WO2011137443A1 - Methods for developing breast phantoms - Google Patents

Abstract

The invention relates to developing breast phantoms for digital breast imaging. Specifically, the present invention relates to developing three-dimensional physical breast phantoms by simulating a plurality of breast anatomical parameters comprising a spatial distribution of adipose compartments and Cooper's ligaments.

Description

METHODS FOR DEVELOPING BREAST PHANTOMS

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to United States Provisional Patent Application 61/329,848, filed April 30, 2010, which is incorporated by reference herein in its entirety.

FIELD OF THE INVENTION

[0002] The invention relates to developing breast phantoms for digital breast imaging. Specifically, the invention relates to developing three-dimensional physical breast phantoms by simulating a plurality of breast anatomical parameters.

BACKGROUND OF THE INVENTION

[0003] Breast phantoms have been used as test objects for a number of applications in clinical breast imaging including quality assessment in mammography, optimization of imaging modalities, and evaluation of image post-processing algorithms. Common physical phantoms are made using uniform material (e.g., Lucite) which can mimic X-ray attenuation in breast tissues. These phantoms do not simulate parenchymal patterns, the complex mammographic background generated by projection of breast anatomical structures. Parenchymal patterns, in the form of anatomical noise, may affect the visibility of breast lesions. Some breast phantoms are designed to mimic the appearance of a unique breast clinical image (e.g., an anthropomorphic mammography phantom); however, they cannot simulate the population variation in breast anatomy.

[0004] Computer simulated phantoms provide a flexible alternative to physical models, offering the possibility to generate very complex, subtle anatomical variations. There are two-dimensional (2D) breast models that can be used to generate synthetic images with matching statistical properties of clinical images. However, those models cannot provide realistic three-dimensional (3D) rendering of the underlying parenchymal patterns. Only 3D simulation of the breast anatomy can provide insight into the formation of parenchymal patterns. [0005] There have been several efforts to design 3D software breast phantoms, but they resulted in limitations, mostly relating to sharp, overly geometric, and uniform shape of a breast, thus diminishing realism.

[0006] To date, no realistic three-dimensional (3D) physical anthropomorphic breast phantoms are available. Accordingly, there exists a need to develop improved breast phantoms.

SUMMARY OF THE INVENTION

[0007] In one embodiment, the invention provides a method for developing a 3D physical breast phantom, the method comprising: simulating a breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast; segmenting said simulated breast phantom into a plurality of tissue types; separating said simulated breast phantom into one or more slabs, as needed; fabricating each of said one or more slabs; and assembling the fabricated slabs, thereby developing said 3D physical breast phantom. In an exemplary embodiment, the step of simulation is performed using a region growing algorithm that simulates a plurality of breast anatomical parameters comprising a spatial distribution of adipose compartments and/or Cooper's ligaments.

[0008] In another embodiment, the invention provides a method for developing a software breast phantom, the method comprising: simulating a breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast, thereby developing said software breast phantom.

[0009] In another embodiment, the invention provides a physical breast phantom, wherein said phantom fabricated by the steps of: simulating a breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast; segmenting said simulated breast phantom into a plurality of tissue types; separating said simulated breast phantom into one or more slabs, as needed; fabricating each of said one or more slabs; and assembling the fabricated slabs, thereby fabricating said physical breast phantom. In some embodiments, the physical breast phantom is analyzed to image a breast of a subject. [00010] In another embodiment, the invention provides a method for imaging a breast in a subject, the method comprising developing a breast phantom by simulating said breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast; and analyzing said simulated breast phantom to image said breast in said subject.

[00011] In another embodiment, the invention provides a method for imaging a breast in a subject, the method comprising fabricating a physical breast phantom comprising the steps of: simulating a breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast; segmenting said simulated breast phantom into a plurality of tissue types; separating said simulated breast phantom into one or more slabs, as needed; fabricating each of said one or more slabs; and assembling the fabricated slabs, thereby fabricating said physical breast phantom; and analyzing said physical breast phantom to image said breast in said subject.

[00012] Other features and advantages of the present invention will become apparent from the following detailed description examples and figures. It should be understood, however, that the detailed description and the specific examples while indicating preferred embodiments of the invention are given by way of illustration only, since various changes and modifications within the spirit and scope of the invention will become apparent to those skilled in the art from this detailed description.

BRIEF DESCRIPTION OF THE DRAWINGS

[00013] Figure 1: Flow chart illustrating a method for developing a breast phantom, according to one embodiment of the invention.

[00014] Figure 2: Flow chart of the iterative region growing procedure used for simulation of the adipose compartments and Cooper's ligaments, according to one embodiment of the invention.

[00015] Figure 3: (a) Subgross (thick) histologic slide showing a predominantly FGT region surrounded by a predominantly AT region, (b) A section of the mammary gland. [00016] Figure 4: (a) A breast CT slice of a mastectomy specimen, (b) A clinical mammogram.

[00017] Figure 5: Three orthogonal slices of a phantom that includes only the skin, the large scale structures: the FGT and the AT regions, (a) Coronal view, (b) Sagittal view, (c) Transverse view.

