INTRODUCTION

Top Abstract Introduction Methods Results Discussion References

SEVERAL ASPECTS of organ structure can have profound impact on the function of the organ. Examples relevant to the circulation include 1) the opposing effect of factors increasing the efficacy of a vascular tree in terms of energy needed to fabricate and operate the tree relative to the proportion of the organ occupied by the vascular tree; 2) the influences of the orientation and interconnection of the muscle fibers and the intermuscle fiber stroma on the mechanical efficacy of the heart wall; and 3) the detailed intertwining of the postglomerular blood vessels and the renal tubules on renal excretory function. These examples have in common the question as to how the packing, interconnectivity, perfusion, or drainage of an organ's basic functional unit (BFU; the smallest integrated assembly of diverse cells that functions like an organ, e.g., the nephron and the hepatic lobule, which are generally of the order of 0.1 mm³ in volume) impacts on the interaction (especially structural) of the BFUs within an organ. This information must be known if the functional characteristics of the BFUs [increasingly measured in terms of the function of its component cell(s) by molecular biological methods] are to be translated into the integrated organ function. This interaction is complex due to the sheer number of BFUs within an organ, the fact that each BFU's function can be modulated independently, and because BFUs can impact on each other structurally and functionally.

We believe that three-dimensional (3-D) imaging provides the ability to quantitate the independent BFUs and their physical locations and orientations relative to each other, which are factors that are needed to address the physiological questions posed above. Such a 3-D image will need to resolve individual BFUs and to simultaneously detect and measure all BFUs within an entire organ (or at least a meaningful portion of it, such as a lobe of the organ). This means that, for mice and rats, we need to image a volume of up to 2 cm³ with at least 10 µm, isotropic resolution.

The methodology described here is an attempt at addressing the need for linking our rapidly increasing knowledge about the expression of genes (in cells) to the structure and function of the whole organ (9). Because much of our detailed knowledge of genes and our ability to control gene expression is available in mice and rats, the ability to quantitate the impact of highly targeted genetic manipulations on organ function would appear to be a step in the direction of answering the question as to how genes link to whole body pathophysiology. The ability to effectively detect and quantitate individual BFUs, quantitate their common mechanical linkage, blood supply, and "drainage" trees, as well as quantitate their packing within an intact organ has the potential for contributing to our understanding the "emergent function" of these huge assemblies of interacting BFUs (26).

	METHODS

Top Abstract Introduction Methods Results Discussion References

Microcomputed Tomography System Hardware

View larger version (43K): [in this window] [in a new window]   Fig. 1.   Schematic diagram of microtomography data acquisition camera (28). CCD, charge-coupled device (modified from Ref. 34).

View larger version (134K): [in this window] [in a new window]   Fig. 2.   Photographs of our bench top microcomputed tomography (micro-CT) scanner. A: general view of the micro-CT scanner. A "floating" table isolates the scanner from the vibrations in the floor, and a lead-lined box surrounding the scanner shields personnel from scattered X-rays. On the left is the spectroscopy X-ray source. The collimated X-ray beam is shielded by a lead-wrapped brass tube as the beam passes from the source into the shielded cover of the scanner. A personal computer (right) is used to control the scanner and to store the CCD image data recorded at each angle of view during the scan. B: detailed view of the scanner within the shielded enclosure. See text for details of components.

Step 1. Our scanner uses a Philips spectroscopy X-ray tube (PW2275/20 molybdenum anode, long line focus) with a 12 × 0.4-mm focal spot that is projected at a 4° angle to provide an effective spot size of 0.8 × 0.4 mm. We use an angulated line source rather than a microfocus source because a line source generates the desired X-ray flux at 50 kW/cm², whereas a same size, conventional focal spot would generate the desired flux at 625 kW/cm² (7). The geometry and intensity distribution of the X-ray focal spot are of concern if the X-ray beam geometry is used to achieve magnification, but in our system the long X-ray focal spot-to-specimen distance and the close proximity of the specimen to the X-ray-to-light conversion crystal greatly reduce this concern. The tube is operated at 35 kV peak to minimize spectral contamination from X-ray photons between the short wavelength limit (60 kV peak) of the X-ray generator and the anode's molybdenum K (energy, in kiloelectron volts, of principal X-ray emission line generated by the K shell electrons of an element) line (~17 keV). Energies above 17 keV as well as low energy radiation are suppressed by the 84-µm-thick zirconium foil. Lower energies (e.g., 8 keV) would require exchanging our current molybdenum X-ray source target with a copper anode, and higher energies (e.g., 25 keV) would require use of a silver anode.

