Small Animal Computed Tomography Imaging - Ingenta Connect

Current Medical Imaging Reviews, 2007, 3, 45-59

45

Small Animal Computed Tomography Imaging Soenke H. Bartling1,2,*, Wolfram Stiller2,*, Wolfhard Semmler2 and Fabian Kiessling1,2 1

Junior Group Molecular Imaging, 2Department of Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany Abstract: Micro Computed Tomography (micro-CT) was suggested in biomedical research to investigate tissues and small animals. Its use to characterize bone structures, vessels (e.g. tumor vascularization), tumors and soft tissues such as lung parenchyma has been shown. When co-registered, micro-CT can add structural information to other small animal imaging modalities. However, due to fundamental CT principles, high-resolution imaging with micro-CT demands for high x-ray doses and long scan times to generate a sufficiently high signal-to-noise ratio. Long scan times in turn make the use of extravascular contrast agents difficult. Recently introduced flat-panel based mini-CT systems offer a valuable tradeoff between resolution (~200 µm), scan time (0.5 s), applied x-ray dose and scan field-of-view. This allows for angiography scans and follow-up examinations using iodinated contrast agents having a similar performance compared to patient scans. Furthermore, dynamic examinations such as perfusion studies as well as retrospective motion gating are currently implemented using flat-panel CT. This review summarizes applications of experimental CT in basic research and provides an overview of current hardware developments making CT a powerful tool to study tissue morphology and function in small laboratory animals such as rodents.

Keywords: Small animal imaging, Computed Tomography (CT), micro-CT, mini-CT, flat-panel detector, motion gating. INTRODUCTION Micro- and mini-CT are scaled down CT-imaging modalities for small animals, which in principle provide the same information about morphology and disease status or disease progression for animals as clinical-scale CT does for humans. Nonetheless, several major differences compared to clinical CT scanning exist. This review provides an overview of basic concepts and techniques that apply to CT when used for in-vivo small animal imaging. Technical and physical limitations of these concepts as well as applications of micro- and mini-CT and an outlook on future developments in the growing field of research in small animal CT imaging are given and discussed. Definition There is no unique definition of mini-CT and micro-CT. Often, all CT scanners that provide higher spatial resolution than current clinical scanners are named micro-CT. This definition results in a very broad range of different CT scanner design concepts ranging from CT scanners designed and used for non-destructive material-testing with a resolution in the order of a few µm up to dedicated small animal CT scanners with a resolution of a few hundred µm. Taking the different design constrains of these various types of scanners into account, the term mini-CT should be used to describe CT systems with a resolution ranging from 100 µm *Address correspondence to either author at the Molecular Imaging Group, Department of Medical Physics in Radiology, German Cancer Research Center (DKFZ), Im Neuenheimerfeld 280, 69120 Heidelberg, Germany; Tel: +49 (0) 6221 422686; Fax: +49 (0) 6221 422557; E-mail: [email protected]; [email protected]

to 500 µm; micro-CTs are then scanners with a resolution of below 100 µm [Kalender W (2006), personal communications] [1] (Fig. (1), Table 1). Small Animal Imaging Due to the fact that more and more research is based on animal models the interest in in-vivo small animal imaging rises. With small animal imaging methods cross-sectional study designs (e.g. where cohorts of animals are killed at each time point and histology is examined) can be replaced by an in-vivo study design permitting repeated measurements of the same animals. Each animal acts as its own control, therefore the variability that would normally be present within a cohort of animals is accommodated and changes are more readily detected by paired comparisons [2]. Additionally, CT imaging scaled down to small animals can help to define, optimize and test future performance goals of clinical-scale CT. Scaling down CT imaging to the size of small animals is challenging: The volumes of mammals’ morphological structures and organs are proportional to their weight [3]. So, to acquire CT data, e.g. of mice, that show their internal organs with detail comparable to clinical CT scans of human patients, a small animal imager needs a resolution of about 100 µm [1]. For motion-compensated imaging (i.e. taking into account cardiac and respiratory motion present during image acquisition) stakes are similar. The heart rate of a mouse is about 400-600 min-1 and respiration frequency ranges from 30-60 per minute. To use the diastole as the phase of the heart cycle that shows a minimum amount of motion, a basal temporal resolution of the CT scanner in the order of 50 ms in mice in comparison to 300 ms in humans is necessary [4].

*These authors contributed equally to this article.

1573-4056/07 $50.00+.00

©2007 Bentham Science Publishers Ltd.

46

Current Medical Imaging Reviews, 2007, Vol. 3, No. 1

Bartling et al.

Fig. (1). Examples for three different design concepts of small animal CT scanners. A bench-top micro-CT (A, B) with rotating sample holder, stationary area detector and micro-focus x-ray source offering variable magnification. Such a setup is mostly used for in-vitro imaging. An optimized relationship between scan field-of-view and resolution as well as good animal handling due to a non-rotating sample bed is provided by a rotating gantry based concept (C, D). Imposing fewer demands on spatial resolution, faster scanning and a bigger scan field-of-view can be achieved with the displayed flat-panel detector, rotating gantry-based design with stationary sample bed (E, F). Reprinted with permission from Willi Kalender, “Computed Tomography”, Publicis Corporate Publishing.

BASIC DESIGN CONCEPTS Several design factors of CT scanners such as geometry, the employed x-ray source and the detector technology influence the fundamental characteristics of CT scanners. Inherent relationships between resolution (spatial, temporal as well as soft-tissue contrast), noise, scan time, scan fieldof-view (FOV) and applied x-ray dose are imposed by the

fundamental laws of physics. Therefore there is not one scanner design that is able to optimize all of these fundamental scanner characteristics. Instead, several different scanner designs exist which are suited for different kinds of diagnostic questions (Fig. (1), Table 1). Fundamental CT design concepts will be discussed in the next paragraphs.

Small Animal Computed Tomography Imaging

Table 1.

Current Medical Imaging Reviews, 2007, Vol. 3, No. 1 47

Comparison of Micro-, Mini- and Clinical-Scale CT

Micro-CT

Mini-CT

Clinical-scale CT scanner

Suited for

Tissue samples, insects, mice, rats

Mice, rats, rabbits, primates, mini-pigs

Up to humans

Spatial resolution (isotropic)

5 µm (single limbs) – 100 µm (whole animals)

100-450 µm

> 450 µm (z-axis > 600 µm)

Transaxial scan field-of-view (FOV)

1-5 cm

5-20 cm

> 20 cm

Time to acquire a “standard” volume (e.g. a whole animal)

Seconds to hours (CT scanners with single slice acquisition within subsecond times exist)

Subsecond (0.5 s) to a few seconds

A few seconds (with rotation times down to 0.33 s)

Radiation dose

> 1 Gy can be reached

~ 10-500 mGy

< 50 mGy

Design

Bench-top, rotating sample (with variable geometry, resolution, scan field-of-view, etc.) or rotating gantry

Rotating specimen or rotating gantry (fixed geometry)

Rotating gantry (fixed geometry)

Cardiac- & respiratory motion compensation

Prospective triggering

Prospective triggering, retrospective gating

Scan modulation, retrospective gating

Example figures

Fig. (1) A, B, C, D, (3), (4)

Fig. (1) E, F, (2), (5), (6)

Overview of different in-vivo small animal CT scanner designs (values given are approximations). The term micro-CT is often used for all scanners that are smaller than clinical scale scanners, however, taking the performance differences into account the term mini-CT could be used for scanners with a resolution of around 100 µm.