[00018] Figure 6: Illustration of the compartment orientation scheme at the initialization of the region growing procedure, (a) The compartment seeded at point A is oriented so that its shortest axis corresponds to the surface normal of an ellipsoid passing through the nipple and the seed point A. (b). A coronal section of the phantom illustrating distribution of compartments approximately radial to the nipple-chest wall midline, (c) The compartments fan out from the nipple to the chest wall, as seen in this sagittal section through the nipple.

[00019] Figure 7: Pseudo code for the region growing procedure.

[00020] Figure 8: Three orthogonal sections of a software breast phantom: (a) Coronal section; (b) Central sagittal section (corresponding to mammographic MLO view); (c) Transverse section, and (d) a simulated X-ray projection through the phantom.

[00021] Figure 9: Sagittal sections through five realizations of the software breast phantom, corresponding to bra cup sizes (left to right) of A, B, C, D and DD. The number of adipose compartments was 200, 330, 500, 710 and 1050.

[00022] Figure 10: (a) Central sagittal sections of three phantoms generated with PD values of 25% (left), 50% (center), and 75% (right)., (b) Simulated X-ray projections through the phantoms from (a).

[00023] Figure 11: (a) Central sections of three phantom realizations with different ranges of compartment size, (b) Simulated X-ray projections through the phantoms from (a).

[00024] Figure 12: Comparison of histograms for the three phantom realizations with 130, 350 and 700 adipose compartments.

[00025] Figure 13: Central sagittal sections of three phantoms generated with compartment walls thickness of (left to right): 1 voxel, 3 voxels, and mixture of 50% of compartments with 1 voxel and 50% with 3 voxels thick compartment walls, (b) Simulated X-ray projections through the phantoms from (a). [00026] Figure 14: Tomographic section of the segmented software phantom (left) and the corresponding slab for rapid prototyping (right).

[00027] Figure 15: Phantom slabs fabricated by rapid prototyping of the glandular portion, skin and Coopers' ligaments (left) and phantom slabs after filling with an epoxy resin base to simulate adipose tissue (right).

[00028] Figure 16: DM image (left) and a reconstructed DBT slice (right) of the first prototype physical anthropomorphic phantom.

DETAILED DESCRIPTION OF THE INVENTION

[00029] The invention is directed to breast phantoms and methods for developing breast phantoms for digital breast imaging. Specifically, the invention is directed to developing three-dimensional physical breast phantoms by simulating a plurality of breast anatomical parameters.

[00030] In one embodiment, provided herein is a method for developing a breast phantom, the method comprising: simulating a breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast, thereby developing said breast phantom. In another embodiment, provided herein is a method for developing a physical breast phantom, the method comprising: simulating a breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast; segmenting said simulated breast phantom into a plurality of tissue types; separating said simulated breast phantom into one or more slabs, as needed; fabricating each of said one or more slabs; and assembling the fabricated slabs, thereby developing said physical breast phantom.

[00031] In another embodiment, provided herein is a physical breast phantom, wherein said phantom fabricated by the steps of: simulating a breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast; segmenting said simulated breast phantom into a plurality of tissue types; separating said simulated breast phantom into one or more slabs, as needed; fabricating each of said one or more slabs; and assembling the fabricated slabs, thereby fabricating said physical breast phantom.

[00032] Earlier designs of software breast phantoms have several limitations. These limitations are primarily related to the sharp, overly geometric boundary between adipose tissue (AT) and fibroglandular tissue (FGT) regions, and uniform spherical shape of the adipose compartments. These features have been noticeable in simulated x-ray images of the breast phantom, thus diminishing its realism, especially at small spatial scales.

[00033] Inventors of the instant application developed an algorithm for computer simulation of breast anatomy to generate a software phantom. In this phantom, the internal adipose compartments and/or Cooper's ligaments have been simulated using a seeded region-growing algorithm, described herein.

[00034] Figure 1 illustrates a method for developing a breast phantom, according to one embodiment of the invention. As shown in Figure 1, item 10, a breast phantom is simulated based on a plurality of breast anatomical parameters comprising a spatial distribution of adipose compartments and/or Cooper's ligaments. Any one or more of breast anatomical parameters, known to one of skilled in that art, may be used to simulate the breast phantom. An example of breast anatomical parameter includes, but not limited to, a spatial distribution of compartments of a breast such as adipose compartments and Cooper's ligaments. Other examples of breast anatomical parameters include, but are not limited to, breast outline, breast size, breast volume, bra cup-size, shape, glandularity, border between adipose tissue region and fibroglandular tissue region, one or more seed points, breast orientation, growth speed for each adipose compartment, skin, and a combination thereof.

[00035] The simulation may be performed by any algorithm known to one of skilled in the art. In a preferred embodiment, the simulation is performed using a region growing approach, known to one of skilled in the art. For example, the simulation may be performed by a region growing algorithm, as illustrated in Figure 2. In one embodiment, the breast phantom is simulated synthetically based on one or more subjects, for example, human female subjects. The plurality of breast anatomical parameters may be simulated simultaneously or sequentially. In a particular embodiment, shape (e.g., ellipsoid), position, and orientation of adipose compartments are used to provide a realistic appearance of the compartments. [00036] The simulation is a computer-implemented method. In one embodiment, the simulation is performed in a network of computers. In another embodiment, the simulation is performed in a stand-alone computer. In some embodiments, the simulation is performed based on data received from a user. In other embodiments, the simulation is performed based on data stored in a server. In one example, the data reflects breast anatomy related data derived from a subject, for example, a patient. In another example, the data reflects breast anatomy related data derived from a plurality of subjects, for example, human female subjects.