Suitable matching of the specimen thickness and the photon energy of the quasi-monochromatic X-rays generated with the X-ray source is achieved if ~5-10% of the radiation incident on the specimen is transmitted (16). At 18 keV, 1.5 cm water attenuate by 85%, and most rat or mouse organs are less than this diameter.

Step 2. The specimen is mounted on a stack of computer-controlled precision stages. At the base of the stack is an inclination stage (Klinger TG 160) that is used to ensure that the rotation axis of the specimen is at right angles to the X-ray beam joining the center of the X-ray focal spot and center of the charge-coupled device (CCD) chip. It has a sensitivity of 1 arcsecond. On top of this stage is a linear translation stage (UT 100.50PP) that has a 0.1-µm resolution and 1.4-µm repeatability. It is used to intermittently remove the specimen from the X-ray beam so as to permit recording of the "raw" beam, which is needed for calibration to compensate for X-ray source instability during the course of a scan. On top of this stage is a rotation stage (RTN 120PP) with 0.01° step size, 0.01° accuracy, and 0.001° resolution. This is arranged so that the axis of rotation of the specimen is always aligned parallel to the central vertical column of pixels of the CCD chip. On top of the rotation stage is a Huber goniometer stage (model 1006, 70 mm), which allows us to place the specimen in a symmetrical (or at desired asymmetrical) position about the axis of rotation.

Step 3. The transmitted X-rays are converted to visible light by a fluorescent scintillating crystal plate (scintillator) of cesium iodide doped with thallium [CsI(Tl)]. This crystal generates ~1,000 light photons per absorbed 18-keV X-ray photon. The 1-mm-thick 4.5 g/cm³ crystal absorbs essentially all 18-keV X-rays. The CsI(Tl) plate is covered with blackened aluminum on the X-ray side to minimize the light reflections at the front surface of the crystal. Although this cuts in half the light that we can detect, it is essential for preventing the loss of spatial resolution that would result from the light internally reflected from the X-ray source side of the crystal. The CCD side of the crystal is covered with a sapphire plate that is coupled to the CsI(Tl) with silicon oil and is transparent to light but provides a barrier to the moisture in ambient air. The light image generated within the CsI(Tl) crystal plate is in turn transferred by a lens onto the surface of a CCD-based imaging camera. This flat field lens currently consists of an f/2.8 enlarger objective for 1- to 10-fold magnification. Thus, if the magnification is twofold, a 1-cm² X-ray image is imaged by a 2-cm² area on the CCD chip, effectively making the CCD pixels 12 µm on a side (referenced to specimen). The use of a coupling lens is much less efficient in converting X-ray scintillations into a signal from the CCD than is a fiber-optic coupling. However, the lens allows us to conveniently alter the magnification of the fluorescent image, a flexibility that we need while we establish the imaging needs for different organs and scientific questions.

Step 4. The continuous illumination pattern generated within the clear CsI(Tl) crystal is converted into a discretized pattern of stored charge in the CCD's (SITe CCD 1024 TKB) pixels. The charge in each of the 1,024 × 1,024, 24 µm on-a-side, square pixels is digitized and stored as an array in a computer until 721 projections, at small increments of specimen rotation around 360°, have been acquired. The actual number of projections needed is governed in large measure by the tomographic resolution desired (8), although the imaged slice thickness cannot be less than the CCD pixel "height." The CCD camera is cooled, both by chilled water and thermoelectrically, to 40°C so as to reduce thermal noise in the signal. The CCD dark current at 40°C is 1 electron · s¹ · pixel¹. The CCD is backside thinned and illuminated, coated with antireflective coating, is ~80% quantum efficient to the 550-nm light generated by CsI(Tl), and can be read out at up to 500,000 pixels per second while maintaining the 16-bit gray scale fidelity.

Step 5. The 16-bit analog-to-digital (A/D) converter has been set up so that each of the A/D conversion steps is equivalent to five electrons, which is approximately equivalent to the readout noise of the CCD. With the CCD's nominal full-well capacity of 450,000 electrons, this means that the medium 16-bit A/D value is reached at ~75% of the nominal full-well condition. This ensures that we utilize the most linear portion of the light input-to-electronic signal output transfer curve and avoid nonlinearity issues in individual CCD pixels, which may fall short of the normal full-well capacity. The CCD chip readout operation is via a controller (ST138 Princeton Instruments CCD camera controller) that is programmed to alter binning, region of interest (ROI), A/D conversion rate, and depth and integration time.

