional image data, e.g., to 2-dimensional, 3 -dimensional, or 4-dimensional images, acquired by various acquisition modalities such as, but not limited to, standard X-ray Imaging, Computed Tomography (CT), Magnetic Resonance Imaging (MRI), Ultrasound (US), Positron Emission Tomography (PET), Single Photon Emission Computed Tomography (SPECT), and Nuclear Medicine (NM).

BRIEF DESCRIPTION OF THE DRAWINGS

These and other aspects of the invention will become apparent from and will be elucidated with respect to the implementations and embodiments described hereinafter and with reference to the accompanying drawings, wherein: Fig. 1 schematically shows a block diagram of an exemplary embodiment of the system;

Fig. 2 shows an exemplary core model of the heart adapted to the image data and the same core model of the heart plus the aorta adapted to the image data;

Fig. 3 shows flowcharts of two exemplary implementations of the method; Fig. 4 schematically shows an exemplary embodiment of the image acquisition apparatus; and

Fig. 5 schematically shows an exemplary embodiment of the workstation.

Identical reference numerals are used to denote similar parts throughout the Figures. DETAILED DESCRIPTION OF EMBODIMENTS

Fig. 1 schematically shows a block diagram of an exemplary embodiment of the system 100 for segmentation of image data describing an object comprising a first and second object component, the system comprising: a selection unit 120 for selecting a first component model of a plurality of first component models for adapting to the image data in order to delineate the first object component and for selecting a second component model of a plurality of second component models for adapting to the image data in order to delineate the second object component; - a connection unit 130 for connecting the first and second component model; an initialization unit 140 for initializing the first and second component model in the image data volume; and an adaptation unit 150 for adapting the first and second component model to the image data, thereby segmenting the image data. The exemplary embodiment of the system 100 further comprises the following units: a component model bank 110 comprising the plurality of the first and second component models; a control unit 160 for controlling the workflow in the system 100; - a user interface 165 for communicating with a user of the system 100; and a memory unit 170 for storing data.

In an embodiment of the system 100, there are three input connectors 181, 182 and 183 for the incoming data. The first input connector 181 is arranged to receive data coming in from a data storage means such as, but not limited to, a hard disk, a magnetic tape, a flash memory, or an optical disk. The second input connector 182 is arranged to receive data coming in from a user input device such as, but not limited to, a mouse or a touch screen. The third input connector 183 is arranged to receive data coming in from a user input device such as a keyboard. The input connectors 181, 182 and 183 are connected to an input control unit 180. In an embodiment of the system 100, there are two output connectors 191 and

192 for the outgoing data. The first output connector 191 is arranged to output the data to a data storage means such as a hard disk, a magnetic tape, a flash memory, or an optical disk. The second output connector 192 is arranged to output the data to a display device. The output connectors 191 and 192 receive the respective data via an output control unit 190. The skilled person will understand that there are many ways to connect input devices to the input connectors 181, 182 and 183 and the output devices to the output connectors 191 and 192 of the system 100. These ways comprise, but are not limited to, a wired and a wireless connection, a digital network such as, but not limited to, a Local Area Network (LAN) and a Wide Area Network (WAN), the Internet, a digital telephone network, and an analogue telephone network.

In an embodiment of the system 100, the system 100 comprises a memory unit 170. The system 100 is arranged to receive input data from external devices via any of the input connectors 181, 182, and 183 and to store the received input data in the memory unit 170. Loading the input data into the memory unit 170 allows quick access to relevant data portions by the units of the system 100. The input data may comprise, for example, the image data. The memory unit 170 may be implemented by devices such as, but not limited to, a Random Access Memory (RAM) chip, a Read Only Memory (ROM) chip, and/or a hard disk drive and a hard disk. The memory unit 170 may be further arranged to store the output data. The output data may comprise, for example, the first and second component model adapted to the image data. The memory unit 170 may be also arranged to receive data from and/or deliver data to the units of the system 100 comprising the component model bank 110, the selection unit 120, the connection unit 130, the initialization unit 140, the adaptation unit 150, the control unit 160, and the user interface 165, via a memory bus 175. The memory unit 170 is further arranged to make the output data available to external devices via any of the output connectors 191 and 192. Storing data from the units of the system 100 in the memory unit 170 may advantageously improve performance of the units of the system 100 as well as the rate of transfer of the output data from the units of the system 100 to external devices. Alternatively, the system 100 may comprise no memory unit 170 and no memory bus 175. The input data used by the system 100 may be supplied by at least one external device, such as an external memory or a processor, connected to the units of the system 100. Similarly, the output data produced by the system 100 may be supplied to at least one external device, such as an external memory or a processor, connected to the units of the system 100. The units of the system 100 may be arranged to receive the data from each other via internal connections or via a data bus.