[00037] In one embodiment, a high resolution three dimensional computational model of the breast anatomy is generated with any suitable voxel size. The voxel size may range from about 10 μιη³ to about 500 μιη³. In one embodiment, the voxel size is 10, 50, 100, 150, 200, 250, 300, 350, 400, or 500 μιη . In another embodiment, a high resolution three dimensional dataset is obtained from the breast of an actual patient. This can be obtained from a breast computed tomography (B-CT).

[00038] The three dimensional computational breast model or the patient data set may be pre- processed, for example, various tissue types may be divided into a small number of classes. When using a 3D B-CT data set or when using a non-labeled model, tissue segmentation may be performed. Corrections may need to be performed to condition noise and scatter in B-CT data.

[00039] In some embodiments, a preprocessing step may be applied to the compartment walls (e.g., adipose compartment or Cooper's ligament walls) following the region growing simulation. The preprocessing may include, for example, 3D thinning and/or dilation. The compartment walls may be thinned to one voxel thickness. This step may be followed by a 3D morphological dilation operation, which is applied iteratively until the desired physical thickness of compartment walls is met.

[00040] Prior to building the physical phantom, additional editing may be performed. For example, depending on the printing method, one or more of the tissue types may be manually or automatically edited to ensure continuity of those tissues.

[00041] For ease of fabrication and for access to the various individual parts of the phantom, it may be preferable to construct the phantom in segments. In this instance, a model may be cut into layers. One of skilled in the art may also model specific physical or physiological changes in the breast, such as simulating contrast-agent uptake in parts of the tissue, or simulating lesions.

[00042] As shown in Figure 1, item 20, the simulated breast may be segmented. In one embodiment, simulated breast is segmented into a plurality of tissue types. Each tissue type may correspond to a specific material for fabrication. In another embodiment, simulated breast is segmented into two or more constituent materials. In one example, one tissue portion, for example, a glandular portion is segmented from another tissue portion, for example, an adipose portion of the simulated breast phantom. The adipose portion may include a plurality of adipose compartments and the glandular portion may include a plurality of Cooper's ligaments.

[00043] As shown in Figure 1, item 30, in one embodiment, the simulated breast phantom is separated into one or more slabs. The number of slabs may depend on the size of a breast. For example, a larger sized breast may be separated into more number of slabs, but a smaller sized breast may be separated into relatively less number of slabs. In some embodiments, a slab may be segmented for an object of interest. In some embodiments, the slabs are indexed by providing a unique identifier to each slab. The unique identifier is selected in such way that it would facilitate precisely assembling the slabs. In some embodiments, the slabs are indexed by an indexing method or algorithm, known to one of skilled in the art.

[00044] In one embodiment, the computational phantom data set may be converted to a printable data format, for example, stereolithography (STL) data format to allow rapid prototype printing.

[00045] As shown in Figure 1, item 40, the slabs may be fabricated by a single step or a multi-step process, known to one of skilled in the art. In an exemplary embodiment, each slab is fabricated by a high resolution rapid prototyping. In one embodiment, the simulated breast phantom is fabricated by rapid prototyping in one shot (e.g., a single step) without the need for segmenting it into slabs. The invention provides a method for fabricating a phantom in one shot, for example, by printing both adipose and glandular tissue on a rapid prototyping printer, thus simplifying and accelerating the whole process. Any suitable material may be used for fabrication. In some embodiments, different materials are used to to fabricate different portions of a breast. For example, an adipose portion may be fabricated using one material, but the glandular portion may be fabricated using another material. In one embodiment, one or more portions (e.g., adipose portion) of each slab is filled with an epoxy resin base. In another embodiment, one or more portions of each slab is filled with iodine, for example, to simulate contrast enhanced lesions. In a particular embodiment, a slab is fabricated by filling adipose portions with an epoxy resin base.

[00046] In some embodiments, all regions of a tissue type may be connected. There can be no tissue floating free in space. The remaining tissues (e.g., adipose tissue) in the phantom are created by filling the hollow regions of the rapid prototyping product with epoxy based resins designed to simulate a specific tissue. Alternatively, some rapid prototyping systems can produce objects using multiple materials of different composition simulating the various tissues being simulated. In this instance, the complete phantom may be printed using rapid prototype technology in a single process.

[00047] As shown in Figure 1, item 50, the fabricated slabs are assembled together so as to develop a three-dimensional (3D) physical phantom. In one embodiment, the slabs are aligned in a stack. The unique identifier, discussed herein, may facilitate accurate aligning or assembling of the slabs.