Step 6. The entire camera is suspended from a goniometric cradle (BG 120MS) that has 0.01° readability, 0.001° step size, and 0.05° accuracy. This rotation capability greatly facilitates alignment of the CCD chip array with the axis of rotation of the specimen. The entire imaging "chain" (i.e., scintillator, lens, CCD camera) is attached to a linear translation stage (MT 160) so that the distance of the imaging chain from the specimen rotation axis can be controlled. The position of the scintillator with respect to the lens, along the optical axis, can be independently adjusted (UT-5020PP) with 0.1 µm resolution so as to permit fine focusing adjustment of the fluorescent image onto the CCD chip. The entire imaging chain is suspended from a bridging support stand (Newport/Klinger X95 Structural Rails).

Operation of the Micro-CT Scanner and Image Preprocessing

The processing of the projection data and subsequent CT image reconstruction is currently performed with a Silicon Graphics Origin 2000 computer. For the X-ray CT application, at each angle of view, the system records an array of pixel gray scale values [I(x,y)]. Every 10 or so angles of view the specimen is removed from the X-ray beam and the raw beam is recorded [I₀(,x,y)]. Note that, strictly speaking, I₀ is a function of time, and not of , but, for convenience, we use  as a time reference as our scanning is done sequentially with specimen angle correlated to elapsed time. All of these images first have an averaged "image" of no exposure [I_ABL(x,y)] subtracted, and the final signal(s) is calculated as S(,x,y) = log_e[I₀(,x,y)  I_ABL(x,y)]  log_e[I(,x,y)  I_ABL(x,y)]; I_ABL(x,y) is assumed to change insignificantly during the scan sequence; hence, only one value needs to be used. I₀(,x,y) changes slowly with time (hence with ) so that the average value allows for adequate compensation for the gradual decrease in I₀ over time. The duration of a scan depends on the object diameter, its predominant material content, and resolution desired because these factors impact on the duration of X-ray exposure at each angle and number of angles of view needed. Typically, a scan of a 2-cm-diameter soft tissue specimen, using 360 angles of view, requires 11 h.

Tomographic Image Reconstruction

Backprojection. Because we use an X-ray focal spot that subtends <0.8 mrad at the scintillator, we have reasonably pure cone-beam geometry. In addition, when the scanned volume within the specimen subtends an angle at the X-ray focal spot that is less than the angle subtended by the CCD pixel height divided by the diameter of the scanned volume, we can reasonably assume that the cone-beam geometry is well represented by a stack of contiguous, parallel, transverse fans that ignore the axial divergence of the X-ray beam. This greatly simplifies the backprojection process. In general, however, we use a cone-beam backprojection, modified Feldkamp reconstruction algorithm (12).

Global transaxial image reconstruction. The conventional convolution-based CT reconstruction method (32) requires that all of the line integrals through a single point on a tomographic plane are measured. That implies that the scanner images the entire transaxial extent of the object scanned at every angle of view. An example of a global reconstruction of a single cross section is provided by Fig. 3.

View larger version (132K): [in this window] [in a new window]   Fig. 3.   CT image, generated with global reconstruction, of a single transverse slice through a rat heart. The reconstruction was generated from a whole heart scan. Highlighted square is the region of interested reconstructed at higher resolution as shown in Fig. 4. [From Ritman et al. (34a).]

ROI reconstruction. A region within the object can be imaged selectively by several methods. We have used the following methods (27) for dealing with missing or incomplete sets of scan data.

View larger version (98K): [in this window] [in a new window]   Fig. 4.   Right: global reconstruction of heart cross section, scanned in region of interest only, results in a radially increasing shading artifact. Left: local reconstruction without any assumption about out-of-field-of-view heart. Note sharp detail and no shading artifact in "radial" direction.

	RESULTS

Top Abstract Introduction Methods Results Discussion References

Resolution of the Tomographic Images

1

1

View larger version (113K): [in this window] [in a new window]   Fig. 5.   Volume-rendered display of 500-µm-thick slab through diameter of the cylindrical test phantom used to evaluate 3-D distribution of spatial resolution (see Table 1). Only the image data with gray scale exceeding the automatically determined gray scale threshold, discriminating the microspheres from the background, are displayed. Lower edge coincides with the central plane of the X-ray cone-beam, and the axis of rotation passes vertically through the center of the image. With 6-µm cubic voxels, the 10-µm microspheres show as barely detectable specks of "noise." The 300-, 100-, and 30-µm diameter glass microspheres are clearly visible and their shape discernible.