In an embodiment of the system 100, the system 100 comprises a control unit 160 for controlling the workflow in the system 100. The control unit may be arranged to receive control data from and provide control data to the units of the system 100. For example, after connecting the second component model to the first component model, the connection unit 130 may be arranged to provide control data "the second component model is connected to the first component model" to the control unit 160 and the control unit 160 may be arranged to provide control data "initialize the second component model" to the initialization unit 140, thereby requesting the initialization unit 140 to initialize the second component model. Alternatively, a control function may be implemented in another unit of the system 100.

In an embodiment of the system 100, the system 100 comprises a user interface 165 for communicating with the user of the system 100. The user interface 165 may be arranged to obtain a user input, e.g., a request for selecting a first and second component model as specified by the user. Optionally, the user interface may receive a user input for selecting a mode of operation of the system, such as the order in which the first and second component model are to be adapted to the image data. The skilled person will understand that more functions may be advantageously implemented in the user interface 165 of the system 100. The essential idea of the invention is to construct a compound model from at least two building blocks, i.e., component models, and to adapt the constructed compound model to image data. One possibility is to construct the whole compound model and to adapt the constructed, whole compound model to image data. For example, in heart segmentation, a heart model may be constructed from atria and ventricles with attached major vessel segments. The pulmonary veins attached to the left atrium, the pulmonary veins and the left atrium hereinafter referred to as the left atrium block, exhibit typical anatomic variation which may be described by a few "standard configurations". Various left atrium block models representing these known typical variants may thus be used. The rest of the heart may be treated as one block, and described by one or more main block models. Different compound models of the human heart may be built by connecting a left atrium block model to a main block model. The compound model comprising the main block model and the left atrium block model may be initialized and adapted to the image data.

Alternatively, the system 100 may be arranged to adapt a partial model, e.g., a main block model, consisting of one or more component models. Later on, one or more new component models, e.g., a left atrium model, that are initialized and adapted, may be connected to the already adapted partial model. In this case, the adaptation may involve adapting both the partial model and the new component models. Alternatively, only the new component models may be adapted while the partial model may be locked in its previously adapted geometry and location. It will be appreciated by those skilled in the art that the process may comprise many iterations. At each iteration, one or more new component models may be added to the partial model until all component models needed for modeling the object are connected and adapted to the image data. An updated partial model comprising the newly added component models may be automatically adapted by the system 100 to the image data. Optionally, a decision to adapt the partial model may be based on a user input. The user interface 165 may be arranged to display the partial model. Optionally, the user may be able to invalidate one or more previous iterations when the displayed partial model is not satisfactory, e.g., because the partial model is poorly adapted to the image data. Fig. 2 shows an exemplary heart core model mesh 21 of the heart adapted to the image data and the same heart core model mesh 21 of the heart plus the aorta model mesh 22 adapted to the image data. The heart core comprises two atria and two ventricles of the heart. The aorta is attached to the heart core at the aortic valve 23 located on the left ventricle. The vertices of the aortic valve define an interface of the heart core model mesh. Rim vertices of the aorta end define an interface of the aorta model mesh. In an embodiment, the heart core model and aorta model interface comprise the same number of vertices. Each vertex of the aorta rim corresponds to one vertex of the aortic valve. The rim vertices are merged with the corresponding valve vertices. The model is then initialized in the image data volume and adapted to the image data. Fig. 3 shows flowcharts of two exemplary implementations of the method of segmentation of image data describing an object comprising a plurality of object components. In a first exemplary implementation 301, the method begins with a selection step 321 for selecting a plurality of component models for adapting to the image data in order to segment the image data. After selecting the plurality of component models, the method continues to a connection step 331 for connecting the plurality of component models. After connecting the plurality of component models, the method continues to an initialization step 341 for initializing the plurality of connected component models in the image data volume. After initializing the plurality of models, the method continues to an adaptation step 351 for adapting the plurality of connected component models to the image data. After adapting the plurality of component models, the method terminates.