[00048] In one embodiment, the physical phantom is an anthropomorphic phantom. In another embodiment, the physical phantom is a gynomorphic phantom. In one embodiment, one or more devices are implanted inside the physical phantom. Based on a specific need, a suitable device may be implanted. In one embodiment, a dose meter may be implanted inside the physical phantom. In another embodiment, a device that measures radiation may be implanted inside the physical phantom.

[00049] In another embodiment, provided herein is a method for imaging a breast in a subject, the method comprising developing a breast phantom by simulating said breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast; and analyzing said simulated breast phantom to image said breast in said subject.

[00050] In another embodiment, provided herein is a method for imaging a breast in a subject, the method comprising fabricating a physical breast phantom comprising the steps of: simulating a breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast; segmenting a first tissue portion from a second tissue portion of said simulated breast phantom; separating said simulated breast phantom into one or more slabs; fabricating each of said one or more slabs; and assembling the fabricated slabs, thereby fabricating said physical breast phantom; and analyzing said simulated breast phantom to image said breast in said subject.

[00051] Breast imaging may be performed by any suitable imaging method known to one of skilled in the art. In one embodiment, the Imaging is performed by digital mammography (DM). In another embodiment, the Imaging is performed by digital breast tomosynthesis (DBT). In yet another embodiment, the Imaging is performed by dual energy Imaging using slabs. Other imaging methods such as, for example, X-ray mammography, magnetic resonance Imaging (MRI), ultrasound Imaging, and positron emission tomography (PET) Imaging may also be used.

[00052] In another embodiment, provided herein is a method for calibrating a breast image, the method comprising developing a breast phantom by simulating said breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast; and analyzing said simulated breast phantom to calibrate said breast image. In another embodiment, provided herein is a method for calibrating a breast image, the method comprising fabricating a physical breast phantom, as described herein; and analyzing said simulated breast phantom to calibrate said breast image.

[00053] In another embodiment, the invention provides a method for sensitivity analysis of a breast image. For example, the phantom of the invention may be used to remove noises and improve the quality and resolution of measurements. [00054] In some embodiments, a 3D digital data set and a physical phantom are used together to calibrate an Imaging acquisition, reconstruction or image processing method. For example, an x-ray tomosynthesis reconstruction for an Imaging the breast, can be calibrated by acquiring images of the physical phantom by the x-ray tomosynthesis machine used for breast imaging. The acquired images can be compared to a 3D digital data set, representing a ground truth of a phantom. Conventionally, the acquired image of the phantom is compared manually to a 2D picture of the phantom (ground truth). The automated comparison between phantom image and ground truth ensures a better quality calibration. This method is appropriate for all imaging modalities, including but not limited to mammography, tomosynthesis, computed tomography, magnetic resonance imaging, ultrasound, and positron emission tomography.

[00055] The invention allows creating physical phantoms with high anatomic detail. For instance, a 3D breast anthropomorphic phantom of the invention can be produced with at least a 60 μιη voxel resolution. A unique feature of a phantom of the invention is the knowledge of its ground truth in the form of a companion software phantom.

[00056] The term subject, as used herein, refers to any human or non-human object, including, for example, patient, phantom, simulated patient, and simulated phantom.

[00057] All literature references cited in the present specification are hereby incorporated by reference in their entirety.

[00058] The following examples are presented in order to more fully illustrate the preferred embodiments of the invention. They should in no way be construed, however, as limiting the broad scope of the invention.

EXAMPLE 1

Development of an Anthropomorphic Breast Software Phantom Based on Region

Growing Algorithm

[00059] Software breast phantoms offer greater flexibility in generating synthetic breast images compared to physical phantoms. The realism of such generated synthetic images depends on the method for simulating the three-dimensional breast anatomical structures. We developed a novel algorithm for computer simulation of breast anatomy. The algorithm simulates the skin, regions of predominantly adipose tissue and fibro-glandular tissue, and the matrix of adipose tissue compartments and Cooper's ligaments. The simulation approach is based upon a region growing procedure; adipose compartments are grown from a selected set of seed points with different orientation and growth rate. The simulated adipose compartments vary in shape and size similarly to the anatomical breast variation, resulting in much improved phantom realism compared to our previous simulation based on geometric primitives. The proposed simulation also has an improved control over the breast size and glandularity. Our software breast phantom has been used in a number of applications, including breast tomosynthesis and texture analysis optimization.

Materials and Methods

Breast anatomy and its appearance in clinical X-ray images

[00060] The major types of breast tissue, distinguishable both in histological slides and clinical X-ray images are fibro-glandular tissue (FGT), and adipose tissue (AT). Fibro- glandular tissue refers to a combination of the glandular, parenchymal tissue, and the fibrous connective, stromal tissue physically supporting the breast. Adipose tissue is organized into groups of fatty cells, forming macroscopically visible adipose compartments, surrounded by connective tissue. Extensions of the connective tissue that are attached to the skin for breast support are called the Cooper's ligaments. The breast gland is positioned between the superficial and deep layers of the fascia mammae. The superficial fascia layer is separated from the skin by 0.5-2.5 cm thick layer of subcutaneous fat; the deeper fascia layer is separated from the pectoralis major by a layer of retromammary fat.

