- Email: [email protected]
Diffusion tensor imaging of peripheral nerves in non-fixed post-mortem subjects
Forensic Science International 263 (2016) 139–146
Contents lists available at ScienceDirect
Forensic Science International journal homepage: www.elsevier.com/locate/forsciint
Diffusion tensor imaging of peripheral nerves in non-fixed post-mortem subjects Wieke Haakma a,b,c,*, Michael Pedersen b,d, Martijn Froeling c, Lars Uhrenholt a, Alexander Leemans e, Lene Warner Thorup Boel a a
Department of Forensic Medicine, Aarhus University, Aarhus, Denmark Comparative Medicine Lab, Department of Clinical Medicine, Aarhus University, Aarhus, Denmark Department of Radiology, University Medical Center Utrecht, Utrecht, The Netherlands d MR Research Center, Department of Clinical Medicine, Aarhus University, Aarhus, Denmark e Image Sciences Institute, University Medical Center Utrecht, Utrecht, The Netherlands b c
A R T I C L E I N F O
A B S T R A C T
Article history: Received 26 November 2015 Received in revised form 10 February 2016 Accepted 1 April 2016 Available online 11 April 2016
Purpose: While standard magnetic resonance imaging (MRI) sequences are increasingly employed in post-mortem (PM) examinations, more advanced techniques such as diffusion tensor imaging (DTI) remain unexplored in forensic sciences. Therefore, we studied the temporal stability and reproducibility of DTI and fiber tractography (FT) in non-fixed PM subjects. In addition, we investigated the lumbosacral nerves with PMDTI and compared their tissue characteristics to in vivo findings. Methods: MRI data were acquired on a 1.5 T MRI scanner in seven PM subjects, consisting of six nontrauma deaths and one chronic trauma death, and in six living subjects. Inter-scan (within one session) and inter-session (between days) reproducibility of diffusion parameters, fractional anisotropy (FA), and mean diffusivity (MD), were evaluated for the lumbosacral nerves using Bland–Altman and Jones plots. Diffusion parameters in nerves L3–S2 were compared to living subjects using the non-parametric Mann– Whitney U test. Results: Reproducibility of diffusion values of inter-scan 95% limits of agreement ranged from 0.058 to 0.062 for FA, and ( 0.037 to 0.052) 10 3 mm2/s for MD. For the inter-session this was 0.0423 to 0.0423, and ( 0.0442 to 0.0442) 10 3 mm2/s for FA, and MD, respectively. Although PM subjects showed approximately four-fold lower diffusivity values compared to living subjects, FT results were comparable. The chronic trauma case showed disorganization and asymmetry of the nerves. Conclusion: We demonstrated that DTI was reproducible in characterizing nervous tissue properties and FT in reconstructing the architecture of lumbosacral nerves in PM subjects. We showed differences in diffusion values between PM and in vivo and showed the ability of PMDTI and FT to reconstruct nerve lesions in a chronic trauma case. We expect that PMDTI and FT may become valuable in identification and documentation of PM nerve trauma or pathologies in forensic sciences. ß 2016 Elsevier Ireland Ltd. All rights reserved.
Keywords: Post-mortem imaging Magnetic resonance imaging Diffusion tensor imaging Fiber tractography Peripheral nerves
1. Introduction Imaging techniques are becoming increasingly important in forensic examinations and may contribute to, or partially replace,
Abbreviations: AD, axial diffusivity; DTI, diffusion tensor imaging; EPI, echo planar imaging; FA, fractional anisotropy; FOV, field of view; FT, fiber tractography; MD, mean diffusivity; MRI, magnetic resonance imaging; PM, post-mortem; PMDTI, post-mortem diffusion tensor imaging; PMMRI, post-mortem magnetic resonance imaging; RD, radial diffusivity; TSE, turbo spin echo. * Corresponding author at: Department of Forensic Medicine, Aarhus University, Palle Juul-Jensens Boulevard 99, 8200 Aarhus N, Denmark. Tel.: +45 8716 7500. E-mail address: [email protected] (W. Haakma). http://dx.doi.org/10.1016/j.forsciint.2016.04.001 0379-0738/ß 2016 Elsevier Ireland Ltd. All rights reserved.