In a second exemplary implementation 302, the method begins with the selection step 322 for selecting a partial component model for adapting to the image data in order to segment the image data. After selecting the partial component model, the method continues to initialization step 342 for initializing the partial component model. After initializing the partial component model, the method continues to the adaptation step 352 for adapting the partial component to the image data. After adapting the partial component model, the method continues to the selection step 323 for selecting a next component model. After selecting the next component model, the method continues to the connection step 333 for connecting the next component model to the partial component model. After connecting the next component model to the partial component model, the method continues to the initialization step 343 for initializing the next component model. After initializing the next component model, the method continues to the adaptation step 353 for adapting the next component model to the image data. After adapting the next component model to the image data, the partial component model may be updated in an update step 363. The updated partial component comprises the partial component model and the adapted next component model. After the update step the method returns to the selection step 323 and continues from there. Alternatively, when no more component models are to be added to the partial component model the method terminates. The last updated partial component model becomes the compound model adapted to the image data by the method.

The skilled person may change the order of some steps or perform some steps concurrently using threading models, multi-processor systems or multiple processes without departing from the concept as intended by the present invention. Optionally, two or more steps of the method of the current invention may be combined into one step. Optionally, a step of the method of the current invention may be split into a plurality of steps. Optionally, a step of the method may be omitted.

In an embodiment of the system 100, the selection unit 120 is arranged to select a component model based on a user input. Alternatively, the selection unit 120 may be arranged to evaluate a partial model and select the component model based on this evaluation. For example, the selection unit 120 may be arranged to find an interface of the partial model with a component model that is not connected to the partial model via the found interface. The selection unit 120 may then be arranged to compare an interface of the component model from a component model bank 110 to the found interface of the partial model and select the component model having an interface that best matches the found interface of the partial model.

Further, more advanced techniques to select a component model for connecting to the partial model are possible. For example, the selection unit 120 may be arranged for applying an object detection technique such as the Generalized Hough Transform to a region of interest. A suitable object detection technique is described in "Towards fully automatic object detection and segmentation", Hauke Schramm et al, Proc. SPIE, Volume 6144, 614402, Medical Imaging 2006: Image Processing; Joseph M. Reinhardt, Josien P. Pluim; Eds., pp. 11-20, hereinafter referred to as Ref. 2. The region of interest may be determined based on the position of the initialized or adapted partial model and a free interface of the partial model. A detected object component may be compared to the component models from the component model bank 110 and the selection unit 120 may be arranged to choose the most suitable component model, based on this comparison. For example, each component model may be registered with the detected object component, using a similarity transformation. Alternatively, a number of features of the detected object and the respective features of the component model may be compared to each other and their similarity may be evaluated. The most suitable component model is selected based on this evaluation.

In a further embodiment, a component model is provisionally connected to the partial model and the combined partial and component model are initialized and/or adapted to the image dataset. The adaptation method may be a fast adaptation method, e.g., using a relatively low-resolution subset of the image dataset. The provisionally adapted component model is then evaluated. For example, the selection unit 120 may be arranged to count test features of the component model successfully detected by the provisionally adapted component model. The most suitable component model, e.g., the model with the largest number of detected test features, is selected based on this evaluation.

The selected component model is used by the system for image data segmentation. Image data segmentation means delineating an object comprised in and described by the image data. Delineating an object should be understood as describing and/or representing the object (or an object component) by means of the adapted model (or by means of an adapted component model, respectively).

The connection unit 130 is arranged to connect component models. In an embodiment of the system 100, the component models are constructed in such a way that they comprise matching interfaces which allow them to be connected with each other in a unique way. For example, a triangular mesh of the heart core model 21 of Fig. 2 may comprise interface vertices which match interface vertices of a triangular mesh of the aorta model 22. A list of interface vertices may be comprised in the heart core and aorta model. The connection unit 130 is arranged to connect the aorta model 22 to the heart core model 21 by merging the interface vertices of the aorta model with the corresponding interface vertices of the heart core model. Then the compound model is initialized and adapted to the image data.

The interfaces of the component models may not match each other geometrically. For example, the heart core model 21 may be first initialized in the image data volume and adapted to the image data. Then the geometry of the heart core model 21 interface depends on how the heart core model is adapted to the image data. Typically, the interface of the adapted heart core model 21 will not match the interface of the aorta model 22 from a component model bank. If the interfaces have the same topology, the interface of the aorta model may be transformed to match the interface of the heart core model. For example, when the heart core model and the aorta model are based on polygonal meshes and their interfaces comprise the same number of vertices, then the vertices of the aorta model 22 may be assigned to (i.e., merged with) corresponding vertices of the heart core model 21 mesh interface. This may require bending and stretching or contracting some edges connecting the aorta model mesh 22 interface vertices with the aorta model 22 mesh internal vertices. The aorta model 22 may then be initialized in the image data volume by the initialization unit 140.