Rotating Sample vs. Rotating Gantry Systems Small animal CT-scanners can be classified into two major categories: rotating sample and rotating gantry systems [5]. So-called rotating sample systems feature a stationary xray source and a stationary x-ray detector usually mounted on a mechanical bench. Both detector and x-ray tube are facing each other and have to be precisely aligned with the central axis of the CT system. A rotating sample holder is placed on the mechanical bench in between source and detector, also aligned with the central axis of the scanner system. The sample holder can usually be precisely rotated perpendicular to the central axis of the scanner system by a computer-controlled motor-driven rotating stage (Fig. (1), A, B). For in-vivo animal imaging this implies that the anesthetized animal has to be immobilized and fixated in an upright position (e.g. head up) on the rotating sample stage which demands careful animal handling [6]. In addition, administration of inhalation anaesthetics and injection of contrast agents during a scan is difficult since the needed apparatus may not inhibit or conflict with the sample rotation. Most of the rotating sample systems have the advantage that the scanner geometry can easily be modified in between scans. The source-to-isocenter (also called source-to-object) distance (SOD) can be varied by shifting the x-ray tube along the central axis of the system relative to the sample holder defining the isocenter of the scanner. The source-todetector distance (SDD) can be varied by either shifting only the detector relative to the scanner’s isocenter along its central axis while leaving the source fixed or by moving both source and detector relative to the isocenter. This flexible

design enables the user to choose the optimum scanner geometry for the imaging application. Systems with rotating gantry but stationary sample accommodate x-ray source and detector mounted exactly facing each other on the inside of a ring-shaped mechanical support (gantry). Here, the gantry containing source and detector is no longer stationary but rotates around the central axis of the scanner. The sample which is to be imaged can be placed in a prone or supine position on a “patient bed” or “patient table”. (Fig. (1), C-F). The longitudinal axis of this table is parallel to the central axis of the CT scanner and the table is usually driven by a computer-controlled micrometer stage, allowing exact and reproducible positioning of the sample within the scanner system. Small animal CT systems featuring a rotating gantry assembly can easily be identified as scaled-down clinical CT scanners adapted to the specific requirements of small animal imaging (Fig. (1), C-F). Compared to scanners having a rotating sample holder, rotating gantry based systems usually do not offer the possibility to change the scanner geometry easily, normally SDD and SOD are fixed numbers of the particular scanner system. Animal handling during examination is much easier in rotating gantry systems. In addition, these systems support faster rotation times than rotating sample systems. However, they usually are more complex mechanically and more intricate from the point of view of systems control and therefore more expensive. No matter if the small animal CT scanner is rotating gantry based or not, both system types can be either benchtop systems as shown in Fig. (1) A-D or systems requiring mechanical support structures too large to be scaled to fit a bench top, e.g. if modified clinical x-ray tubes offering higher x-ray photon fluxes than micro-focus tubes or very large area detectors e.g. to cover large FOV sizes (Fig. (1),

48


E, F) are to be used. An overview over commercial and experimental laboratory micro-CT scanners of both types can be found in [7], a compilation of the most recent small animal CT scanner models, including those featuring flatpanel detector systems can be found in [8]. Scanner Geometry Independent of the x-ray source employed or the detector technology chosen for a particular small animal CT-scanner there are two possibilities for CT scanner geometry: the socalled “short” scanner geometry where the SOD is small compared to the object-to-detector distance (ODD) and the so-called “long” scanner geometry where SOD and ODD have about the same size [Kachelriess M (2006), personal communication]. Both geometries have their advantages and disadvantages on the imaging properties of the scanner but share the requirement that the small animals imaged should be entirely covered by the available FOV. Short scanner geometries place the animal close to the xray source, the ODD being larger than the SOD. The image projected onto the detector can thereby be magnified by the factor ODD , M= SOD improving spatial resolution of the scanner system. However, the resulting FOV is being decreased by the same factor requiring larger detectors. With the advent of largearea flat-panel detectors higher magnification factors will be possible, since the active detector area will increase significantly. Placing the animal close to the x-ray source, however, requires the use of micro-focus x-ray tubes since larger focal spot size will introduce a significant image blur (so-called penumbral blurring), which degrades resolution. The problem of penumbral blurring is aggravated by high magnification of scanner systems since the penumbra of the source’s focal spot is also magnified by the factor M. Long scanner geometries also have their advantages. Image blur caused by the finite size of the x-ray source focal spot is decreased if the object is placed closer to the detector instead of being placed close to the source. However, the xray photon flux Φ will decrease exponentially with

1 , SOD 2 which can be compensated by x-ray tubes with higher output (but normally bigger focal spot because of anode heating). Otherwise, the scan time per projection needs to be increased or contrast resolution will be degraded since (statistical) image noise (being related to the attenuated photon flux

Φ att reaching the detector by

1 ) will increase. Φ att

Bartling et al.

The larger the distance between source and object (SOD) is, the smaller the skin entrance dose to the animal imaged will be. This implies that although scanners with short geometry usually are more dose efficient and allow to geometrically magnify the object, thereby improving spatial resolution, long scanner geometries can significantly reduce the skin entrance dose of the animal under investigation. X-Ray Source The scanner geometry and the desired spatial, lowcontrast spatial and temporal resolution govern the choice of x-ray tubes for small animal CT imaging. An x-ray source for micro- and mini-CT has to fulfill three demands: the focal spot size has to be as small as possible, the source has to emit a high photon flux Φ, and x-ray energies should be selectable over a suitable range. As already mentioned above the focal spot size should be as small as possible in order not to have a negative influence on spatial resolution by image blurring. However, the size of the focal spot limits the available x-ray photon flux. For small focal spot sizes and stationary targets, heat dissipation from the target area is proportional to the diameter of the focal spot and radial to first order, the maximum power emission being

Pmax ≈ 1.4 ⋅ ( x f ,FWHM )

0.88

, with

x f ,FWHM

being the focal spot size in microns [9]. Since the emitted xray photon flux Φ is roughly proportional to the product of the x-ray anode current I and the square of the tube voltage U and the tube power is P = U•I, the available x-ray flux is limited by the size of the focal spot, which can only absorb a certain amount of heat (anode heating). The maximum power that an x-ray tube can emit thus also depends on the heat capacity of the anode material and the tube technique used: tubes with rotating anode support much higher output than tubes with stationary anodes since heat is evenly distributed along the focal spot trajectory. High x-ray flux is desirable for small animal CT imaging to achieve high temporal resolution and short scan times: if the photon flux is high enough, sufficient x-ray photons reach the detector and can be collected in short times for each projection. A sufficient amount of x-ray photons is required to limit image noise and allow good low-contrast spatial resolution [6]. Through short scan times in-vivo scanning of animal subjects is eased since the animals do not have to be anesthetized for long times. Also, high photon fluxes are needed to ensure high temporal resolution for perfusion studies or motion compensated imaging (ECG or respiratory gated imaging). In order to achieve the best low-contrast spatial resolution in small animal CT imaging, the x-ray tube should allow choosing a range of tube voltages, thereby selecting appropriate x-ray energies E. For higher x-ray energies the energy-dependent absorption coefficient µ(E) is small and low-contrast spatial resolution is limited due to the small number of x-ray photons absorbed in the animal; if x-ray energy is low and thus µ(E) large, most photons are absorbed


in the animal and the contrast resolution is limited by the small amount of x-ray photons reaching the detector [9]. In small animal CT imaging photon energies in the range of 3050 kV are commonly applied. Hereby, the advantage is that x-ray attenuation is up to two orders of magnitude higher than in clinical CT systems usually operated at 80-140 kV which allows better discrimination of soft tissue types [1]. These low-energy x-ray photons are still able to penetrate the animal since its examined volume is much thinner than that of human patients. Additionally, the highest absorption differences of iodine lie in the low energy range, leading to an improvement in contrast-enhanced scans for the contrast between iodinated contrast agents to the surrounding tissue. For small animal CT scanning three different types of xray sources have been used so far: micro-focus x-ray tubes, diagnostic x-ray tubes as in clinical-scale human patient radiological equipment and diagnostic x-ray tubes modified to fulfill the special requirements of small animal CT. Micro-focus x-ray tubes with focal spot sizes ranging for 5-50 µm (~ 5-200 W power emission) have mostly been employed in micro- and mini-CT and have the advantage of allowing high isotropic spatial resolution down to a range of 10-50 µm for FOV sizes of 30-50 mm [7]. Others have employed diagnostic x-ray tubes with larger focal spot sizes (0.3 mm at 9 kW and 1.0 mm at 11 kW) in order to profit from the high x-ray flux at short exposure times [6, 10, 11] to increase temporal resolution for cardiac and respiratory gating in small animal CT while still achieving an isotropic spatial resolution of 100 µm. Experimental small animal CT systems with modified diagnostics x-ray tube exist [12]. Here, the tube filament was shortened to decrease the focal spot size of the rotating anode x-ray tube while still offering considerable photon fluxes which make high scan speed possible and are needed for the flat-panel area detector. Detector Technology Detector technology plays a crucial role for the performance characteristics of all small animal CT systems. Together with the focal spot size of the x-ray source the size of the detector elements has the biggest influence on the spatial resolution of the images, their size also influences the low-contrast resolution together with x-ray flux. No matter what detector technology is used (see below) in a particular small animal CT system, all of its parts (e.g. image intensifier screen, optical coupling medium, photo detector, etc.) have to be optimized for the following factors: high quantum efficiency and uniform response throughout its whole surface at the chosen x-ray energies and good pixel resolution through small detector pixelation. The detector’s dynamic range needs to be as large as possible to ensure good low-contrast resolution and high dose efficiency, whereas its noise and dark current should be as low as possible. A high read-out rate is favorable for the systems to achieve a maximum temporal resolution and to minimize scanning times. Of course the area of the detector should be as large as possible so large FOVs can be employed and none of the detector elements should introduce geometrical distortions [9]. From the first developments of small animal CT systems onwards different detector types have been used. Apart from