[00061] Overall, the breast consists of a predominantly FGT region, surrounded by a predominantly AT region, as seen in a subgross histologic slide, Figure 3(a). Histologic slides only show the 2D spatial relationship of various breast tissues; their 3D relationship is illustrated in drawings of the breast anatomical preparations made by Sir A. P. Cooper (Fig. 3(b)). The adipose tissue was removed during the preparation of anatomical samples.

[00062] In clinical X-ray images, breast anatomy is visualized based on X-ray attenuation differences between various tissue types. With the X-ray energy normally used in breast imaging, fibrous and glandular tissue structures have practically the same X-ray attenuation coefficient; the attenuation coefficient for the adipose tissue is about half of the attenuation for fibro-glandular tissue. Fig. 4(a) shows a coronal breast CT of a mastectomy specimen, and Fig. 4(b) shows a clinical mammographic image. Adipose compartments and the Cooper's ligaments are visible in both images; they are more easily distinguishable in the CT image. Projections of the adipose compartments and the connective tissue structures in the breast form the characteristic parenchymal pattern seen in clinical X-ray images of the breast.

[00063] Our breast phantom simulates the skin, large scale regions of AT and FGT, the matrix of Cooper's ligaments and adipose tissue compartments. The version of the phantom presented here does not model the ducts explicitly; we have included a simulated ductal network in our previous version of the phantom. The predominantly AT region has compact coverage of adipose compartments of different size, while the predominantly FGT region has fewer, more separated compartments, as seen in Fig. 4(a).

Simulation of predominantly AT and FGT regions

[00064] We start with the design of our software breast phantom assuming the breast is positioned to a mammographic medio-lateral oblique (MLO) view. The sagittal plane is defined to be parallel with the MLO plane; the phantom outline is symmetric about the center sagittal plane; the current version of the phantom does not include the pectoral muscle. The MLO view is also used in digital breast tomosynthesis. We modeled two large-scale breast tissue regions: the FGT region and the surrounding AT region. The sagittal symmetry plane of the model corresponds to the MLO view plane, as shown in Fig. 5 (a).

[00065] The breast outline is approximated by two ellipsoidal surfaces, attached transversely at the nipple level, as seen in Fig. 5(b). The breast phantom is covered by a thin layer simulating skin. Fig. 5 shows orthogonal cross sections of a simple breast phantom with only large scale structures. Without further simulation of middle scale elements, the FGT region is considered to be purely fibro-glandular while the AT region is purely fatty.

[00066] The size and the shape of the software breast phantom outline, and the size, shape, and position of the FGT region can be interactively specified during the simulation. A customized user interface was developed to allow selection of different modeling parameters and display simulation results.

[00067] Breast density is a known independent factor of breast cancer risk. It is typically measured mammographically as the percent density (PD), the percentage of breast area occupied by the dense, non-fatty tissue (i.e., the skin, the fibrous and glandular tissues). For our 3D software breast phantom, we use a volumetric, 3D percent density, defined as the percentage of the phantom volume occupied by the simulated dense tissue.

Region growing-based algorithm for modeling Cooper's ligaments and adipose compartments

[00068] Breast adipose compartments, as visualized by subgross histology images, mastectomy specimen radiographs, and clinical mammograms, appear to have irregular shape, varying size, and non-uniform distribution throughout the breast. The adipose cavities inside the FGT region are relatively small and scattered among fibro-glandular tissue structures, while the adipose compartments inside the AT region are densely packed, separated by the fibrous walls and Cooper's ligaments, and bounded by the skin and the border between the AT and FGT regions.

[00069] We have used a region growing simulation of adipose compartments and Cooper's ligaments in our software breast phantom. In medical image analysis, region growing algorithms have usually been applied for image segmentation based on region homogeneity. In our project, region growing is used for a realistic, simultaneous simulation of the large number of adipose compartments, providing variability in their shape, size, and distribution. The simulation is initialized by a selection of seed pixels. A region is grown from each seed, assuming an ellipsoidal shape and selected orientation and growth rate. This growth continues until the neighboring compartments touch each other, after which the boundary pixels in contact become inactive. With the continuation of this growing procedure, the initial ellipsoidal compartment shape is gradually transformed into a more realistic, irregular one. Adipose compartments in the AT region are simulated first, followed by the simulation of cavities inside the FGT region; this approach provides more control over the glandularity of the software phantom. Initialization of the region growing procedure

[00070] The number of compartment seeds can be roughly estimated based on clinical images or histological slides. We measured approximate compartment volumes and compared them with the total breast volume. In mastectomy specimen breast CT images (see Fig. 4(a)), we estimated around 200 compartments. The seeds are randomly selected with slightly denser distribution near the nipple, resulting in relatively smaller compartments in the retroareolar area. The growth rate of each compartment is randomly set. Each compartment grows with an initial shape approximated by an ellipsoid. Orientation of ellipsoids is selected randomly with several constraints described in Fig. 6: The compartments are positioned approximately radial to the nipple-chest wall midline, and fan out from the nipple to the chest wall. Such an initial orientation scheme is used to provide a realistic appearance of the compartments at the end of the region growing procedure.