Two connected component models interact with each other via their merged interface because the merged interface is a part of and interacts with each of the two connected component models. This interaction, together with internal component model interactions and interactions of the component models with the image features, determine the geometry of the compound model. Additional interactions between component models, e.g., repulsion interactions preventing two component models from connecting with each other outside the interface region, may be also employed by the system 100 of the invention. In an embodiment of the system 100, the connection unit 130 and the initialization unit 140 are combined together. The connection unit 130 and the initialization unit 140 are arranged to place a component model, like an aorta model, in a predetermined position (i.e., location and orientation) relative to an already constructed partial component model, like a heart core model.

In an embodiment of the system 100, the connection unit 130 is arranged to create a virtual component model for connecting the two component models referred to as the first and second component model. For example, using a virtual component model mesh is a way of enabling the system 100 to connect two component model meshes with interfaces of different topology. The virtual component model mesh deformation during adaptation may be governed by the internal forces of the virtual component model mesh. In essence, a virtual component model is an interaction energy interface, which describes how the first and second component model interact with each other. This interaction contributes to the result of the initialization and adaptation of the first and second component model. The definition of the virtual interface may be comprised in the first and/or second component model. The initialization unit 140 is arranged to initialize each component model in the image data volume. In an embodiment, initialization of a component model is based on a user input. The user interface 165 is arranged to allow translating, rotating and scaling the component models. Optionally, the user interface may allow elastic deformations of the component model. In an embodiment of the system 100, the initialization of a component model is automatic. Different approaches may be used for the initialization of component models. These approaches largely depend on the component models, ways of connecting them together, and the flow of the method used by the system 100. For example, in the first exemplary implementation 301 of the method depicted in Fig. 3, the heart core model and the aorta model may be first connected and the compound model comprising the connected component models may be initialized. Alternatively, in the second exemplary implementation 302 of the method depicted in Fig. 3, the heart core model is first initialized and adapted. The initialization of the heart core model may involve object detection and rigid registration of the heart core model with the detected object, for example. Alternatively, affine transformations of the heart core model may be used to register the model with the detected object. A suitable automatic initialization of the human heart model based on a triangular mesh is described in Ref. 2. The initialization of a triangular-mesh-based aorta model connected to a triangular-mesh-based heart core model adapted to the image data may be carried out by minimizing the internal energy of the aorta model mesh as a function of all non- frozen vertices of the aorta model mesh. The frozen vertices of the component model mesh are the interface vertices of the aorta model mesh merged with interface vertices of the heart core model mesh. Next, the initialized heart core model is adapted by the adaptation unit 150.

The adaptation unit 150 is arranged to adapt the initialized component models. The adaptation may involve any number of component models comprised in the compound model. In one embodiment of the adaptation unit 150, the component models, which have been selected, connected to an already existing partial model and initialized during an iteration of the method 302, are adapted to the image data in the adaptation step 353. Alternatively, all component models may be adapted to the image data at each adaptation step 353.

The adaptation method depends on the definition of the model components. A method of adaptation of a triangular mesh to image data is described and discussed in "Modeling shape variability for full heart segmentation in cardiac CT images", Olivier

Ecabert et al., Medical Imaging 2006: Image Processing. Edited by Joseph M. Reinhardt and Josien P. W. Pluim, Proceedings of the SPIE, Volume 6144, 2006, Pages 1199-1210. A method of adaptation of simplex mesh models is described in "General Object Reconstruction based on Simplex Meshes", Herve Delingette, Journal of Computer Vision, Volume 32, September 1999, Pages 111-146. The skilled person will know more models and methods of their adaptation. The scope of the claims must not be construed by means of the methods referred to in the description of the invention.

In an embodiment, different adaptation methods may be used for different component models. Different adaptation methods have different advantages. For example, the adaptation method described in the article by Delingette is suitable for modeling flexible organs such as the left arterial appendage, while other parts of the heart may be adapted using an adaptation method described in the article by Ecabert et al. Optionally, the system 100 of the invention may comprise several adaptation units, each unit using a different adaptation method. The skilled person will understand that initialization and adaptation units or steps may be combined into one unit or step, respectively.

The skilled person will appreciate that the system 100 may be a valuable tool for assisting a physician in many aspects of her/his job.

Those skilled in the art will further understand that other embodiments of the system 100 are also possible. It is possible, among other things, to redefine the units of the system and to redistribute their functions. Although the described embodiments apply to medical images, other applications of the system, outside the medical domain, are also possible.