conventional autopsy [1–4]. With computed tomography (CT) it is possible to identify fractures, pathologic gas development, large hemorrhages, and metal or other foreign bodies [5–9]. Magnetic resonance imaging (MRI) on the other hand is particularly valuable when soft tissue injuries or organ trauma need to be identified [5,8,10]. Regarding some of the more advanced MRI techniques, such as diffusion tensor imaging (DTI) [11–14], their potential value in forensic sciences remains unknown. DTI is an MRI technique which can measure the random movement of water molecules, known as Brownian motion [11]. In nervous tissue this movement is more pronounced along the nerve, so-called anisotropy [12]. Especially in the identification of peripheral nerves, DTI might be valuable, since nerves are difficult to dissect
W. Haakma et al. / Forensic Science International 263 (2016) 139–146
140
during conventional autopsy and are therefore typically neglected in forensic pathology. DTI can be used to characterize the structure of nervous tissue in detail and microstructural properties can be evaluated. DTI may serve as a valuable tool in investigations of accidents, homicides and other traumatic cases. Important differences between post-mortem (PM) and living tissue should be considered. Shortly after death, processes such as autolysis occur. Combined with bacterial degradation, which facilitates tissue decomposition, nervous tissue will deteriorate [15]. Therefore, PMDTI has been performed predominantly on formaldehyde fixed tissue [16]. However, molecular diffusion characteristics are known to change markedly following the fixation process due to formation of intra- and inter-molecular cross-links [17]. Moreover, temperature regulation of PM subjects can be challenging during PMMRI, which can influence DTI results since water diffusion is highly temperature sensitive [18]. Previous PMDTI studies have focused on fixed human brains and demonstrated preserved diffusion anisotropy in nerves [15,17,19], allowing for fiber tractography (FT) and quantitative investigation of diffusion parameters such as fractional anisotropy (FA), which represents the degree in which diffusion is oriented in one direction, and mean diffusivity (MD), which is the average of all eigenvalues (the magnitude of diffusion in a direction). The FA and MD are based on the axial diffusivity (AD), which is the largest of the eigenvalues and in healthy nervous tissue diffusion is mostly represented along the nerve, and radial diffusivity (RD), which is the average of the second and third eigenvalue and in healthy nervous tissue diffusion is perpendicular to the nerve. Only a few studies have employed PMDTI in non-fixed human tissue, including the brain [20–23], brain trauma [24] and cardiac infarction [25]. The aim of this study was to demonstrate that PMDTI can be used for identification and quantification of peripheral nervous tissue in a reproducible way and to identify peripheral nerve trauma. Specifically, we investigated the lumbosacral nerves with PMDTI, because they already have been described in detail in several in vivo DTI studies [26–29], and we compared their tissue characteristics to in vivo findings. In addition, we studied the temporal stability and reproducibility of DTI and FT in non-fixed PM subjects.
of 46 years, range 30–55 years all non-trauma deaths, and one female trauma death; 35 years). PMMRI was performed 1–8 days after the estimated date of death. Specific subject information is listed in Table 1. Acquired data were included in four experiments (described below); data of subject 1–5 were included in experiment 1, data of subject 6 were included in experiment 2, data of subject 1– 5 were included in experiment 3, and data of subject 7 were included in experiment 4. All subjects were cooled in the morgue at 4 8C before and after autopsy without fixative solution preparation. The routine autopsy was performed before PMMRI (except for experiment 2) and did not affect the lumbosacral area. Six healthy living subjects (mean age of 30 years, range 25–42 years) were included as reference subjects for experiment 3. 2.2. Data acquisition
2. Materials and methods
MRI of the PM subjects was performed on a 1.5 T MRI scanner (Achieva; Philips Healthcare, Best, The Netherlands) using a 16-channel phased-array surface coil. PMDTI was performed with diffusion-weighted spin echo single-shot echo planar imaging (EPI) using the following parameters: TE = 82 ms, TR = 13,538 ms, SENSE factor = 2, number of excitations = 8, field of view (FOV) = 384 mm 216 mm, matrix size = 128 72, 72, slice thickness = 3.0 mm, resulting in a voxel size of 3.0 mm 3.0 mm 3.0 mm, EPI factor = 35, b-values of 0 and 2000 s/mm2, and 15 gradient directions. For three PM subjects the protocol was repeated two times within the scan session, and for four PM subjects it was repeated four times within the scan session, resulting in total scan times of 58 min and 1 h 56 min, respectively. MRI of the six living subjects was performed on a 3 T MRI scanner (Achieva; Philips Healthcare, Best, The Netherlands) using a 16channel phased-array surface coil. DTI was performed with diffusion-weighted spin echo single-shot EPI using the following parameters: TE = 58 ms, TR = 4209 ms, SENSE factor = 2, number of excitations = 2, FOV = 336 mm 336 mm, matrix size = 112 112, slice thickness = 3.0 mm, resulting in a voxel size of 3.0 mm 3.0 mm 3.0 mm, EPI factor = 35, b-values of 0 and 800 s/mm2, and 15 gradient directions. The total acquisition time was 4.5 min. For anatomical reference, a 3D turbo spin echo (TSE) sequence was acquired based on the protocol described by [29] for both PM and living subjects.