an overview over the different types that have been used to date we limit our focus to the emerging flat-panel x-ray detectors, which will be elaborated on in more detail. For a detailed look on digital x-ray detector technology please refer to [13]. As in clinical-scale CT systems, detector systems in micro- and mini-CT can consist of linear arrays of photodiode detector elements illuminated by a so-called “fan beam”. This approach to small animal CT was used in some of the early systems but is still used by some commercial small animal CT manufacturers [7]. After the development of the so-called “cone-beam” reconstruction algorithm [14], the first three-dimensional small animal CT systems were described end of the 1980s and beginning of the 1990s until the early 2000s. They employed two-dimensional detectors consisting of phosphor screens which were optically coupled to cooled charge-coupled detector (CCD) arrays via an optical lens or a fiber-optic taper (e.g. [15, 16]). Though being inefficient, the optical coupling via lenses offers the possibility of variable image magnification, whereas fiberoptic coupling is very efficient but uses fixed magnification [1]. This detector setup is technically demanding but remains one of the most sensitive x-ray detection methods [7]. Holdsworth and co-workers (1993) used x-ray image intensifier (XRII) screens coupled to a video readout [17]. The rapid illumination response to x-ray radiation is the main advantage of image intensifiers [1]. Since the 1990s CCD-based detector arrays with scintillating plates (e.g. made of CsI(Tl), GdO2SO4, etc.) have become common in small animal CT systems [6]. Recently, flat-panel area detectors have been introduced to small animal CT imaging [10, 12, 18-20]. An overview of indirect (x-rays are converted to light in the scintillating layer first, then the light is detected by photodiodes and subsequently amplified) and direct conversion (x-ray photons are directly detected) flat-panel detector types and their technical differences can be found in [21]. The flat-panel detectors used in recent flat-panel small animal CT systems are of the indirect conversion type. They usually consist of a scintillating crystal layer like thalliumdoped caesium iodide (CsI(Tl); or GdO 2SO4, etc.) coupled to a photodiode array in form of an amorphous silicone (a-Si) wafer [10, 22, 23]. Others have employed photodiode arrays fabricated by a complementary metal-oxide semiconductor (CMOS) process [18, 19]. Compared to XRII detectors flatpanel detectors have some advantages: their structure is thin and they allow for large-area detection (see below) without geometrical distortions [24]. Because of technological advances they can be produced in good quality and with high precision. The hydrogenated a-Si material used has the advantage that it can be deposited on very large areas (making large FOVs possible) at relatively low temperatures (200-250° C, easing fabrication), has the properties of a semiconductor (photoconductivity in the visible spectral range) and does not age through x-ray exposure [25], which is a drawback of CCD- and CMOS-based detectors. However, a-Si flat-panel detectors have several drawbacks: the pixel size of around 100-200 µm2 is relatively large compared to pixel sizes of ~2.5 µm2 achievable with CCD detectors [16], due to low fill-factors of ~ 45-70 %

50


([18, 26]) the geometric efficiency is not as high as in CCD detectors, and the so-called image lag (ghost images of previously acquired projections due to charge traps in the aSi layer, e.g. dangling bonds in the amorphous structure) has a negative influence on spatial, low-contrast and temporal resolution [27]. The effects of image lag can be quite serious on image quality. They are described for flat-panel computed tomography in general in [28]. Algorithms are applied to correct for the effect of image lag; projection frames previously acquired are weighted and subsequently subtracted from the currently acquired frame [23]. Compared to other detector types, flat-panel detectors in small animal CT also suffer from their smaller dynamic range which can deteriorate image quality: insufficient resolution of pixel gain and offset normalization causes ring artefacts and truncation of low-density anatomical detail due to insufficient dynamic range results in shadow artefacts and incorrect Hounsfield numbers (HU) [29]. In addition, flatpanel detectors need frequent and careful recalibration for offset and pixel gain factors [23, 29]. Nonetheless, the advantages of flat-panel detectors used in small animal CT imaging by far outweigh the disadvantages stated and cited here. Further developments in flatpanel detector technology ([21, 29]) will allow even more studies with and new applications of small animal CT. The use of other detector materials like CdTe or CdZnTe for flatpanel detector fabrication might increase detector efficiency for the x-ray energies employed and thus lead to improved soft tissue contrast [30]. Spatial Resolution In small animal CT spatial resolution is one of the key parameters since the size of morphological structures is approximately proportional to the weight of the (laboratory) animal examined. This leads to the requirement that small animal CT systems should be capable of providing a spatial resolution in the order of 100 µm. Spatial resolution of small animal CT scanner systems is mainly determined by the following factors [9]: the geometry of the scanner system, which can be chosen to allow geometrical magnification of the specimen imaged, the size of focal spot having an influence on the magnitude of penumbral blurring (see below) and the chosen detector technology, which determines the minimum image pixel or voxel size through its intrinsic resolution. Thus, the size of the single detector elements sets the maximum spatial resolution a particular micro- or mini-CT can theoretically achieve, even if image reconstruction features image matrices with smaller pixel or voxel sizes. Reconstruction voxel sizes chosen too large compared to the intrinsic detector pixel size reduce the spatial resolution of a CT scanner, while very small reconstruction voxels size do not increase its intrinsic resolution. Ideally, the reconstruction voxel size should be chosen to be half the size of the intrinsic resolution of a CT scanner complying with Nyquist´s sampling theorem. Per definition spatial resolution is the smallest distance between two objects still being distinguishable as two separate entities. In order to describe the spatial resolution of CT images obtained from a particular scanner, the modulation transfer function (MTF) is employed. The MTF plots the percentage of contrast that is transferred from the

Bartling et al.

imaged object (e.g. a line-pair phantom) to the CT image as a function of spatial frequency (expressed as the number of line pairs per unit length). It is common use to state the spatial frequency which corresponds to 10 % contrast in the CT image as the spatial resolution of the scanner. I.e.: the finer the CT image detail still resolvable the higher the spatial resolution of the scanner and thus the spatial frequency at 10 % modulation transfer [12]. For small animal CT scanners the so-called slant-slit method is often used to calculate the MTF of a scanner [18, 19]. Here the phantom consists of an acrylic plate having a rectangular slit covered by a very thin metal foil (e.g. 18 µm of aluminium [18]). The cross-section of the foil is the slit which can be slightly tilted with respect to the horizontal or vertical axis of the image plane so that the line-profile of the slit in the image domain can be sampled at sub-pixel level. Then the MTF can be calculated as the Fourier transformation of the line profile. In practice, the spatial resolution of clinical scale CT scanners and small animal CT flat-panel scanners featuring a clinical-scale gantry is usually measured by scanning linepair phantoms, which provide sets of metal line-pairs (e.g. tungsten bars) of various thicknesses at decreasing distances cast into plastic material. Here the smallest feature size (in millimeters) that is still visible can easily be calculated by dividing 10 mm by the amount of distinguishable linepairs/cm times two. The maximum theoretically achievable spatial resolution according to the employed detector’s intrinsic resolution and the geometrical magnification of a particular small animal CT system can seldom be achieved [5]. Apart from the mechanical stability (e.g. vibrations, etc.) of the scanner and the influence of the reconstruction algorithm, the image blur introduced by the finite size of the focal spot ( x f ,FWHM ) of the x-ray tube limits the spatial resolution of the real system (“penumbral blurring” is

b=

ODD ODD x f , FWHM = x f ,FWHM SDD − ODD SOD

[6]). To reduce the influence of the focal spot shape and the associated penumbral blurring Popescu and co-workers [23] have modified and measured the focal spot shape and size of the customized clinical x-ray source employed in the flatpanel CT used for small animal imaging [31]. The measurements of Popescu and co-workers [23] included an investigation of the focal spot shapes throughout the whole detector plane. These authors have shown that small distortions of the focal spot introducing a negligible amount of image blurring can still be observed even if the tube is modified to match source and detector setup. In order to achieve the spatial resolution of interest, fluxes of 104-105 photons have to traverse a particular region of interest (ROI) per resolution unit at minimum [32] so a detector signal significantly larger than the statistical noise due to the quantum nature of the x-ray photons can be gained. Higher photon fluxes allowing shorter exposures times (and thus better temporal resolution) require larger x-