Region growing procedure

[00071] Different region growing rates are implemented with the help of a virtual clock in an iterative procedure. The virtual clock's time unit is determined by the fastest growing compartment. Each compartment is updated after a certain number of time units, depending on the compartment's growth rate. A unique table is created for each compartment to keep track of all its active boundary points in the dynamic growing simulation. A table initially contains only the seed point of the corresponding compartment. During the iterative procedure, a compartment expands ellipsoidally and its associated table is updated iteratively to record the active boundary points and discard points that become internal or inactive in the growing process; the inactive points are stored in a separate table used for post-pocessing. A boundary point becomes inactive once it detects another object in its neighborhood. An inactive boundary point prevents the compartment from expanding at this point. When there are no more active points in a table, the compartment has grown to full contact with its neighboring compartments. Fig. 7 shows the pseudo code for the iterative region growing procedure.

[00072] Two different stopping criteria are used to simulate different distributions of compartments in the AT and FGT regions. In the AT region, all compartments grow until they are in full contact with their neighboring compartments, the skin, or the AT/FGT border. The AT region compartments may grow beyond the AT/FGT region border, up to individually selected distances, to provide for a less geometric appearance of the AT/FGT border in the synthetic images.

[00073] As visualized in clinical images and histologic slides, the adipose compartments in the FGT region are smaller and less compact compared to the AT region compartments. This compartment distribution can be simulated by stopping the region growing simulation while all or most of the compartments are still active. During the FGT region growing procedure, we track the volumetric PD of the software phantom, and stop the growing when a desired PD value is obtained. Post-processing of simulated compartment walls and imaging acquisition model

[00074] The average thickness of the adipose compartment walls, at the end of the region growing procedure, is 3 voxels, In order to have more control over the thickness, we applied a post-processing step to the compartment walls following the region growing simulation. The post-processing includes 3D thinning and dilation. The compartment walls are first thinned to one voxel thickness, based on the analysis of the tables of inactive, points generated during the region growing procedure. This step is followed by a 3D morphological dilation operation, which is applied iteratively until the desired physical thickness of compartment walls is met. Synthetic images generated with different thickness of compartment walls are shown in Figure 13.

[00075] The described region growing procedure was implemented using MATLAB (version 7.3, The MathWorks, Inc., Natick, MA). A customized user interface was developed to allow selection of different modeling parameters and display simulation results. The phantoms shown in this paper were generated with spatial resolution of 200 or 500μιη/νοχ6ΐ. To visualize the effects of different simulation parameters on phantom images, we generated synthetic X-ray projections of the uncompressed phantom assuming a mono- energetic, parallel beam X-ray acquisition model. The linear X-ray attenuation coefficients were selected as μΐ = 0.456 cm^"1 for adipose tissue and μ2 = 0.802 cm^"1 for glandular and connective tissue and skin at 20keV. RESULTS

[00076] We have analyzed the synthetic images of the phantoms with varying breast size, percent density, compartment size, and compartment wall thicknesses. Fig. 8 shows three orthogonal slices and a simulated X -ray projection through a software breast phantom. In the phantom sections, the tissue distribution is visualized using the look-up table of unique gray values assigned to each tissue type.

Breast phantoms of different breast size and percent density

[00077] To cover anatomical variations in breast size, we generated phantoms of different volumes by changing the simulation parameters. Based on the ranges of breast volume corresponding to different bra cup sizes, found in the literature, we simulated five phantoms corresponding to bra cup sizes: A (volume=250ml), B (450ml), C (700ml), D (950ml), and DD (1500ml). Fig. 9 shows the central sagittal sections of the five phantoms with different bra cup sizes. The five phantoms were simulated assuming the same range of the adipose compartments size; as a result, the number of compartments in each phantom varied from 200 to 1050.

[00078] The five phantoms shown in Fig. 9 were generated with similar PD values (PD=38+3%). For a fixed breast size, the tissue PD is primarily determined by the size of the predominantly large scale FGT region. An accurate glandular PD value can be obtained by modifying the stopping criteria of the region growing simulation inside the FGT region. Fig. 10 shows three different phantoms of 450 mL volume with PD values of (left to right): 25%, 50%, and 75%. Breast phantoms of different compartment sizes

[00079] Software phantoms generated with different sizes of the adipose compartments are illustrated in Fig. 11. Three phantom realizations were generated using (left to right): 80, 200, and 400 seeds in the AT region, and 50, 150, and 300 seeds in the FGT region. The corresponding PD values are equal to 31%, 38% and 37%. The average volumes of the adipose compartments in the AT (FGT) region are 3.22 cm³ (2.19 cm³), 1.30 cm³ (0.57 cm³) and 0.66 cm³ (0.34 cm³), for the three phantom realizations (from left to right) in Fig. 9(a). The histograms of compartment volumes for the three phantom realizations are shown in Fig. 12.

Breast phantoms of different thickness of compartment walls

[00080] Fig. 13 shows sections and simulated projections through the phantoms generated with compartment walls thickness of (left to right): 1 voxel, 3 voxels, and a mixture in which 50% of compartments are 1 voxel thick and 50% with 3 voxels thick.