2.1. Subjects
2.3. Data analysis
In this study, seven non-fixed PM subjects with normal anatomy of the lower spine were included (five male, one female; median age
DTI data from the PM and living subjects were processed identically, using the ExploreDTI diffusion MRI toolbox
Table 1 Subject specifications. Identification number
Sex
Age in years
Approximate number of days between death and scan
Number of scans within one session
History
Cause of death
1 2 3 4
Male Male Male Male
55 46 38 30
3 1 5 5
2 2 4 4
Alcohol abuse Depression, posttraumatic stress disorder Drug abuse Schizophrenia
5 6
Male Female
46 49
3 5
4 4
Drug abuse Schizophrenia, drug and alcohol abuse
Cardiac death Gunshot through head Overdose methadone Overdose of potassium tablets General physical decline Pancreatic cancer
7
Female
35
2
2
Traffic accident during life, amputation right leg, injury to the lower lumbar vertebra years before death, incontinence and intestinal problems, hypoesthesia in the legs
Gunshot through head
Extra information
Scanned three days before autopsy and once after autopsy
W. Haakma et al. / Forensic Science International 263 (2016) 139–146
(www.ExploreDTI.com) [30], and processing the data comprised the following steps. First, the acquired PMDTI scans were corrected for eddy current induced geometrical distortions and EPI deformations [31,32]. Secondly, diffusion tensors were estimated using the REKINDLE procedure with the iteratively reweighted linear least squares approach [33,34]. Finally, a deterministic streamline tractography approach was used to reconstruct the fiber pathways [35]. Diffusion parameters were calculated for each nerve bundle from L3 to S2 using a tract based analysis. Per nerve, ‘SEED’ and ‘AND’ ROIs were defined to select the fiber tracts belonging to the bundle of interest. ‘NOT’ ROIs were used to remove spurious tracts that did not correspond with known anatomy. Next, a segment of 3 cm was selected starting at the level where the nerve branched with the spine. Tractography threshold parameters were set to: FA = 0.10, MD = 0.0001 mm2/s, minimum fiber length = 15 mm, and angle deviation = 208 per integration step (step size = 1 mm). In addition to the lumbar and sacral nerves, pathways of the spine fiber bundle were reconstructed. Because the MD is higher in the spine compared to surrounding tissue, the MD threshold was increased to 0.00055 mm2/s. The diffusion parameters for each structure of interest were derived from the FT results. Statistical analysis of the diffusion parameters was performed with SPSS version 20.0 (SPSS Inc., Chicago, IL, USA). The specific statistical tests for each of the four experiments are described below. 2.4. Experimental procedures This study focused on different aspects of the PMDTI temporal stability and reproducibility and consisted of four experiments: inter-scan reproducibility and effect of temperature, inter-session reproducibility and autopsy, PM versus living subjects, and the examination of a pathological case. 2.4.1. Experiment 1: inter-scan reproducibility and effect of temperature Inter-scan reproducibility of the diffusion properties for all nerve levels of the first two consecutive scans of PM subjects 1–5 were investigated with a Bland–Altman plot [36]. The 95% limits of inter-scan agreement per diffusion parameter were defined as the mean of paired differences 1.96 the standard deviation (SD). The potential effect of temperature change on the diffusion parameters was investigated by recording the body temperature in the abdomen before and directly after every MRI session. Changes in temperature were correlated to mean values of FA, MD, AD and RD calculated from two ROIs drawn in the erector spinae muscle (lower back muscle). The diffusion parameters of the first and last scan were then compared using the Wilcoxon signed rank-sum test. FT results between the consecutive scans were visually compared. 