ray tube focal spot sizes but dictate a relaxation of the requirements of spatial resolution and geometrical magnification of a small animal CT scanner system. Low-Contrast Resolution The highest spatial resolution achievable in small animal CT is of little use unless the scanner is able to resolve small differences between structures and neighbouring tissues having a poor inherent contrast. In any CT system, contrast in the images is a consequence of the difference in the energy-dependent linear absorption coefficient µ(E) (also called “attenuation coefficient”) of different tissues [7]. Thus, low-contrast detectability (LCD) can be defined as the ability to detect fine variations in the (electron) density of an object over the background [12]. The low-contrast resolution capabilities of a small animal CT scanner largely depend on the type of detector used: the more quantum efficient the detector is, the lower the amount of incident photons to generate a detector signal can be; and the larger the detectors dynamic range is, the smaller the differences in photon flux still detectable and differentiable can be. This implies that the low-contrast resolution capability of a scanner also depends on the x-ray dose applied per projection (view) of a scan (see below), since resolving low-contrast structures actually means collecting detector signals whose difference is significantly larger than the noise level present. Micro-CTs that are not specifically designed for small animal imaging usually have inherent limitations in signal-to-noise ratio (SNR) performance because of their small detector voxel size (to achieve high spatial resolution) and their low x-ray exposure level (caused by the employed micro-focus x-ray tubes). Therefore, their low-contrast resolution is traded for high spatial resolution of high-contrast structures such as bones [18]. Other investigators trying to achieve high low-contrast detectability have employed x-ray tubes featuring large foci [6, 33]. In addition, a careful selection of the x-ray energy appropriate to the imaging task is necessary if low-contrast detectability should be optimal. Experimental evaluations of the low-contrast resolution of small animal CT systems usually employ custom made contrast phantoms. These may consist of water-filled acrylic hollow cylinders with small cylindrical low-contrast inserts with densities close to the density of water [18]. They allow to either determine the minimum dose level at which the low-contrast inserts can be discriminated from the surrounding water or used to quantify the low-contrast resolution performance of a small animal CT at a fixed x-ray exposure by measuring its contrast-to-noise ratio (CNR):

CNRi =

Si − Sb σ i2 + σb2

,

with S and σ being the mean CT number and the standard deviation of the image pixel values in a specific ROI in HU, the subscripts stand for the inserts (i) and the background ( b) [19]. Ideally, the CNR in CT images is proportional to the square root of the applied dose [18, 34].

Contrast agents are used in small animal CT like in clinical CT to increase low-contrast resolution. The increase in CT number (∆CT Number) caused by the contrast agent applied can be approximated by

∆CT Number ∝

Here

′ µ contrast medium (E ) ′ (E ) µ water

⋅ C ([9, 35]).

C is the concentration of the contrast agent in the

tissue (in mg/ml) and

µ′ ( µ′ = µ ρ

[cm g]) is the 2

attenuation coefficient normalized for density. Image Reconstruction In small animal CT the same image reconstruction techniques as in clinical-scale CT are applied. CT systems using fan-beam geometries apply fan-beam reconstructions based on filtered back-projection algorithms originating from the early days of clinical CT. The recently introduced flatpanel based micro- and mini-CTs employ variants of the Feldkamp-David-Kress (FDK) algorithm [14] for cone-beam reconstruction. Both reconstruction techniques have to be adapted to the specific small animal scanner geometry. An overview over the underlying principles of image reconstruction can be found in [9, 35]. Contrast Media Non-ionic extravascular water-soluble contrast agents, as typically used for clinical CT examinations, rapidly clear from the blood within minutes after intravenous injection. The first-pass of the contrast media with the highest vascular contrast is only retained for a few heart beats. This time frame is too short for image acquisition with most micro-CT scanners. Contrast media that remain in the vasculature for longer are called blood-pool contrast agents. In order to achieve blood-pool effects the size of contrast media molecules can be increased to larger sizes than that of capillary fenestrations. Furthermore, phagocytosis by the reticuloendothelial system should be avoided as long as possible through the chemical design of the contrast media. Several different agents have shown potential as blood-pool agent for CT imaging [36]. A water-soluble macromolecular agent (dysprosiumDTPA-dextran) was reported to provide blood-pool contrast enhancement for up to 45 minutes [37]. Liposomal encapsulation of iohexol resulted in a blood-pool effect for up to 3 hours after injection in rabbits [38]. In rats, iodine-containing micelles caused enhancement in the blood, liver and spleen for more than 3 hours [39]. An iodinated triglyceride emulsion (ITG-LE) packed into the lipophilic core of a synthetic chylomicron remnant has been described in [40]. It can be used to image liver parenchyma because it is internalized by the hepatocytes, whereas liver tumor cells do have less functional lipoprotein receptors and therefore only show low enhancement. The

52


liver sequestration of ITG-LE from the blood-pool can be delayed by adding polyethylene glycol-modified phospholipids (ITG-PEG) to the polyiodinated triglyceride emulsion resulting in a longer blood-pool time. The combined application of both formulations of ITG enhances the healthy liver parenchyma as well as the intrahepatical vessels at the same time. This reduces the risk to misinterpret vessels as tumor lesions. False-positive results may occur with hepatocytespecific contrast media alone, because liver vessels as well as liver tumors are non-enhancing and can therefore be confounded [41]. Dual-phase contrast enhancement can be achieved with contrast media that first circulate in the blood and are subsequently cleared via the hepatobiliary pathway hereby enhancing the liver and the spleen for hours [36]. Also nanoparticles may deal as CT contrast agents. An example is nanosized bismuth-sulfide with a polymer coating. These nanoparticles were described to have x-ray absorption characteristics that are five times better than that of iodine-containing compounds, which reduces the risk of viscosity problems [42]. The vascular half-life is longer than that of iodine-based contrast media and their efficacy/safety profile is comparable or better. Imaging of tracheobronchial lymph nodes with nanoparticles in CT has already been shown in dogs [43]. X-Ray Dose Dose applied during image acquisition is of special concern in small animal CT, especially if follow-up examinations of the same animal are necessary. When all other hardware and acquisition parameters are kept constant a good low-contrast spatial resolution requires the application of high x-ray doses: in order to be able to differentiate neighboring tissues with similar (electron) densities with a significant SNR, a particular voxel of the imaged animal being the smallest unit of resolution needs to interact with a certain minimum amount of photons. The minimum amount of required photons is constant regardless of the voxel size. However, dose is a function of the energy deposited through photon absorption and energy loss via Compton scattering interactions per volume, therefore dose needs to increase with the power of four with the resolution element if photon noise is supposed to be constant. In other words: “An increase in resolution from 1.0 mm to 0.5 mm has to be paid for by a 16-fold increase in exposure if we expect to see low-contrast lesions of 0.5 mm diameter with the same clarity as lesions of 1.0 mm diameter in the previous case” [34]. Small animal CT needs to be performed with higher resolution than human CT scanning because anatomical structures within laboratory animals are smaller than in human bodies. In order to achieve spatial and low-contrast resolution (and thus noise levels) comparable to human CT imaging higher x-ray doses are physically needed. For example, mice imaging with an isotropic resolution of 135 µm requires a dose of 250 mGy supposing an ideal small animal CT-scanner is used for the examination [44], which amounts to a 10-30 fold increase in dose compared to a CT examination of a human patient [45, 46]. An isotropic resolution of 135 µm results in a three-fold increase in

Bartling et al.

spatial resolution compared to clinical CT, whereas the theoretically expected x-ray dose would increase by a factor of 3 4=27 (cf. [34]). This would mean that imaging mice with the same image quality at an isotropic resolution of 35 µm would require a dose of 2500 Gy. From these studies one can conclude that fundamental laws of physics impose the limits of dose in small animal imaging and thereby the limits of invivo small animal imaging [44]. Depending on the diagnostic demand, however, x-ray dose can be decreased while maintaining high spatial resolution if cut-backs in other image quality parameters such as image noise and soft-tissue contrast can be accepted. Much higher image noise is acceptable e.g. when high-contrast structures such as bones, lung or contrast media filled vessels are the focus of the examination. So far only whole body doses have been discussed, which means that depending on the experimental focus higher doses could be applied to organs or parts of the animal which are less dose-sensitive. Dose limitations for imaging living animals are usually stated in terms of the lethal dose (LD): LD50/30 is the whole body radiation dose which would kill 50% of an exposed population within 30 days of exposure. Previous studies indicated that the lethal dose of mice ranges between 5-7.6 Gy [44, 47, 48], depending on the strain [49], age and other factors. This range certainly sets the upper limit for the whole body x-ray exposure of a single examination of a living animal. Fortunately, small animal CT examinations usually do not even get close to this limit. Since the main advantage of in-vivo small animal CT lies in the fact that follow-up studies can be performed the effects of sublethal dosages become relevant. The sum of the biological effects of many sublethal doses is a complex function that is influenced by many factors. While the lethal dose might give an absolute limit, exposure to much smaller x-ray dosages can already have biological effects, which might interfere with experimental results. Several low-dose radiation effects are described including effects on tumor growth, hematopoiesis [45, 50] and bone growth [2]. If follow-up studies are performed over a long timeframe a possible contribution of radiation effects to the observed results needs to be assessed: a group of animals that is only imaged at the end of the time series could act as a control in comparison to the scanned animals [51] or the contralateral limb not having been exposed to radiation could be taken as a non-irradiated control [2]. Single organ doses for mice imaging were calculated using a Monte Carlo simulation. Here the “typical” whole body dose was 80 mGy (at 80 kV tube voltage) to 160 mGy (at 50 kV tube voltage). Furthermore, a strong dependance on the tube voltage as well as on the position within the animal was found [50]. To allow the computation of x-ray exposure doses from air kerma measurements dose coefficients for standardized mice, typical scanner geometries and x-ray spectra were calculated using Monte Carlo simulations [45]. NEW DEVELOPMENTS Flat-Panel Based Small Animal CT Flat-panel detector based micro- and mini-CT scanners for small animal imaging have been described in bench-top


