|Home | About | Journals | Submit | Contact Us | Français|
Diffusion Tensor imaging (DTI) is an emerging noninvasive method for evaluating tissue microstructure, but is highly susceptible to in vivo motion artifact. Ex vivo experiments on fixed tissues are needed to improve DTI techniques, which require fixed tissue specimens. Several efforts have been made to study the effect of fixation on both human and mouse tissue, with varying results. Four human cervical cords and three segments of pig vivo cervical spinal cord specimens were imaged both before and after tissue fixation using 3D multi-shot diffusion weighted imaging (ms-DWEPI). Fixation caused a significant decrease in the longitudinal diffusivity while the relative anisotropy (RA), and radial diffusivity remained unaffected. Additionally, once adequately preserved the diffusivity parameters of fixed tissue remain constant over time. Fixation has important effects on the diffusivity of tissue specimens. These findings have important implications for the determination of tissue microstructure and function using DTI technologies.
Diffusion Tensor imaging (DTI) is emerging as an important method for evaluating the microscopic anatomy of central nervous system using noninvasive imaging techniques (Horsfield and Jones, 2002; Budde et al., 2007; Ciccarelli et al., 2007). Diffusion weighted imaging has already proven to be an important diagnostic tool in the identification of acute ischemia. By applying a magnetic gradient in a given direction, restricted water diffusion in areas of cytoxic edema can be readily identified. DTI goes a step further by measuring both the magnitude and direction of diffusion of water in multiple directions. By measuring multiple vectors of diffusion, DTI can characterize the microscopic anatomy of white matter tracts within the brain and spinal cord. Water molecules within these highly organized tissues with elongated morphology, preferentially move in direction parallel to the axonal orientation rather than perpendicular to axons. This property of directional diffusion is referred to as anisotropy. By measuring movement of water within neurons, DTI can provide a highly sensitive method of characterizing cellular microstructure. DTI techniques can measure and map out the diffusion of water along white matter tracts within the columns of the spinal cord, and derive information on concerning the integrity of both myelin sheaths and axons. The measured longitudinal diffusion is thought to be a marker for axonal disease, while the diffusion perpendicular to axonal tracts can provide information on the integrity of myelin sheaths (Song et al., 2002). In normal white matter, diffusion should be greater in the longitudinal direction than in the perpendicular direction. With common neurodegenerative diseases such as multiple sclerosis, both myelin and axonal integrity are affected, leading to alterations in longitudinal and radial diffusion values.
However given the sensitivity of DTI imaging parameters to motion, the presence cardiac activity, respiratory motion, and CSF pulsation in vivo has limited its diagnostic accuracy. Furthermore, the size and location of spinal cord within the spinal canal also introduces difficulties with spatial resolution, and geometric distortion artifacts from the surrounding vertebral column (Holder et al., 2000; Ries et al., 2000; Wheeler-Kingshott et al., 2002; Ozanne et al., 2007; Valsasine et al., 2007; Wilm et al., 2007; Rossi et al., 2008). For these reasons, many ex vivo DTI imaging experiments have been conducted, on both animal and human tissue specimens, with the hope that information would lead to improved in vivo techniques for evaluating tissue microstructure (Nijeholt et al., 2001; Chen et al., 2003; Bot et al., 2004; Schwartz et al., 2005; Sun et al., 2005; Kroenke et al., 2006; D'Arceuil and Crespigny, 2007; D'Arceuil et al., 2007; Wu et al., 2007; Schmierer et al., 2008).