The units of the system 100 may be implemented using a processor. Normally, their functions are performed under the control of a software program product. During the execution, the software program product is normally loaded into a memory, like a RAM, and executed from there. The program may be loaded from a background memory, such as a ROM, hard disk, or magnetic and/or optical storage, or may be loaded via a network like the Internet. Optionally, an application-specific integrated circuit may provide the described functionality.

Fig. 4 schematically shows an exemplary embodiment of the image acquisition apparatus 400 employing the system 100, said image acquisition apparatus 400 comprising a CT image acquisition unit 410 connected via an internal connection with the system 100, an input connector 401, and an output connector 402. This arrangement advantageously increases the capabilities of the image acquisition apparatus 400, providing said image acquisition apparatus 400 with advantageous capabilities of the system 100.

Fig. 5 schematically shows an exemplary embodiment of the workstation 500. The workstation comprises a system bus 501. A processor 510, a memory 520, a disk input/output (I/O) adapter 530, and a user interface (UI) 540 are operatively connected to the system bus 501. A disk storage device 531 is operatively coupled to the disk I/O adapter 530. A keyboard 541, a mouse 542, and a display 543 are operatively coupled to the UI 540. The system 100 of the invention, implemented as a computer program, is stored in the disk storage device 531. The workstation 500 is arranged to load the program and input data into memory 520 and execute the program on the processor 510. The user can input information to the workstation 500, using the keyboard 541 and/or the mouse 542. The workstation is arranged to output information to the display device 543 and/or to the disk 531. The skilled person will understand that there are numerous other embodiments of the workstation 500 known in the art and that the present embodiment serves the purpose of illustrating the invention and must not be interpreted as limiting the invention to this particular embodiment. It should be noted that the above-mentioned embodiments illustrate rather than limit the invention and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim or in the description. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several distinct elements and by means of a programmed computer. In the system claims enumerating several units, several of these units can be embodied by one and the same item of hardware or software. The usage of the words first, second, third, etc., does not indicate any ordering. These words are to be interpreted as names.

Claims

CLAIMS:

1. A system (100) for segmentation of image data describing an object comprising a first and second object component, the system comprising: a selection unit (120) for selecting a first component model of a plurality of first component models for adapting to the image data in order to delineate the first object component and for selecting a second component model of a plurality of second component models for adapting to the image data in order to delineate the second object component; a connection unit (130) for connecting the first and second component model; an initialization unit (140) for initializing the first and second component model in the image data volume; and - an adaptation unit (150) for adapting the first and second component model to the image data, thereby segmenting the image data.

2. A system (100) as claimed in claim 1, further comprising a component model bank (110) comprising the plurality of the first and second component models.

3. A system (100) as claimed in claim 1, wherein the connection unit (130) is arranged to match an interface of the first component model with an interface of the second component model.

4. A system (100) as claimed in claim 1, wherein the connection unit (130) is arranged to create a virtual component model for connecting the first and second component model.

5. A system (100) as claimed in claim 1, wherein the first component model is adapted to the image data, using a first adaptation method, and the second component model is adapted to the image data, using a second adaptation method.

6. A method (301; 302) of segmentation of image data describing an object comprising a first and second object component, the method comprising: a selection step (321; 322; 323) for selecting a first component model of a plurality of first component models for adapting to the image data in order to delineate the first object component and for selecting a second component model of a plurality of second component models for adapting to the image data in order to delineate the second object component; a connection step (331; 333) for connecting the first and second component model; an initialization step (341; 342; 343) for initializing the first and second component model in the image data volume; and - an adaptation step (351; 352;353) for adapting the first and second component model to the image data, thereby segmenting the image data.

7. A method (301; 302) as claimed in claim 6, wherein the selection, connection, initialization, and adaptation steps are carried out in the following order: - initializing (342; 343) the first component model; adapting (352; 353) the first component model; selecting (323) the second component model; connecting (333) the second component model to the first component; initializing (343) the second component model; and - adapting (353) the second component model.

8. An image acquisition apparatus (400) comprising a system (100) as claimed in claim 1.

9. A workstation (500) comprising a system (100) as claimed in claim 1.

10. A computer program product to be loaded by a computer arrangement, comprising instructions for segmentation of image data describing an object comprising a first and second object component, the computer arrangement comprising a processing unit and a memory, the computer program product, after being loaded, providing said processing unit with the capability to carry out the tasks of: selecting a first component model of a plurality of first component models for adapting to the image data in order to delineate the first object component and for selecting a second component model of a plurality of second component models for adapting to the image data in order to delineate the second object component; connecting the first and second component model; initializing the first and second component model in the image data volume; and - adapting the first and second component model to the image data, thereby segmenting the image data.