systems [18] as well as in gantry-based systems [10, 12, 20, 52]. As discussed in Detector Technology, flat-panel detectors have several benefits over detectors that were used for micro-CT earlier. Large flat-panel detectors are ideally suited for gantrybased small animal CT imaging systems with a geometry that allows a reasonably high resolution with a large gantry bore and a wide z-coverage so that bigger animals such as rabbits and piglets [10, 12] can be scanned. The data of even bigger animals can be acquired in one rotation. A decrease in spatial resolution allows a relaxation in dose performance and therefore a very fast acquisition of single projections and therefore the whole volume in short scan times. Fast, continuous scanning becomes possible. Acquiring the same volume in short succession is a prerequisite for perfusion imaging of fast contrast media dynamics. Slow contrast media dynamics such as kidney or liver uptake and bladder or gall bladder filling could already be assessed with slower micro-CT scanners. However, fast contrast media dynamics such as brain, tumor and scar perfusion is of much higher interest. Using flat-panel detector CT scanners, exemplary perfusion imaging of fast dynamics for (whole) small animals has been shown [52]. This technique could provide valuable physiological and functional information of many small animal models and diseases. Several post-processing methods could be applied to the contrast media time course data. First experiments were promising: the time course of contrast media could be traced, and parameters such as Mean Transit Time, Blood Volume and Blood Flow of experimental tumors could be calculated [52]. Fast and continuous scanning could also make retrospective cardiac and respiratory gating possible. For this goal, the short exposure time of flat-panel CT that virtually freezes the heart motion on the level of projections is especially advantageous. Earlier micro-CT scanners were limited with regard to motion gating by the long exposure times per projection.


edge of the selected element). When both data sets are subtracted, density differences only remain for the selected elements. However, the availability of synchrotron radiation is currently limited to expensive synchrotron generators. Furthermore, multi-source CT imaging could decrease the scan time by radiating the object from several directions at once. APPLICATIONS The following chapter gives an overview of the applications of micro- and mini-CT in small animals. In the compiled studies herein various scanner designs with a range of scan parameters, scan times and spatial resolution are used. Most studies are proof of principle studies. Nevertheless, the conclusions drawn from one study hold true for other studies in case that similar mini-/micro-CT scanners with comparable characteristics were used. Lung Imaging In mice, the total lung volume is 1.3 cm3. The thorax of mice is about 2 cm in diameter. Proportionally, a mini-CT scanner that provides an insight on the mice lung structures as in a human lung should offer a resolution in the order of 75 µm [54]. In vivo lung imaging is challenged by the respiratory and cardiac motions and its compensated imaging was described by Badea and co-workers [33]. Beside the lung parenchyma, the larger thoracic structures such as heart, oesophagus, trachea, bronchi and large vessels can be imaged in non-enhanced scans (Fig. (2)). Lung Tumors The detection of experimentally induced lung tumors is possible in mice using micro-CT. This was shown in a study, in which urethane induced lung tumours were imaged in living mice. The scans were performed without motion compensation and with an isotropic resolution of 35 µm [55]. Only few, small tumors that were found in histologic slices were missed in the micro-CT scans.

Compared to prospective triggering, retrospective gating could decrease the scan times significantly. This could make the use of standard, iodinated contrast media possible and may open perspectives to lung or cardiac perfusion studies. Future Perspectives For a good overview of future developments of micro-CT with focus on in-vivo small animal imaging please read the review from Ritman [53]. If big technical and engineering challenges could be solved, techniques such as x-ray fluorescence, x-ray diffraction or x-ray scatter imaging would be big steps for CT imaging. Their potential for the improvement of soft-tissue contrast and contrast media sensitivity are enormous. However, a long time might still pass until these technologies will reach operational readiness. Another method possibly available in the near future is K-edge subtraction imaging. It is based on synchrotron generated narrow bandwidth rays that allow imaging of certain chemical elements by imaging the volume twice at different photon energies (once with a frequency just above and once with a frequency just below the characteristic K-

Fig. (2). In-vivo thoracic imaging of a mouse (A) and a rat (B) from flat-panel based CT. Data acquisition time was 5 s. However, motion artefacts close to the ribs and diaphragm are pronounced. Bronchi (white arrow) and vessels (black arrow) can be distinguished from the lung parenchyma.

54


Follow-up studies of lung tumor size changes over time could be performed [56-58]. Size distortions of the lung tumors due to motion artefacts should be taken into account when micro-CT without appropriate motion-compensation algorithms is used for absolute measurement studies. The minimal sizes of lung tumors that were found in several studies were: 6.6 mm3 in volume and 0.85 mm in diameter in the free lung parenchyma and 1.4 mm in diameter perihilar [58]. In other studies the size of lung nodules regardless of their localization was reported to be 0.63 mm3 minimum in volume [57] and 0.5 mm [56], respectively, smaller than 0.2 mm in diameter [55]. In this context, using a ventilator and prospective respiratory gating the accuracy of tumor volume determination could be improved [57]. In the reviewed studies almost all tumors that were identified by micro-CT were confirmed histologically, but vice versa not all tumors that were found in histology were reliably identified in micro-CT even if they had the same size as tumors that were found. Juxtahilar tumors were most easily to detect, while the tumors next to the thoracic wall and the big vessels were most difficult to visualize, which is most likely due to motion artefacts that lead to a smearing of the thoracic walls. Furthermore, it is difficult to differentiate tumors from hilar structures and from vessels. Similarly, because the lung tissue in the reviewed studies was normal apart from the induced tumors, it remains to be proven by other studies, how accurate micro-CT can differentiate between scar tissue, microgranulomas, hyperplastic lymphocellular or other non-malignant lesions in lung tissue. The use of macroscopic differential features to distinguish between tissues as known from human thoracic imaging is certainly limited by the resolution. Blood-pool contrast media proved to be valuable to improve differentiation of tumors and vessels [58].

Bartling et al.

Even pleural effusions occurring as the consequence of tumor disease could be successfully imaged with micro-CT [58]. General Lung Parenchyma Changes (Emphysema and Fibrosis) Using micro-CT the assessment of generalized lung parenchyma changes is possible as shown in following experiment: For example lung emphysema was induced in C57BL/6J mice by intratrachael instillation of pancreatic elastase. The mice were scanned using a micro-CT without motion compensation and an isotropic resolution of 35 µm. On CT images mice treated with the highest dosage of elastase showed the highest amount of pixels with low HU values within their lungs. Furthermore, the lung volume was larger in this group. In conclusion, lung emphysema can be detected by micro-CT indirectly by the lower density of the lung and by an increase in lung volume [59]. Unfortunately, despite the sufficiently up-scaled spatial resolution a direct assessment of the altered emphysematous lung structure as in human high resolution (HR)CT-imaging is not possible [54]. This is mainly due to motion artefacts and resolution limitations. Vice versa, also the assessment of lung fibrosis in mice is possible by analysis of CT numbers. Here the disease progress was shown to be associated with an increase in CT numbers on a clinical CT-scanner [60]. Using mini-CT it was shown that not only the general density could be assessed but also structural changes of bleomycin-induced fibrosis such as ground glass opacity, septal thickening, fibrotic strands and secondary dilation of the bronchial system (bronchiectasis) [52]. Bone Imaging Similar to its applications in very high resolution ex-vivo bone imaging micro-CT can provide information about bone

Fig. (3). 3D reconstruction (A) and 2D slice (B) of a rat knee and its adjacent bones scanned with a bench-top micro-CT shown in Fig. (1) B. The rat was rotated between the detector and the x-ray source while only the limb was radiated. High spatial resolution of 6 µm3 could be achieved because a single limb fits into a very small scan field-of-view (image courtesy Prof. W. Kalender, Institute of Medical Physics, Erlangen, Germany).


structure and its changes in vivo in small animals, however, with stricter resolution constraints. Microstructural Bone Imaging Using micro-CT a morphological bone analysis of its micro-architecture can be performed in vivo providing several descriptors such as bone volume ratio, bone surface ratio, trabecular thickness, trabecula separation, trabecular number, connectivity density and structural model index [6163]. CT density values as a questionable surrogate marker for bone status could also be acquired using micro-CT. Dualenergy scans as used in clinical CT setups to assess the bone density have not been reported using micro-CT yet. Small animal bone scans can be performed on single limbs. Customized jigs can assure immobilization. Due to the low diameter of those limbs resolution optimized scanner geometries (e.g. 15 µm isotropic [2]) with only a small scan FOV can be applied (Fig. (3)). Macroscopic Skeletal Imaging Micro-CT can be used to image the whole macroscopic animal skeleton. Applications are studies on the skeletal development [64] or studies in which structural changes in the skeleton anatomy occur as a consequence of disease. For these applications, the demand for large scan volumes as provided by clinical-scale CT or flat-panel based mini-CT scanners is higher than the need for high resolution. Clinical scale scanners as well as flat-panel based mini-CT systems can provide an FOV that is big enough to image a whole mouse or rat during one acquisition. Using appropriate post processing the whole skeleton can be displayed in high spatial resolution three dimensionally [20, 52].