There is a great reliance on archived tissue specimens in order to conduct such MRI experiments. Immersion fixation of specimens with formaldehyde is common practice to maintain tissue integrity during imaging experiments and prior to histological analyses. Several efforts have been made to study the effect of fixation on DTI parameters using mouse tissue, with varying results (Kim et al., 2007). Few studies have been conducted on human tissue specimens (Nijeholt et al., 2001; Schmierer et al., 2008). It had been previously accepted that there is no change in longitudinal and radial diffusion values when comparing in vivo state to ex vivo fixed tissues using mouse models (Sun et al., 2003; Kim et al., 2007). However Schmiere et al. recently published data from experiments performed on human brain specimens which have important implications on both previous and future DTI experiments performed on fixed tissue (Madi et al., 2005; Schmierer et al., 2008). They demonstrated a change in both longitudinal and radial diffusion when comparing the values in ex vivo human brain sections both before and after immersion fixation (Schmierer et al., 2008) suggesting the tissue microstructure is altered.
More importantly no studies to date have evaluated the long term microstructural effects of fixation on human specimens and how it relates to measured diffusivity values. Kim et al. has previously reported the relative anisotropy decreases overtime using a mouse model comparing in situ measurement to ex vivo fixed tissues at two and fifteen weeks (Kim et al., 2007).
The results of this study emerged as part of a larger project designed to evaluate the reliability of the measured diffusion values in there ability to characterize tissue microstructure compared to histological analyses. The goal of the experiments was to use the data from ex vivo specimens to improve future functional imaging experiments on the human nervous system performed in vivo. In the present study, we 1) compared DTI indices in ex vivo tissues both before and after fixation to closely emulate common postmortem delays in fixation of autopsy specimens, and 2) evaluated the long term effects of tissue fixation on the common DTI indices and on tissue microstructure.
In this report, we present the DTI results of human and pig cervical spinal cord (CSC) using 3D ms-DWEPI on ex-vivo CSC specimens before and after fixation with temperature modulation. DTI experiments were conducted on four human CSC specimens, and three pig CSC segments before and after fixation. The pig specimens were scanned repeatedly over eight weeks and again at 16 weeks post fixation.
Four unfixed postmortem human cervical spinal cords approximately 5 cm in length were provided by the Human Brain and Spinal Fluid Resource center, the National Neurological Research Specimen Bank in Los Angeles, California. The patients were women with a known history of Alzheimer's disease, but no reported spinal cord pathology. The average age of the patient's was 83 years (SD=2.5; range 80-87 years) at the time of death. The human CSCs were initially flash frozen a mean of 15h (SD=6.1 h; range 8-21h) after death. Laminectomies were performed from the second cervical spinal level (C2) through C7. The meninges and spinal roots were excised en bloc. The cords were stored at -80° C, and subsequently thawed over 48 hours; the first 24 hours at -20° C and the second 24 hours at 4° C prior to imaging. The cords were thawed in this fashion in effort to protect specimens from possible microstructural damage that could theoretically occur with abrupt temperature change. All the human CSC specimens were scanned twice: first under unfixed conditions at 4 °C, and the second scan following 2-3 days after being immersed in a 4 °C, 4% Paraformaldehyde and phosphate-buffered saline (PBS) solution for fixation. The human CSCs were initially acquired with the intention of performing additional histological analyses, not reported in this document. Therefore careful temperature modulation was employed to minimize factors that could contribute to autolysis of the tissue. The second scan was also performed while maintaining a temperature of 4 °C which allowed comparison of diffusivity indices before and after fixation with other conditions held constant.
Three segments of CSC were obtained from a pig euthanized using intravenous sodium pentobarbital (100 mg/kg; Abbot Laboratories, North Chicago, Illinois). Animal studies had been approved by institutional animal care and use committee (IACUC). The pig CSC specimens were scanned a total of six times. The first experiment was done on unfixed specimens. The second scan was conducted one week following fixation, using the same methods described above for the human samples (2-3 days of immersion in a chilled 4% Paraformaldehyde and PBS). The fixed pig cords were subsequently two, three, four, eight and sixteen weeks after completing the initial fixation. Unlike the human samples, pig samples were scanned at the room temperature (~25 °C) as the samples were acquired for analysis of the effects of fixation specifically, and not for further histological analyses. The initial autolysis time was approximately four hours.