View larger version (20K): [in this window] [in a new window]   Fig. 6.   Tomographic images of glass microspheres suspended in a plastic cylinder. These images were computed from the stack of transaxial images generated by the scanner so that they are oriented in the axial plane at right angles to the transaxial plane. The X-ray CT number (i.e., image brightness) values along diameters of different diameter microspheres are plotted below each image. Ideally these profiles are square waves, but several features of the CT scanner and reconstruction process blur the edges of objects and thereby round off the profiles.

Because the voxel size is 6 µm in this analysis, dimensional analysis of the images of the 10-µm-diameter microspheres was not pursued. Nondimensional estimates of those small microspheres can, however, be made with the brightness area product method (5) in which case our estimated diameter was 16 µm.

Table 1 summarizes the accuracy of our estimates of microsphere volume and the role of the axial and radial locations of the microspheres in the scanned volume. We expect there to be increased distortion of the microsphere in the axial direction the further the sphere is away from the central plane of symmetry of the X-ray cone-beam. Similarly, the radial distance from the axis of rotation also has an impact on the accuracy because the peripheral zone of the CT image suffers from reduction of the number of transmitted rays passing through each voxel; hence, a "noisier" image is to be expected. These estimates were made from a scan involving 721 equispaced angles of view around 360°. If smaller numbers of views are used (thereby speeding up the scanning process), the estimated volumes are progressively underestimated as the number of angles is decreased. This effect is proportionately greatest for the smallest microspheres.

Examples of 3-D Imaging Capability of the Cardiovascular System

View larger version (58K): [in this window] [in a new window]   Fig. 7.   Display and example of analysis of a three-dimensional (3-D) micro-CT image of a rat heart with contrast agent injected into the coronary arteries. Left: brightest-point projection image of the entire heart. As illustrated in middle, selected artery can be identified and traced for obtaining the branching geometry in terms of branch lengths, diameters, and angle with the mother and daughter branches as a function of vessel segment branching order. Right: example of one form analysis, the testing of the predicted equality of the cubed diameter of the parent vessel (here indicated by cross-sectional area to the 3/2 power) to the sum of the cubes of the daughter vessel diameters.

Figure 8 illustrates the ability to isolate individual vessels from the complex network perfusing a block of porcine heart wall and the ability to quantitatively analyze the vessel dimensions. This analysis is performed with a computer program of the sort developed by Higgins et al. (17).

View larger version (80K): [in this window] [in a new window]   Fig. 8.   A: brightest-point projection image of a transmural slab of a pig's left ventricular wall. B: volume-rendered display of the same data set. C: one isolated transmural vessel isolated by a computer algorithm that connects all voxels of comparable gray scale that are connected. D: plot of vascular cross-sectional area along the length of this "isolated" vessel.

Myocardial muscle fiber "architecture." It is a common observation that almost all tubular muscular organs have helically arranged muscle fibers in the wall. Of particular interest is the muscle fiber architectural arrangement in the myocardium. Figure 9 illustrates the sort of data that can be obtained from the 3-D image data. Examples of hypotheses that "explain" the spiral arrangement of muscle fibers in the myocardium include maximization of a vessel's lumen volume for minimum length of fibers and that the particular arrangement in the left ventricular wall tends to result in equal stress on the muscle fibers despite their different locations and radii of curvature in the wall (3). Our measurements will be available throughout the entire heart wall; hence, the universality of these hypotheses can be tested, which is something that has been well-nigh impossible to date because of the prohibitive logistic task that would be involved.

View larger version (60K): [in this window] [in a new window]   Fig. 9.   Myocardial muscle bundles are discernible in the micro-CT images of a rat heart. Center: image of a transaxial cross section, and "insets" on left are images of sections through the heart wall. Right: angles the muscle fibers make to the transaxial plane. These tomographic images show that the circumferential variation of the fiber angles is clearly visible and can be measured from these images as illustrated in middle and plotted on right. In addition, the physical connection between fiber bundles is also visualized. This capability should permit detailed quantitative assessment of the impact of modulating influences such as hypertrophic stimuli.