Bone Metastasis Screening Micro-CT can be used to screen with a high sensitivity and specificity for osteolytic bone metastases in whole mouse skeleton [58]. Visualization of the soft tissue of the metastases that extend beyond the bone surfaces, however, is a challenge for the soft-tissue contrast resolution capability of the micro-CT. Its detection has not yet been described using micro-CT. Angiographic Imaging In-vitro imaging using micro-CT resulted in astonishing results of the vasculature and microvasculature of small animals. Often, these images were generated after casting the vasculature with contrast media such as silicon-based compound (Microfil MV-122, Flow Tech, Carver, MA, USA) [65]. For in-vivo imaging various contrast media have been discussed in Section. Due to the long scan times of micro-CT mainly blood-pool contrast agents have been used to image the vasculature (Fig. (4)) [37-39]. The detection of the aorta, pulmonary trunk, inferior vena cava, renal artery and vein in mice and rats [35, 66] has been described. Standard iodinated contrast media typically used for clinical examinations clear too fast from the vessels to be used in micro-CT scanners. However, mini-CT-scanning in such short times became recently feasible. In-vivo imaging of all central vessels of mice and even that of dilated subcutaneous vessels that drain implanted tumors as well as smaller vessels inside the tumor were demonstrated using flat-panel mini-CT in combination with standard iodinated contrast media (Fig. (5)). Characteristics of tumor vessel architecture could be assessed [52] and their changes traced

Fig. (4). Blood-pool contrast media enhanced scan of a rat in an early (A) and later (B) phase as scanned by a micro-CT shown in Fig. (1) D. The scanner’s spatial resolution is optimized for the dimensions of rats and mice. Due to the scan time of 180 s blood-pool contrast media were used (image courtesy Prof. W. Kalender, Institute of Medical Physics, Erlangen, Germany).

56


Bartling et al.

artefacts, which reduce the accuracy of thoracic imaging of small animals and make reliable measurements e.g. of lung tumors more problematic [57, 67] (Fig. (2)). To compensate for such movements prospective as well as retrospective gating and triggering methods can be applied. In prospective methods, the imaging process is modulated according to the movements of the small animal. Therefore only one animal can be imaged during one acquisition. Prospective methods take longer scan times than retrospective methods but are usually more dose efficient than retrospective methods. Regardless of whether prospective or retrospective methods are used, both methods add up projections acquired during defined phases of cardiac or respiratory cycles to form one complete dataset for CT image reconstruction. Therefore, position reproducibility between several breathing as well as cardiac cycles is crucial. However, this reproducibility is not better than 100 µm in small animals [68], limiting the maximum spatial resolution of thoracic imaging of free-breathing small animals to a resolution in this order.

Fig. (5). Mini-CT angiography using standard, iodinated contrast media of a mouse scanned using a flat-panel detector CT as in Fig. (1) F [12]. 100 µl contrast media were injected three seconds before scan; data acquisition took three seconds, resulting in a late arterial/venous phase. A volume rendering, in which the slice positions of the axial slices (A-D) are indicated, is given in (A). Furthermore, a coronal slice is given in (E), showing the kidneys (long black arrow) and suprarenal glands (long white arrow). The contrast enhancement in the caval vein (short black arrow) is more pronounced than in the aorta (short white arrow). Small structures such as the splenal vein (white dashed arrow), the hepatic venous confluence (white arrow head) and the suprarenal vessels (black dashed arrow) can be detected. Regardless of cardiac motion the right and left ventricle (black asterisk) can be distinguished clearly.

So far predominantly prospective methods have been implemented in small animal motion compensated CT imaging: To physically stop breathing excursions small animals can be intubated and ventilated. Breathing can be arrested for a short time period. Due to the long scan times of micro-CT complete datasets can not be acquired during one breath hold. However, the short breathing movement arrests that occur during the physiological ventilation plateau phase can

in follow-up studies [20]. Assessment of Slow Contrast Agent Kinetics To assess the slow time-resolved contrast media enhancement in micro-CT long circulating contrast media were used [39, 66]. Other authors assumed that even with iodinated contrast agents quantification of several aspects of renal function as shown in clinical CT imaging may be successful [1, 35]. The renal cortex, medulla, pelvicalyceal system and ureter could be differentiated in contrast enhanced studies. A hydronephrosis model has been characterized using this method successfully [35]. Through the use of faster flat-panel CT scanners, also fast dynamic contrast-enhanced applications may become broadly available [52] (Fig. (6)). Cardiac and Thoracic Motion-Compensated Imaging Thoracic imaging of living animals is challenged by heart and lung movements. These movements lead to imaging

Fig. (6). Perfusion CT using standard iodinated contrast media of a bone metastasis model in rats [76] as scanned with flat-panel CT. The contralateral (black arrow) as well as the partially destroyed ipsilateral tibia (white arrow) can be seen. Rotation time was 6 s, half-scan reconstruction was performed every 3 s using 180° data. Injection of contrast media was initiated together with scanning. At 0 s (A) no contrast media was detected in the selected ROI, which covered the bone metastasis, while at 24 s (B) pronounced rim and vessel enhancement can be found. The time-course of averaged CT numbers within the ROI is given in (C) (unpublished data by Dr. Tobias Baeuerle, German Cancer Research Center (DKFZ), Heidelberg, Germany).


be used to acquire at least a part (e.g. one projection) of the imaging dataset [4, 33, 57]. The trigger point can be optimized so that data acquisition is performed in phases of the respiratory cycle that show least motion [69]. In another study, the animal could breathe freely. Its breathing excursions were tracked by a pneumatic cushion that was rigidly attached to the animal. In a definable breathing excursion position projections were acquired. After each acquisition the animal was rotated to the next projection view [70]. Acquiring various datasets at various breathing excursion points allowed four-dimensional imaging of the thorax and lung. With a variation of the study design multiple projections at one angle were acquired and retrospectively sorted according to various breathing cycles. Intrinsic image information was used for the projection sorting, thereby not requiring an additional device to collect gating information [71]. Cardiac motion compensation can be implemented using a prospective ECG triggering [4] combined with breathing triggering in intubated and ventilated mice. Repeating at various trigger points resulted in 4D imaging of a rodent heart [67]. The benefits of lung motion compensation are obvious. Lung parenchyma, diaphragm and vessels were displayed sharper than in non-gated scans [4, 70]. Respiratory gating made the measurement of lung tumor sizes more precise because smearing caused by lung motion artefacts could be reduced or avoided. Smearing results in inaccurate size measurements [57]. ECG-gating further improved the display of lung and cardiac anatomy [4]. Soft Tissue Imaging While micro- and mini-CT provide high resolution of high contrast structures such as bone and contrast media filled vessels, the soft-tissue contrast is relatively low. As mentioned, the soft-tissue contrast is strongly disturbed by image noise and is therefore strongly depending on the applied radiation dose. To this point it is uncertain, how suited micro- and miniCT are to differentiate soft-tissue structures because the differentiation of soft tissue in mini- and micro-CT is usually not as good as in clinical scale CT scanners. However, differentiation of fat from other soft tissue structures is possible. Measurement of fat volume as well as fat distribution in a mouse disease model has been demonstrated [35]. Fat measurements open promising perspectives in imaging diabetic animal models or in studying kachexia (e.g. during neoplastic disease). Most likely, if fat can be differentiated from surrounding soft tissue, determination of muscle volume also becomes feasible and can be used to characterize musculoskeletal disorders, e.g. in transgenic mice with muscle dystrophia. Organs in the peritoneal cavity can already be differentiated in non-enhanced scans. Liver, kidney, spleen, heart, lung, stomach, adrenal gland, gut and bladder can be identified and their borders be delineated. To enhance the contrast between organs or to differenttiate tumors and organs in the peritoneal cavity, standard,