The temperature of the specimens was maintained and monitored during the imaging experiments using a constructed temperature-control-device. It consisted of a Styrofoam box containing an ice bath cooled with liquid nitrogen. The liquid nitrogen was sequestered from a specialized quadrature rf-coil which was built for the experiments. The coil is 2 inches in diameter and 3 inches in length. All of the samples were placed in a rectangle shaped acryl box not only to maintain moisture but also to help localize the cords within the coil during the MRI experiment. The coil and specimens were placed within the Styrofoam temperature-control-device during the imaging.
T2 weighted images using 3D Turbo Spin Echo (TSE) have been employed to verify the pathology of specimens. For TSE imaging, typical MR imaging parameters were TR 4000 ms, TE 81 ms, (0.5 mm)3 isotropic spatial resolution, echotrain length (ETL) 11, and receive bandwidth 255 Hz/pixel, respectively.
A 3D ms-DWEPI was developed based on a segmented spin-echo EPI (Fig. 1), using Siemens pulse sequence development software (IDEA). 3D ms-DWEPI was applied on both the human and pig CSC specimens using a 3T whole-body MRI system (Trio, Siemens Medical Solution, Erlangen, Germany) with Avanto gradients (45 mT/m strength and 150 T/m/s slew rate), and a home-built quadrature rf-coil. Typical MR imaging parameters were TR 400 ms, (0.5 mm)3 isotropic spatial resolution, b of 1000 s/mm2 in 20 non-collinear directions, echotrain length (ETL) 5, and receiver bandwidth of 500 Hz/pixel. Imaging time for 3D ms-DWEPI was 30 minutes.
The Stejskal-Tanner diffusion-weighting gradient scheme was used for diffusion weighted imaging measurements (Stejskal and Tanner, 1965). In this scheme, a pair of diffusion gradients are positioned before and after 180° rf-pulse for de-phasing and re-phasing spins. In a diffusion-weighted spin-echo MR measurement, the signal intensity (S) of the diffusion weighted images can be expressed with diffusion coefficient, D, and gradient attenuation factor, b in eq. (1):
where So is the non-diffusion weighted signal, TE is the echo time and T2 is the spin-spin relaxation time. Here, b is the amount of diffusion weighting which is presented in eq. (2).
γ is the gyromagnetic ratio, and Δ, GD and δ are the duration, amplitude, and separating time of diffusion gradients, respectively. Diffusion coefficient, D for anisotropic system is expressed in a 3×3 symmetric tensor matrix (Dxx, Dxy, Dxz, Dxy, Dyy, Dyz, Dxz, Dyz, Dzz).
The acquired diffusion tensor imaging (DTI) data set was processed in pixel-by-pixel manner using a home-made DTI analysis software written with interactive data language (IDL). (ITT Visual Information Solutions Inc., Boulder, CO) The apparent diffusion coefficient (ADC) was first calculated using two images with b=0 and b=1000 s/mm2 for all 20 diffusion encoding directions. These ADC values were used to extract the 3×3 diffusion matrix using the singular value decomposition (SDV), which was then diagonalized to obtain the three rotationally invariant eigenvalues (diffusivities) and the three corresponding eigenvectors. Multiple diffusivity measurements that can be derived include relative anisotropy (RA), fractional anisotropy (FA), trace diffusivity (TrD), longitudinal diffusivity (LongD), and radial diffusivity (RadD). From the obtained directional diffusivities, LongD (λ) and RadD (λ) are defined in eq. (3).
where <λ> is the mean diffusivity.
FA and RA are nearly equivalent measurements that describe what fraction of the water molecules are moving in a given direction. In previous studies, various anisotropy indices have been introduced and developed to help in the understanding of CSC anisotropic properties, for instance, the ratios of the principal diffusivities such as (λ1/λ3) and (λ/λ), which are intuitive and simple indices (Pierpaoli and Basser, 1996).