2.4.2. Experiment 2: inter-session reproducibility and autopsy To determine the inter-session reproducibility PM subject 6 was scanned three times with intervals of 24 h (scan time point 1, 2, and 3, respectively) after which autopsy was performed. Reproducibility of the diffusion parameters of these three scans was investigated with a Jones plot [37]. The 95% limits of intersession agreement per diffusion parameter were defined as the mean of the differences 1.96 the standard deviation (SD). During the subsequent autopsy L1–L4 was removed to expose the cauda equina. Twenty-four hours after the third scan, the subject was scanned again and the FT results were then visually compared with the FT results obtained from the three scans before autopsy. 2.4.3. Experiment 3: post-mortem versus living subjects The diffusion parameters of the lumbosacral nerves of PM subjects were compared to those of living subjects using the
141
Mann–Whitney U test. The FT results of the lumbosacral nerves of PM subjects were then visually compared to the living subjects. 2.4.4. Experiment 4: pathological case Subject 7, a female PM subject of 35 years, with previous medical history of injury of the lower lumbar vertebra years before death, was included. She suffered from incontinence, intestinal problems and hypoesthesia in the legs. The FT results were visually compared to those of the PM subjects with normal anatomy.
3. Results In this PMDTI study, we were able to reconstruct the 3D architecture of the lumbar and sacral nerves in all seven PM subjects and all six living subjects, detailing the individual pathway trajectories and the microstructural properties of L3–S2. 3.1. Experiment 1: inter-scan reproducibility and effect of temperature Fig. 1 shows Bland–Altman plots of the first two consecutive scans of PM subjects 1–5 with limits of agreement of 0.058 to 0.062, ( 0.037 to 0.052) 10 3 mm2/s, ( 0.051 to 3 2 0.069) 10 mm /s, and ( 0.038 to 0.048) 10 3 mm2/s for FA, MD, AD, and RD, respectively. The measured temperature of the PM subjects was on average 15.5 8C (range 11.9–18.2 8C) at the start of the examination, and 18.2 8C (range 14.0–21.5 8C) at the end of the examination. There was a small but significant increase in RD (0.0117 10 3 mm2/s, p = 0.037) in the erector spinae muscle between the first and last scan. The architecture of lumbosacral nerves visualized with FT was comparable between the consecutive scans for all PM subjects (Fig. 2). 3.2. Experiment 2: inter-session reproducibility and autopsy Fig. 3 shows Jones plots of the first three scans (before autopsy) of subject 6 with limit of agreement of 0.0423 to 0.0423, ( 0.0442 to 0.0442) 10 3 mm2/s, ( 0.0564 to 0.0564) 10 3 mm2/s, and ( 0.0407 to 0.0407) 10 3 mm2/s for FA, MD, AD and RD, respectively. Mean diffusion values over all nerve levels of FA, MD, AD, and RD remained constant (FA = 0.37 0.05, 0.36 0.03, and 0.36 0.04, MD = 0.35 0.03 10 3 mm2/s, 0.35 0.03 10 3 mm2/ s, and 0.34 0.04 10 3 mm2/s, AD = 0.50 0.06 10 3 mm2/s, 0.49 0.05 10 3 mm2/s, 0.47 0.05 10 3 mm2/s, and RD = 0.27 0.02 10 3 mm2/s, 0.28 0.03 10 3 mm2/s, 0.27 0.04 10 3 mm2/s for scan time points 1, 2, and 3, respectively). The architecture of lumbosacral nerves visualized with FT of the first three scans obtained before autopsy showed no qualitative changes in FT results (Fig. 4A–C). After the autopsy, where the vertebrae at the level of L1–L4 were removed (Fig. 4E and I), changes in FT of the nerves at that level and at lower levels (L4–S2) were identified (Fig. 4D). A decrease of FA and an increase of MD, AD, and RD were observed (0.28 0.03, 0.41 0.04 10 3 mm2/s, 0.53 0.02 10 3 mm2/s, 0.34 0.03 10 3 mm2/s, respectively). 3.3. Experiment 3: post-mortem versus living subjects Nerves of PM subjects showed approximately four-fold lower MD, AD, and RD compared to those of living subjects (Table 2). The FT results of PM subjects were comparable to living subjects (Fig. 5A and C). It was possible to track the nerves at each level. The lumbosacral nerves visualized with FT were similar in architecture for both PM and living subjects.