[00081] The proposed breast modeling method is capable of simulating breast phantoms of different breast size, glandularity and distribution of adipose compartments. The phantom size is determined by the volume of breast as parameterized by the ellipsoidal approximation of the breast outline. The phantom PD is determined by the size of the FGT region and the size of FGT adipose compartments. The distribution of adipose compartments is determined by the parameters of the region growing procedure. The number of adipose compartments determines the average compartment size; for a fixed phantom size, the more compartments included, the smaller the average compartment size. The variation in compartment volumes depends on the compartment growth rates; the larger the range of growth rates, the larger the variation in compartment volumes. The final shape of the compartments depends on their initial distribution, orientations and growth rates.

[00082] We have developed a realistic simulation of the breast anatomy and clinical images in the form of a software breast phantom. The software phantom represents an extension of our previous computer breast model. The matrix of adipose compartments and Cooper's ligaments is simulated using a region growing procedure. Two different stopping criteria were implemented to simulate different distribution of adipose compartments in the AT and FGT regions. Using the proposed modeling approach, we were able to simulate a wide range of phantoms with different volume, tissue glandularity, or compartment size distribution.

EXAMPLE 2

Development of a 3D Physical Anthropomorphic Breast Phantom

[00083] Physical phantoms have been extensively used in the development and optimization of new imaging systems. Geometric phantoms are used to evaluate observer- independent image quality metrics such as contrast, noise and spatial resolution. For the more realistic imaging tasks, a more complex phantom that represents the patient anatomy is required. We have developed a 3D physical anthropomorphic breast phantom designed for comparative assessment of digital mammography (DM) and digital breast tomosynthesis (DBT).

Phantom Design

[00084] A prototype physical breast phantom was designed using an existing software breast phantom, representing the breast anatomy in the form of a 3D voxel array. Each voxel belongs to a unique tissue structure and is characterized by the corresponding physical properties. The breast outline is shaped using ellipsoidal approximations. The skin, regions of predominantly adipose (AT) and fibroglandular tissue (FGT), and Cooper's ligaments are simulated (Figure 8). Previous studies on the analysis of parenchymal pattern have suggested a high degree of realism in the simulated tissue structures.

[00085] Mammographic breast compression is an essential part of clinical DM and DBT exams. The breast deformation due to compression was simulated using a finite element model. We assumed uniform elastic properties of the phantom and select compression forces from the clinically observed range. Compressed tissue structures were obtained by interpolation of their undeformed shape to the compressed phantom volume.

[00086] The flexible design of the software phantom provides the ability to cover wide variations in breast size, shape, glandularity, and internal composition. It also allows generation of synthetic images with known ground truth. The software phantom used to construct the physical phantom corresponded to a 450 mL volume, 5cm thick compressed breast with 25% volume glandularity. The software phantom was computed with a 200μιη3 voxel size.

Phantom Fabrication

[00087] Preprocessing of the compressed software breast phantom data was performed to provide an appropriate data format for physical phantom fabrication. To ensure structural stability of the physical phantom we increased the Cooper's ligament thickness using 3 steps of morphological dilation. The glandular tissue portion, skin and Coopers' ligaments were segmented from the adipose tissue (Figure 14, left). The segmented volume was then separated into six slabs and converted to STL data format (Figure 14, right) using Mimics software (Materialise NV, Leuven, Belgium).

[00088] The six slabs were fabricated by high-resolution (60μιη3 voxels) rapid prototyping using the same material for the glandular portion, skin and Coopers' ligaments (Figure 15, left). A material with linear x-ray attenuation equivalent to 70% glandular-30% adipose tissue was used. This material was the most rigid among the materials available for the use with the prototyping printer.

[00089] The adipose regions in such prototyped slabs were filled using an epoxy resin base (Figure 15, right). The material properties of the epoxy were adjusted to closely match the linear attenuation coefficients of adipose breast tissue over the diagnostic energy range used in breast x-ray imaging. The mass density was adjusted by adding phenolic microspheres while the interaction coefficients were adjusted by adding polyvinylidene-fluoride. The epoxy resin was prepared under vacuum. The filled phantom slabs were sanded. The slabs were stacked and held together with interchangeable pins. Phantom Imaging with DM and DBT

[00090] DM and DBT images of the prototype physical phantom were acquired with a clinical DBT machine (Hologic, Inc., Bedford, MA) using automatic exposure control (Figure 16). The tomographic projection images were reconstructed using the filtered backprojection algorithm.

[00091] To our knowledge this is the first 3D physical anthropomorphic breast phantom. A unique feature of this phantom is the knowledge of its ground truth in the form of a companion software phantom. We believe that this makes this phantom an excellent tool for both qualitative and quantitative assessment of the image quality across various DM and DBT systems.

[00092] Having described preferred embodiments of the invention with reference to the accompanying drawings, it is to be understood that the invention is not limited to the precise embodiments, and that various changes and modifications may be effected therein by those skilled in the art without departing from the scope or spirit of the invention as defined in the appended claims.

Claims

WHAT IS CLAIMED IS:

1. A method for fabricating a physical breast phantom, the method comprising:

simulating a breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast;

segmenting said simulated breast phantom into a plurality of tissue types;

separating said simulated breast phantom into one or more slabs;

fabricating each of said one or more slabs; and

assembling the fabricated slabs, thereby fabricating said physical breast phantom.