Skeletal muscle vasculature. Figure 10 is a display of the 3-D image of a rat hindlimb made with the micro-CT scanner. The blood vessels have been filled with PbO₄ suspended in a silicon polymer (Microfil), and these appear, as well as the bone, in the CT images. Detailed information about the diameters and lengths of the connected vascular segments and the branching angles between the "mother" and "daughter" branches would allow us to perform quantitative evaluation of the distribution of vascular resistances and how this changes in the presence of vascular remodeling. In addition, we could evaluate how this branching geometry might impact on the spatial heterogeneity of perfusion of the muscle.

View larger version (135K): [in this window] [in a new window]   Fig. 10.   Volume-rendered display of vascular tree within a rat's hindlimb. Vessels only around the fibula are shown. Tibia and the vessels in the muscles immediately around it have been "deleted" from this 3-D image data set to reduce the clutter. Relationship of the vessels to the muscle bundles as well as the quantitative indexes of the branching geometry can be used to quantitate the influence of pathophysiological stimuli.

Coronary artery vasa vasorum. Figure 11 is an example of our ability to visualize the 3-D anatomy of vasa vasorum in a section of porcine coronary arterial wall. A question of interest relates to the role of angiogenesis in neointimal formation after coronary artery injury (24). The micro-CT images are uniquely suited not only to quantitate angiogenesis within the arterial wall but also to integrate arterial structure and vascular biology through 3-D image data combined with molecular biological methods using immunohistological methods applied to histological sections of the same specimens after completion of the CT scan. Preliminary analyses of this type (25) show marked spatial disorganization and increased spatial density of coronary vasa vasorum after intracoronary artery balloon dilatation.

View larger version (64K): [in this window] [in a new window]   Fig. 11.   Volume-rendered display of a segment of coronary artery showing the primary lumen and the associated intramural vasa vasorum. Density and source of perfusion of these vasa vasorum may be related to the local formation of the atheromatous plaque. (Image courtesy of Dr. Robert Schwartz, Div. Cardiovascular Diseases, Mayo Clinic.)

Renal vascular system. Figures 12 and 13 illustrate the type of image data that we acquire with micro-CT of a rat kidney. The 3-D microstructure visualized with these images is important for reasons such as in the following hypothetical example. The efferent arterioles of the juxtamedullary nephrons branch into a "horsetail" bundle of capillaries. They are arranged as a parallel set of capillaries of increasing length. If vascular resistance of these parallel branches are equal per unit length then the longer vessels will receive less flow than the shorter. If the pressure head is increased, more flow will occur through the longer vessels with the additional consequence that more of the papillary medulla is perfused with subsequent differences in solute exchange consistent with the extravascular milieu changes down toward the papilla. This is an example where knowledge of the vessel function is insufficient, because the location of the vessel is of critical importance. Now, if the kidney grows in size in response to a uninephrectomy (20), the anatomic proportions may change sufficiently so that the position-related aspect of the function could change. As an example of this methodology, Ortiz et al. (31) have shown that cortical microvasculature is markedly affected in rats subjected to experimental hepatic cirrhosis.

View larger version (90K): [in this window] [in a new window]   Fig. 12.   A: tomographic image of a thin coronal slice through a rat kidney. B: volume-rendered display of the same kidney with gray scale settings selected for the microvasculature. From images such as these, we can readily quantitate the number, spatial distribution, and volume of the glomeruli, features that may reflect various pathophysiological influences on the kidney. (Images courtesy of Dr. Carlos Romero of Dept. of Physiology, Mayo Clinic.)

View larger version (96K): [in this window] [in a new window]   Fig. 13.   Volume-rendered displays of basic functional unit components, i.e., the pre- and postglomerular microvasculature and the glomeruli. (Image courtesy of Dr. Carlos Romero of the Dept. of Physiology, Mayo Clinic, and Dr. Michael D. Bentley, Mankato State University.)

	DISCUSSION

Top Abstract Introduction Methods Results Discussion References

Performance Relative to Biological Measurement Needs

Performance Relative to Other X-Ray Micro 3-D Imaging Methods

In regard to relative costs of the various 3-D microimaging approaches that are available, the cost of the components used in our system was less than $150,000.