The regions of interest were selected using a combination of the imaging from the T2 weighted data, color (red-green-blue-RGB) map, FA map, and RA map using coronal plane images. The T2 weighted images were evaluated for the presence of lesions by our radiologist (Dr. Zollinger). The RGB.map indicates the direction of the principle eigenvectors: blue represents longitudinal axis, green the left-to-right, and red anterior-to-posterior. Two ROI's were selected from each human CSC specimen, and from each of the three pig CSC segments. The measured radial diffusivity, longitudinal diffusivity, and the ratios of longitudinal to radial diffusivities indices were obtained using the RA map. The ROIs were manually delineated within the ventrolateral white matter from the RA map data set, as shown in figure 4. The coronal plane was utilized in order to reduce possible volume averaging error from errantly selecting adjacent gray matter, and to maximize the size of the analyzed area in figure 4
Quantitative statistical comparison of anisotropic parameters from both the human CSC and pig CSC segments specimens was conducted using the paired Student t-test and analysis of variance (ANOVA) (SPSS version 15: SPSS, Chicago, IL, USA). Analysis was conducted on tissue both before and after fixation.
High resolution T2 weighted images from the human CSC are displayed in figure 2. There were no areas of increased signal intensity within any of the cord specimens, both human and pig. The FA map generated is scaled by color, with the grey matter demonstrating a dark blue color which indicates a low FA value in figure 3 (a). While the white matter had a FA value closer to 1, therefore a bright green color. Again the FA data confirmed that there was no pathology within the white matter tracts, as there were no areas of color gradation. The RGB map was consistent with the data from the FA map and T2 weighted images, with the principle eigenvector (blue) aligned with the longitudinal axis of the spinal cord representing the white matter tracts in figure 3 (b) The grey matter had a predominantly green color designation consistent with decussating fibers, and neuronal synapses. The vector plot showed the direction of the principle eigenvector which is parallel to the longitudinal axis.
The signal-to-noise ratio (SNR) generated using the ROI-based analysis of b=0 non-diffusion imaging remained fairly constant from 40 to 60 for regions of white matter within both human and pig CSC specimens before and after fixation (table 1). Prior mouse models maintained a SNR of 45 after utilizing a signal-noise calibration using Rayleigh statistic which multiplies the standard deviation of the background noise by 0.66 (Kim et al., 2007). Pierpaoli and Basser reported that values of common anisotropic parameters of diffusion such as the longitudinal diffusivity, radial diffusivity, and the subsequent ratio of the two diffusivities can radically change in low SNR ranges based with their Monte Carlo simulation (Pierpaoli and Basser, 1996). As the SNR decreases below the level of 20, measured anisotropy values can be falsely high. Unlike FA and RA, the λ, λ, λ/λ remain constant in a system with SNR of >20 (Pierpaoli and Basser, 1996). We were able to maintain a range of SNR between 40~60 using our specialized quadrature RF coil.
Fixation caused a significant decreased in the longitudinal diffusivity for both human and pig specimens in our experiments. The RA, radial diffusivity and the subsequent ratio of longitudinal to radial diffusivity remained unaffected or were not significantly changed by fixation (tables 2 and and3).3). We observed in our study, 16% change in radial diffusivity (P=0.158) and 24 % difference (P=0.001) in the longitudinal diffusivity from the human CSC specimens pre and post fixation, respectively (table 3). The LongD of the pig cords decreased by 25 % or from 0.83 to 0.62 (10-3 mm2/s) with fixation. The reason for this drop is not well understood, but may be secondary to structural changes that occur with molecular cross-linking during the process of fixation. Our findings were consistent to those reported by Schmierer et al. who also evaluated ex vivo specimens before and after fixation.
To further examine the long term effects of fixation on diffusivity indices (λ, λ, λ/λ), the pig CSC segments were scanned multiple times over sixteen weeks under constant temperature conditions. The λ, λ and λ/λ remained constant over time as displayed in figure 5.