2. The method of claim 1, wherein said breast phantom is a physical breast phantom.

3. The method of claim 1, wherein said breast phantom is a three-dimensional breast phantom.

4. The method of claim 1, wherein said breast phantom is an anthropomorphic breast phantom.

5. The method of claim 1, wherein said breast phantom is generated synthetically based on one or more subjects.

6. The method of claim 1, wherein the step of simulation is performed using a region growing algorithm.

7. The method of claim 1, wherein at least one of said plurality of parameters is a spatial distribution of adipose compartments or Cooper's ligaments.

8. The method of claim 1, wherein said plurality of parameters further comprises breast outline, breast size, breast volume, bra cup-size, shape, glandularity, border between adipose tissue region and fibroglandular tissue region, one or more seed points, breast orientation, adipose compartments, growth speed for each adipose compartment, skin, Cooper's ligaments, or a combination thereof.

9. The method of claim 1, wherein said first tissue portion is the glandular portion and a second tissue portion is the adipose portion of a breast.

10. The method of claim 1, wherein the step of simulation is performed in a network of computers.

11. The method of claim 1, wherein the step of simulation is performed in a stand-alone computer.

12. The method of claim 1, wherein the step of simulation is performed based on data received from a user.

13. The method of claim 1, wherein the step of simulation is performed based on data stored in a server.

14. The method of claim 12 or 13, wherein the data reflects breast anatomy related data derived from a patient.

15. The method of claim 12 or 13, wherein the data reflects breast anatomy related data derived from a plurality of subjects.

16. The method of claim 1, further comprising the steps of indexing said one or more slabs by providing a unique identifier to each slab.

17. The method of claim 1, wherein each of said one or more slabs is fabricated by a high resolution rapid prototyping.

18. The method of claim 1, wherein each of said one or more slabs is fabricated by filling one or more portions in each slab with an epoxy resin base or iodinated materials.

19. The method of claim 1, further comprising the steps of implanting a device inside said fabricated physical breast phantom.

20. The method of claim 19, wherein said device is a dose meter.

21. A method for developing a breast phantom, the method comprising: simulating a breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast, thereby developing said breast phantom.

22. The method of claim 21, wherein said breast phantom is a physical breast phantom.

23. The method of claim 21, wherein said breast phantom is a three-dimensional breast phantom.

24. The method of claim 21, wherein said breast phantom is a software breast phantom.

25. The method of claim 21, wherein said breast phantom is an anthropomorphic breast phantom.

26. The method of claim 21, wherein said breast phantom is generated synthetically based on one or more subjects.

27. The method of claim 21, wherein the step of simulation is performed using a region growing algorithm.

28. The method of claim 21, wherein at least one of said plurality of parameters is a spatial distribution of adipose compartments or Cooper's ligaments.

29. The method of claim 21, wherein said plurality of parameters further comprises breast outline, breast size, breast volume, bra cup-size, shape, glandularity, border between adipose tissue region and fibroglandular tissue region, one or more seed points, breast orientation, adipose compartments, growth speed for each adipose compartment, skin, Cooper's ligaments, or a combination thereof.

30. The method of claim 21, wherein said first tissue portion is the glandular portion and a second tissue portion is the adipose portion of a breast.

31. The method of claim 21, wherein said method is performed in a network of computers.

32. The method of claim 21, wherein said method is performed in a stand-alone computer.

33. The method of claim 21, wherein the step of simulation is performed based on data received from a user.

34. The method of claim 21, wherein the step of simulation is performed based on data stored in a server.

35. The method of claim 33 or 34, wherein the data reflects breast anatomy related data derived from a patient.

35. The method of claim 33 or 34, wherein the data reflects breast anatomy related data derived from a plurality of subjects.

36. A physical breast phantom, wherein said phantom fabricated by the steps of:

simulating a breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast;

segmenting said simulated breast phantom into a plurality of tissue types;

separating said simulated breast phantom into one or more slabs;

fabricating each of said one or more slabs; and

assembling the fabricated slabs, thereby fabricating said physical breast phantom.

37. A method for imaging a breast in a subject, the method comprising developing a breast phantom by simulating said breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast; and analyzing said simulated breast phantom to image said breast in said subject.

38. A method for imaging a breast in a subject, the method comprising

fabricating a physical breast phantom comprising the steps of:

simulating a breast phantom based on a plurality of breast anatomical parameters, wherein at least one of said plurality of breast anatomical parameters is a spatial distribution of compartments of a breast;

segmenting said simulated breast phantom into a plurality of tissue types;

separating said simulated breast phantom into one or more slabs; fabricating each of said one or more slabs; and assembling the fabricated slabs, thereby fabricating said physical breast phantom; and

analyzing said simulated breast phantom to image said breast in said subject.

39. The method of claim 37 or 38, wherein the Imaging is performed by digital mammography (DM).

40. The method of claim 37 or 38, wherein the Imaging is performed by digital breast tomosynthesis (DBT).

41. The method of claim 37 or 38, wherein the Imaging is performed by dual energy Imaging using slabs.