iodinated contrast media have been injected in to the peritoneal cavity, facilitating visualization of their boundaries. In contrast-enhanced flat-panel micro-CT scans also the longitudinal assessment of urethral tumors in the bladder became feasible [72], where tumors were identified as negative shapes in the contrast media filled bladder. Image Registration with Other Modalities The diagnostic value of studies can be improved by fusing and matching the data from different imaging modalities thus directly combining morphological and functional features: registration is the method to align the modalities in space, while fusion is a function that combines all or parts of the information of the modalities into a new dataset. Potential complementary imaging data for registration with micro-CT can derive from radionuclide, optical and MR imaging as well as histology. In general, registration methods can be characterized as follows: intrinsic (registration based on the image information itself) or extrinsic (registration based on additional markers or frames), retrospective (after scanning) or prospective (before scanning, e.g. through reference frames or scan beds), manual (user based) or automatic (software algorithm), rigid (no spatial distortion is allowed to register both datasets) or non-rigid (deformation is allowed). Small animal multimodality registration can be done by combining both imaging modalities in one scanner and perform simultaneous or near simultaneous data acquisitions [16, 73]. The advantage of combining two modalities in one scanner system is that there is no need to move the animal between acquisitions, meaning both scanners can use the same coordinate system, resulting in very good registration accuracies. The main disadvantage of simultaneous imaging devices is a certain loss of the flexible choice of scanner geometry. Another method is the use of a scan bed or a framework, to which the animal can be fixated. The scan bed can then be used to transfer the animal from one modality to the next insuring accurate animal positioning. This method has been suggested as the most feasible method in small animal imaging that provides a very good accuracy [74]. If multi-modality imaging acquisition is not conducted in a combined scanner or using fixed reference frames, retrospective methods such as software-based or manual methods need to be applied. Multiple software-based approaches for retrospective, automatic methods are also described [75]. Registration accuracy was in the order of 1 mm. The used software approaches still needed significant user interaction. Alternatively, marker based approaches could be chosen, here registration markers that are visible in both modalities were attached rigidly to the animal. So far radionuclide image registration with micro-CT has been published mostly. Registration between 18F-fluoride PET bone images and micro-CT of the rat skull [75], 18F-FDG PET of the rat and micro-CT (Fig. (7)) [74, 75] as well as SPECT and micro-CT [16, 74] was described.

58


Bartling et al.

Fig. (7). Micro-PET and micro-CT co-registration. Structural information of a mouse from micro-CT (A) is registered to 18F-FDG micro-PET data (B), resulting in a combination of both modalities (C). The combination image provides anatomic information on the tumor location (white arrows) and its metabolism (from Meei-Ling Jan et al., Institute of Nuclear Energy Research, Longtan, Taiwan, © 2005 IEEE).

ACKNOWLEDGEMENTS

[17]

We thank Willi Kalender and Michael Grasruck for their support by providing image material.

[18]

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11]

[12] [13] [14] [15] [16]

Ritman EL. Micro-computed tomography-current status and developments. Ann Rev Biomed Eng 2004; 6: 185-208. Boyd SK, Davison P, Muller R, Gasser JA. Monitoring individual morphological changes over time in ovariectomized rats by in vivo micro-computed tomography. Bone 2006; 39: 854-62. Calder WA. Size, function and life history. 2 ed: Dover Publications 1996. Badea C, Hedlund LW, Johnson GA. Micro-CT with respiratory and cardiac gating. Med Phys 2004; 31: 3324-9. Holdsworth DW, Thornton MM. Micro-CT in small animal and specimen imaging. Trends Biotechnol 2002; 20: 34-9. Badea C, Hedlund LW, Wheeler CT, Mai W, Johnson GA. Volumetric micro-CT system for in vivo microscopy. IEEE Int Symp Biomed Imag: Macro to Nano 2004: 1377-80. Wang G, Vannier M. Micro-CT scanners for biomedical applications: an overview. Adv Imag 2001; 16: 18-27. Knollmann F, Valencia R, Buhk JH, Obenauer S. Characteristics and applications of a flat panel computer tomography system. RoFo 2006; 178: 862-71. Paulus MJ, Gleason SS, Kennel SJ, Hunsicker PR, Johnson DK. High resolution X-ray computed tomography: an emerging tool for small animal cancer research. Neoplasia 2000; 2: 62-70. Ross W, Cody DD, Hazle JD. Design and performance characteristics of a digital flat-panel computed tomography system. Med Phys 2006; 33: 1888-901. Siewerdsen JH, Moseley DJ, Bakhtiar B, Richard S, Jaffray DA. The influence of antiscatter grids on soft-tissue detectability in cone-beam computed tomography with flat-panel detectors. Med Phys 2004; 31: 3506-20. Gupta R, Grasruck M, Suess C, et al. Ultra-high resolution flatpanel volume CT: fundamental principles, design architecture, and system characterization. Eur Radiol 2006; 16: 1191-205. Yaffe MJ, Rowlands JA. X-ray detectors for digital radiography. Phys Med Biol 1997; 42: 1-39. Feldkamp LA, Davis LC, Kress JW. Practical cone-beam algorithm. J Opt Soc Am 1984; 1: 612-9. Kohlbrenner A, Koller B, Hammerle S, Ruegsegger P. In vivo micro tomography. Adv Exp Med Biol 2001; 496: 213-24. Kastis GA, Furenlid LR, Wilson DW, Peterson TE, Barber HB, Barrett HH. Compact CT/SPECT small-animal imaging system. IEEE Trans Nucl Sci 2004; 51: 63-7.

[19] [20]

[21] [22] [23] [24] [25]

[26] [27] [28] [29] [30] [31] [32] [33]

Holdsworth DW, Drangova M, Fenster A. A high-resolution XRIIbased quantitative volume CT scanner. Med Phys 1993; 20: 44962. Lee SC, Kim HK, Chun IK, Cho MH, Lee SY, Cho MH. A flatpanel detector based micro-CT system: performance evaluation for small-animal imaging. Phys Med Biol 2003; 48: 4173-85. Kim HK, Lee SC, Chun IK, et al. Performance evaluation of a flatpanel detector-based microtomography system for small-animal imaging. IEEE Nucl Sci Symp 2003; 3: 2108-113. Kiessling F, Greschus S, Lichy MP, et al. Volumetric computed tomography (VCT): a new technology for noninvasive, highresolution monitoring of tumor angiogenesis. Nat Med 2004; 10: 1133-8. Asahina, H. Selenium-based flat panel x-ray detector for digital fluoroscopy and radiography. Visions 2001; 1: 31-40. Kalender WA. The use of flat-panel detectors for CT imaging. Radiologe 2003; 43: 379-87. Popescu S, Stierstorfer K, Flohr T, Suess C, Grasruck M. Design and evaluation of a prototype volume CT scanner. Proc SPIE 2005; 5745: 600-608. Joon Kim H, Kyung Kim H, Cho G, Choi J. Construction and characterization of an amorphous silicon flat-panel detector based on ion-shower doping process. NIMA 2003; 505: 155-8. Voelk M, Hamer OW, Feuerbach S, Strotzer M. Dose reduction in skeletal and chest radiography using a large-area flat-panel detector based on amorphous silicon and thallium-doped cesium iodide: technical background, basic image quality parameters, and review of the literature. Eur Radiol 2004; 14: 827-34. Colbeth RE, Boyce S, Fong R, et al. 40 x 30 cm flat-panel imager for angiography, R&F, and cone-beam CT applications. Proc SPIE 2001; 4320: 94-102. Siewerdsen JH, Jaffray DA. A ghost story: spatio-temporal response characteristics of an indirect-detection flat-panel imager. Med Phys 1999; 26: 1624-41. Siewerdsen JH, Jaffray DA. Cone-beam computed tomography with a flat-panel imager: effects of image lag. Med Phys 1999; 26: 2635-47. Roos PG, Colbeth RE, Mollov I, et al. Multiple-gain-ranging readout method to extend the dynamic range of amorphous silicon flat-panel imagers. Proc SPIE 2004; 5368: 139. Khodaverdi M, Pauly F, Weber S, et al. Preliminary studies of a micro-CT for a combined small animal PET/CT scanner. IEEE Nucl Sci Symp 2001; 3. Grasruck M, Gupta R, Reichardt B, et al. Combination of CT scanning and fluoroscopy imaging on a flat-panel CT scanner. Proc SPIE 2006; 875-882 Motz JW, Danos M. Image information content and patient exposure. Med Phys 1978; 5: 8-22. Badea CT, Fubara B, Hedlund LW, Johnson GA. 4-D micro-CT of the mouse heart. Mol Imaging 2005; 4: 110-6.