X-ray source. X-ray microscopic 3-D imaging depends importantly on the characteristics of the X-ray source. The synchrotron (29) provides a nearly ideal source in that it produces a very intense beam of electromagnetic radiation so that selection of a narrow bandwidth, essentially monochromatic X-ray beam is still sufficiently intense to permit X-ray imaging with adequate photon statistics within a reasonable (biologically speaking) period of time (21). In addition, the mean wavelength of the near monochromatic beam can be tuned to optimally accommodate the thickness of the specimen. Finally, the X-ray beam is very close to parallel, which permits use of a parallel-beam CT image reconstruction algorithm that greatly reduces the computational burden. However, synchrotron X-ray sources are not without limitations. They include the fact that very few beam lines are available and those that exist are available only intermittently for this purpose. Also, sophisticated X-ray optical devices are needed to spread the beam to accommodate specimens greater than ~1 cm diameter if mechanical transverse scanning is to be avoided. Nonetheless, because of these salutory characteristics, synchrotron-based micro-CT scanners will remain important for continued development and evaluation of new techniques applied to micro-CT approaches.

Bench top quasi-monochromatic X-ray sources such as the one we use are available, but other methods are under development at present. These sources all suffer in that the X-ray intensity is limited, resulting in greatly increased duration of imaging. The X-ray generation approach that we have taken appears to be the best compromise at present, although it will be superceded when some of the new techniques are implemented on a routine basis.

X-ray and matter interaction for image contrast generation. Although we use the common method based on X-ray attenuation as the source of image contrast, there are several other uses of X-ray in micro-CT that utilize different aspects of the interaction of X-ray and matter. One such approach is to use phase contrast (i.e., measure different time delays of X-ray propagation through different materials). Although phase contrast has much greater inherent contrast in tissues than does attenuation contrast (4), this method is as yet very much in the early development stages and involves a complex, lengthy scanning and signal analysis process. Another approach is the detection of angle-dependent scatter of the X-ray by the different tissues. This X-ray diffraction tomography has potential for differentiating molecules that are made up of the same atoms, but the molecular arrangement of these atoms differs sufficiently to result in different diffraction "signatures" (15). Another approach uses the emission of fluorescent X-radiation from selected atoms purposely introduced into the specimen (35). If the illuminating X-ray energy is precisely known and of narrow bandwidth, and the fluorescence of the "contrast" atoms is also of known wavelength, the Compton and Thompson scattered X-rays (broad band) can be excluded from the measurement by using a high-purity germanium detector.

Specimen damage by X-ray exposure. The overall limitation of X-ray imaging is the damage caused by the X-rays (28). One limitation is set by the damage of critical proteins that would ultimately kill the specimen. As our current approach uses fixed specimens, the limit is set by the heating caused by the radiation. This limit can be exceeded with synchrotron radiation but not with the spectroscopy source at the focal spot-to-specimen distance we use. We estimate the exposure of our specimens during a complete scan to be ~104 rad. Although this does not noticeably damage our specimens, it could cause problems for in vivo, longitudinal studies. This, in addition to needing to speed up the scan so as to minimize motion blurring, is another reason to reduce the spatial resolution of a scan in longitudinal studies.

Performance Relative to Other, Non-X-Ray Micro-3-D Imaging Methods

Confocal microscopy. Confocal microscopy was developed primarily to reduce the loss of contrast due to the contiguous tissue that is blurred and mixed with the image of the slice of interest within an intact volume of tissue. This approach is used to also scan sequential slices with the advantage that the volume is not physically sliced. Practically speaking, the largest volumes that are imaged with confocal imaging (including 2 photon) are generally <1 mm³ so that continuous measurements are not possible in an intact organ ~1 cm³ in volume.

Scanning electron microscopy. Scanning electron microscopy is limited in that it images a projection of the specimen so that the surfaces only are conveyed, which does not provide the 3-D information we seek.

Micromagnetic resonance imaging. Micromagnetic resonance imaging (micro-MRI) has been successfully used, but the features imaged are quite different from those in X-ray images. Most commonly, MRI images are generated with factors related to the proton mobility. As the resolution of these systems depends directly on the magnetic field strength, special magnets with superconducting coils have been developed to give a microscopic resolution of 10 µm at 7 tesla. Reese et al. (33) have used diffusion-tensor magnetic resonance to image the spatial distribution of myocardial fiber architecture (i.e., angle to the short axis plane). However, that method is limited because it does not provide information about the mechanical interconnection of fibers from plane to plane. Such information is available in the micro-CT 3-D images and could be useful for modeling electrical and mechanical aspects of the myocardium.

Future Development Possibilities