142
W. Haakma et al. / Forensic Science International 263 (2016) 139–146
Fig. 1. Bland–Altman plots for each diffusion parameter (fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD)) in all postmortem subjects. The y-axis is the difference of the diffusion parameter between the first and second scan and the x-axis is the mean value of the first and second scan. Agreement is shown between the first two scans within one MRI session for all diffusion parameters. The upper dashed line shows the upper 95% limit of agreement, the bottom dashed line shows the lower 95% limit of agreement, and the red line shows the mean of the differences.
3.4. Experiment 4: pathological case
4. Discussion
At the level of the damaged cauda equina (L5, S1, S2), the reconstructed nerves with FT showed disorganization and asymmetry (Fig. 5B) compared to the non-trauma FT results (Fig. 5A). At the non-affected levels (L3 and L4) it was possible to reconstruct the nerves up to the spine with FT which was also possible in non-trauma PM subjects.
In this study, we demonstrated the ability of PMDTI to reconstruct the 3D architecture of the lumbar and sacral nerves in PM subjects, detailing the individual pathway trajectories and the microstructural properties at the level of L3–S2. Diffusion parameters and FT were found to be reproducible over time. The FT results in PM subjects were comparable to those obtained in living
Fig. 2. Post-mortem fiber tractography results of the lumbosacral nerves in a post-mortem subject from the consecutive scans within one MRI session (A: first scan, B: second scan, C: third scan, D: fourth scan). The color-encoding is according to the mean diffusivity (in units mm2/s). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
W. Haakma et al. / Forensic Science International 263 (2016) 139–146
143
Fig. 3. Jones plots of the inter-session measurements for each diffusion parameter (fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD)). The y-axis is the difference between an individual diffusion measurement (of scan time point 1, 2, and 3, respectively) and the mean of the three diffusion measurements (of scan time point 1, 2, and 3). The x-axis is the mean of the three diffusion measurements (of scan time point 1, 2, and 3). Agreement is shown between the three scans acquired in three consecutive days. The upper dashed line shows the upper 95% limit of agreement, the bottom dashed line shows the lower 95% limit of agreement, and the red line shows the mean of the differences (zero).
subjects in terms of architectural configuration. We demonstrated that it was possible to identify nerve damage in a chronic trauma case using PMDTI. Bland–Altman plots showed good agreement between the first two scans within one MRI session indicating a good inter-scan
Table 2 Diffusion parameters (fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD)) with standard deviation (SD) of postmortem (PM) and living subjects.
FA Living PM
L3
L4
L5
S1
S2
Mean SD
Mean SD
Mean SD
Mean SD
Mean SD
0.27 0.05 0.29 0.07
0.27 0.04 0.29 0.06
0.29 0.03 0.33 0.07
0.26 0.04 0.29 0.08
0.25 0.03 0.29 0.07
1.43 0.17* 0.37 0.04*
1.41 0.17* 0.36 0.03*
1.47 0.17* 0.39 0.07*
1.49 0.10* 0.36 0.08*
1.84 0.18* 0.48 0.04*
1.87 0.19* 0.49 0.03*
1.89 0.16* 0.52 0.07*
1.88 0.12* 0.48 0.08*
1.22 0.17* 0.31 0.05*
1.19 0.16* 0.29 0.04*
1.26 0.17* 0.33 0.08*
1.29 0.10* 0.31 0.08*
MD (10 3 mm2/s) Living 1.47 0.16* PM 0.38 0.05* AD (10 Living PM
3
mm2/s) 1.90 0.17* 0.50 0.05*
RD (10 Living PM
3
mm2/s) 1.25 0.16* 0.31 0.05*
* p < 0.0005, significant difference at each nerve level between living and PM in MD, AD, RD.