Small Animal Computed Tomography Imaging [34] [35] [36]

[37] [38] [39] [40]

[41] [42]

[43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56]


Kalender WA. Computed tomography: fundamentals, system technology, image quality, applications. Munich, VCH Verlagsgesellschaft 2005. Paulus MJ, Gleason SS, Easterly ME, Foltz CJ. A review of highresolution X-ray computed tomography and other imaging modalities for small animal research. Lab Anim 2001; 30: 36-45. Ford NL, Graham KC, Groom AC, Macdonald IC, Chambers AF, Holdsworth DW. Time-course characterization of the computed tomography contrast enhancement of an iodinated blood-pool contrast agent in mice using a volumetric flat-panel equipped computed tomography scanner. Invest Radiol 2006; 41: 384-90. Vera DR, Mattrey RF. A molecular CT blood pool contrast agent. Acad Radiol 2002; 9: 784-92. Kao CY, Hoffman EA, Beck KC, Bellamkonda RV, Annapragada AV. Long-residence-time nano-scale liposomal iohexol for X-raybased blood pool imaging. Acad Radiol 2003; 10: 475-83. Torchilin VP, Frank-Kamenetsky MD, Wolf GL. CT visualization of blood pool in rats by using long-circulating, iodine-containing micelles. Acad Radiol 1999; 6: 61-5. Bakan DA, Doerr-Stevens JK, Weichert JP, Longino MA, Lee FT, Counsell RE. Imaging efficacy of a hepatocyte-selective polyiodinated triglyceride for contrast-enhanced computed tomography. Am J Ther 2001; 8: 359-65. Weichert JP, Lee FT, Chosy SG, et al. Combined hepatocyteselective and blood-pool contrast agents for the CT detection of experimental liver tumors in rabbits. Radiology 2000; 216: 865-71. Rabin O, Manuel Perez J, Grimm J, Wojtkiewicz G, Weissleder R. An X-ray computed tomography imaging agent based on longcirculating bismuth sulphide nanoparticles. Nat Mater 2006; 5: 118-22. Ketai LH, Muggenberg BA, McIntire GL, et al. CT imaging of intrathoracic lymph nodes in dogs with bronchoscopically administered iodinated nanoparticles. Acad Radiol 1999; 6: 49-54. Ford NL, Thornton MM, Holdsworth DW. Fundamental image quality limits for microcomputed tomography in small animals. Med Phys 2003; 30: 2869-77. Boone JM, Velazquez O, Cherry SR. Small-animal x-ray dose from micro-CT. Mol Imaging 2004; 3: 149-58. Goertzen AL, Meadors AK, Silverman RW, Cherry SR. Simultaneous molecular and anatomical imaging of the mouse in vivo. Phys Med Biol 2002; 47: 4315-28. Sato F, Sasaki N, Kawashima N, Chino F. Late effects of whole or partial body x-irradiation on mice: life shortening. Int J Radiat Biol Relat Stud Phys Chem Med 1981; 39: 607-15. Mole RH. Quantitative observations on recovery from whole body irradiation in mice II. Recovery during and after daily irradiation. Br J Radiol 1957; 30: 40-6. Kohn HI, Kallmann RF. The influence of strain on acute x-ray lethality in the mouse I. LD50 and death rate studies. Radiat Res 1956; 5: 309-17. Taschereau R, Chow PL, Chatziioannou AF. Monte Carlo simulations of dose from microCT imaging procedures in a realistic mouse phantom. Med Phys 2006; 33: 216-224. Waarsing JH, Day JS, Verhaar JA, Ederveen AG, Weinans H. Bone loss dynamics result in trabecular alignment in aging and ovariectomized rats. J Orthop Res 2006; 24: 926-35. Greschus S, Kiessling F, Lichy MP, et al. Potential applications of flat-panel volumetric CT in morphologic and functional small animal imaging. Neoplasia 2005; 7: 730-40. Ritman EL. Molecular imaging in small animals--roles for microCT. J Cell Biochem Suppl 2002; 39: 116-24. Ritman EL. Micro-computed tomography of the lungs and pulmonary-vascular system. Proc Am Thorac Soc 2005; 2: 477-80. De Clerck NM, Meurrens K, Weiler H, et al. High-resolution x-ray microtomography for the detection of lung tumors in living mice. Neoplasia 2004; 6: 374-9. Kennel SJ, Davis IA, Branning J, Pan H, Kabalka GW, Paulus MJ. High resolution computed tomography and MRI for monitoring

Received: October 27, 2006

[57] [58]

[59] [60] [61]

[62]

[63] [64] [65] [66] [67] [68]

[69] [70] [71]

[72] [73] [74] [75] [76]

lung tumor growth in mice undergoing radioimmunotherapy: correlation with histology. Med Phys 2000; 27: 1101-7. Cody DD, Nelson CL, Bradley WM, et al. Murine lung tumor measurement using respiratory-gated micro-computed tomography. Invest Radiol 2005; 40: 263-9. Li XF, Zanzonico P, Ling CC, O'Donoghue J. Visualization of experimental lung and bone metastases in live nude mice by x-ray micro-computed tomography. Technol Cancer Res Treat 2006; 5: 147-55. Postnov AA, Meurrens K, Weiler H, et al. In vivo assessment of emphysema in mice by high resolution x-ray microtomography. J Microsc 2005; 220: 70-5. Plathow C, Li M, Gong P, et al. Computed tomography monitoring of radiation-induced lung fibrosis in mice. Invest Radiol 2004; 39: 600-9. Tamada T, Sone T, Jo Y, Imai S, Kajihara Y, Fukunaga M. Threedimensional trabecular bone architecture of the lumbar spine in bone metastasis from prostate cancer: comparison with degenerative sclerosis. Skeletal Radiol 2005; 34: 149-55. David V, Laroche N, Boudignon B, et al. Noninvasive in vivo monitoring of bone architecture alterations in hindlimb-unloaded female rats using novel three-dimensional microcomputed tomography. J Bone Miner Res 2003; 18: 1622-31. Waarsing JH, Day JS, Weinans H. Longitudinal micro-CT scans to evaluate bone architecture. J Musculoskelet Neuronal Interact 2005; 5: 310-2. Guldberg RE, Lin AS, Coleman R, Robertson G, Duvall C. Microcomputed tomography imaging of skeletal development and growth. Birth Defects Res C Embryo Today 2004; 72: 250-9. Jorgensen SM, Demirkaya O, Ritman EL. Three-dimensional imaging of vasculature and parenchyma in intact rodent organs with x-ray micro-CT. Am J Physiol 1998; 275: 1103-14. Mukundan S, Ghaghada KB, Badea CT, et al. A liposomal nanoscale contrast agent for preclinical CT in mice. Am J Roentgenol 2006; 186: 300-7. Badea CT, Bucholz E, Hedlund LW, Rockman HA, Johnson GA. Imaging methods for morphological and functional phenotyping of the rodent heart. Toxicol Pathol 2006; 34: 111-7. Mai W, Badea CT, Wheeler CT, Hedlund LW, Johnson GA. Effects of breathing and cardiac motion on spatial resolution in the microscopic imaging of rodents. Magn Reson Med 2005; 53: 85865. Walters EB, Panda K, Bankson JA, Brown E, Cody DD. Improved method of in vivo respiratory-gated micro-CT imaging. Phys Med Biol 2004; 49: 4163-72. Ford NL, Nikolov HN, Norley CJ, et al. Prospective respiratorygated micro-CT of free breathing rodents. Med Phys 2005; 32: 2888-98. Hu J, Haworth ST, Molthen RC, Dawson CA. Dynamic small animal lung imaging via a postacquisition respiratory gating technique using micro-cone beam computed tomography. Acad Radiol 2004; 11: 961-70. Johnson AM, Conover DL, Huang J, et al. Early detection and measurement of urothelial tumors in mice. Urology 2006; 67: 1309-14. Meei-Ling J, Yu-Ching N, Kou-Wei C, et al . A combined microPET/CT scanner for small animal imaging. NIM A 2006; in press. Jan ML, Chuang KS, Chen GW, et al. A three-dimensional registration method for automated fusion of micro PET-CT-SPECT whole-body images. IEEE Trans Med Imaging 2005; 24: 886-93. Vaquero J, Desco M, Pascau J, et al. PET, CT and MR image registration of the rat brain and skull. IEEE Trans Nucl Sci 2001; 48: 1440-5. Bauerle T, Adwan H, Kiessling F, Hilbig H, Armbruster FP, Berger MR. Characterization of a rat model with site-specific bone metastasis induced by MDA-MB-231 breast cancer cells and its application to the effects of an antibody against bone sialoprotein. Int J Cancer 2005; 115: 177-86.

Revised: December 13, 2006

Accepted: December 14, 2006