With current limitations in performing diffusion tensor imaging evaluation of the spinal cord in vivo, the importance of utilizing human autopsy and archived tissue specimens is vital for future research. Archived tissue specimens are an invaluable resource for the development of new in vivo MRI techniques and for intensive study of human anatomy in both health and with disease. Because there is often a delay in the postmortem period for the acquisition and fixation of tissue specimens, understanding the changes in tissue microstructure, before and after fixation has important implications. This study emerged in order to define the effects of fixation, both immediate and long term, on DTI parameters used to characterize tissue microstructure in fixed tissues.
The first finding of this study is that the regional cellular microstructure changes as a result of fixation, as verified by changes in the measured diffusivity indices. Prior studies conducted on mouse specimens demonstrated that there was no change in the LongD between in situ after death before fixation and ex vivo immersion fixed 2 week after death reported by Kim et al. in table 3 (Kim et al., 2007). The exact reasons for this discrepancy are unclear, as the mechanics of tissue fixation have also not been fully elucidated. The differences in diffusion may be attributed to basic anatomical differences. Sun et al. used mouse brains in their experiments and found that the diffusivity indices remain stable between unfixed and fixed state. Our study utilized both human and pig spinal cords. An additional potential cause for measured diffusion differences could be related to axon density and susceptibility in the different regions of the central nervous system evaluated. Deluca et al. postulated that neurons with different axonal diameters may be more susceptible to injury within multiple sclerosis (MS) lesions (DeLuca et al., 2006). Kim et al. used mouse thoracic and lumbar cord sections, and compared in situ to ex vivo post fixation states, while our experiment focused on the cervical cord. The change of radial diffusivity indices discussed by Schmierer et al. may be related to the anatomical differences of between the white matter within the cerebral cortex compared to the spinal cord.
Though fixation was found to have immediate effects on longitudinal diffusivity, we also demonstrated that once the specimen is adequately preserved, there is no significant change in the RA, longitudinal diffusivity, radial diffusivity, or the ratio of longitudinal to radial diffusivity over time. The principal results of the experiments conducted on the pig specimens show that the diffusivity parameters of fixed tissue remain constant over time as shown in figure 5. This is in contradistinction to Kim et al. who reported that the RA and longitudinal diffusivity decreases over time in fixed tissues of their mouse model, while radial diffusivities remain the same (Kim et al., 2007). The mouse study utilized data from two time periods; two and fifteen weeks after fixation, which may not be sufficient to determine a trend. However, we used five data points over sixteen weeks and studied structural change of specimens over time. Again this point has important implications for research conducted on cervical spinal cord specimens, as it supports the practice of repeated imaging experiments over time, without having a significant effect on the measured anisotropic indices. We believe this finding will be beneficial to all those studying neurological diseases such as multiple sclerosis (MS) which target the cervical spinal cord.
An important future direction in evaluating the reliability of anisotropic indices over time and with fixation, are the effects of temperature. Kim et al. performed their in vivo experiments at approximately 37 °C, and the ex vivo experiment at approximately 20 °C. In general, the diffusivity of water increases proportionally by ~2 % per 1° C increase (Bihan et al., 1989). Without temperature control, diffusivity and RA values may be biased. We observed reduced water diffusivity in our controlled temperature (4 °C) experiments. Constant temperatures were maintained throughout the imaging experiments, human cords at 4 °C and the pig specimens at 25 °C.
In conclusion, we found that fixation caused a significant structural changes causing a decrease in the longitudinal diffusivity while the relative anisotropy (RA), and structure in radial direction (radial diffusivity) remained unaffected. Therefore when considering measured values of LongD, in ex vivo specimens, the effects of fixation on these values must be considered before assumptions about tissue microstructure can be made. Additionally, once adequately preserved the cellular microstructure and the diffusivity parameters of fixed tissue remain constant over time.
This work was supported by National Multiple Sclerosis Society, the Cumming Foundation, the Benning Foundation, and NIH grants R21NS052424 and R21EB005705. We thank to Dr. Brian T. Saam in Physics at University of Utah for his technical support on the temperature controlling system.
Grant sponsors: National Multiple Sclerosis Society, the Cumming Foundation, the Benning Foundation, and NIH grants R21NS052424 and R21EB005705