reproducibility. The effect of temperature was measured in the muscle rather than in nerves as it was expected that the muscle provided a more robust measurement. Due to the larger area, the variability of the diffusion parameters will be smaller and will suffer less from partial volume effects in comparison to nervous tissue. The observed differences in RD between the first and last scan are in line with the fact that an increase in temperature leads to an increased diffusion. However, this difference was relatively low and had a negligible effect on the diffusion parameter estimates and FT results. Despite the temperature differences between the PM subjects, FT and diffusion parameters were reproducible as reflected by the low range of limits of agreement (Fig. 3) and the unchanged architecture of the nerves over days (Fig. 4). This implies that degeneration processes were limited due to cooling of the PM subjects shortly after death, retarding the cellular breakdown. In general, diffusion parameters are affected by the breakdown of cellular membranes, but this is limited the first few days after death when the body is preserved in a cooled environment. A previous report showed that leaving tissue unfixed led to deleterious effects on diffusion properties in brain tissue over time [15]. Diffusion values may decrease further when the PM subject is preserved in the morgue for a longer period of time. However, the same authors also reported that diffusion parameters in the brain decreased in the first three days when cooled immediately after death [15]. We were unable to confirm this finding based on the data of our study.
144
W. Haakma et al. / Forensic Science International 263 (2016) 139–146
Fig. 4. Post-mortem fiber tractography (FT) results of a subject which was scanned four days in a row. A, B, and C show the FT results obtained on day 1, 2 and 3, respectively. F, G, and H show the matching T2-weighted images of the spinal cord. D shows the FT results after autopsy on the fourth day. I shows the cauda equina exposed during autopsy where L1–L4 were removed (E and I). Note that there are small air cavities in the spinal cord (I) indicated with red arrows which were not present during the first three scans (F, G, and H). Color-encoding is the same as in Fig. 2. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
There were differences in PM FT and diffusion parameters between post and pre autopsy. However, the exact mechanism explaining these differences is unknown. Possible explanations can be related to different processes of which we discuss three examples. First, mechanical (e.g. due to force applied to remove L1–L4) and morphological (e.g. reduction in cerebrospinal fluid due to leakage) processes might change FT and diffusion parameters. Secondly, these differences may be related to small air cavities observed in the spine and lower part of the abdomen after the autopsy, due to leakage of the cerebrospinal fluid from the spinal
cord. Fig. 4F–H shows the spinal cord fully intact, where Fig. 4I shows small air cavities in the spinal cord. These air cavities lead to susceptibility artifacts with subsequent adverse effects on the FT results. Lastly, it is likely that the observed temperature changes in the subjects have an effect on the diffusion parameters. PM FA values were comparable to in vivo results. FA values describe the ratio between AD and RD values. As the ratio between AD and RD remains relatively unchanged in a PM setting, the anisotropic properties of the nervous tissue PM are still intact. Schmidt et al. investigated the brain with diffusion weighted
Fig. 5. Fiber tractography results of (A) a post-mortem (PM) subject, (B) a PM case with injury to the lower vertebra, and (C) a living subject. The architecture of the lumbosacral nerves PM is comparable to living subjects. The traumatic case (B) shows assymetry and disorganization at the level of the damaged cauda equina. Colorencoding is the same as in Fig. 2. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
W. Haakma et al. / Forensic Science International 263 (2016) 139–146
imaging (DWI) and found MD values which were approximately 3–4 times lower than in vivo [23], which was similar to our results. Part of the observed differences in MD, AD and RD between PM subjects and in living nervous tissue are presumed to be caused by temperature differences [18]. A drop in temperature also influences the viscosity of liquids and could therefore be another explanation for our results [38]. However, a previous study showed that when correcting for temperature between PM and in vivo results, differences were still present [20]. Therefore, there might be other processes which could explain these differences. For practical reasons the PM subjects were scanned on a 1.5 T scanner and the living subjects on a 3 T scanner. Furthermore, different b-values were used in the in vivo and PM protocol. A different field strength and b-value may induce differences in calculated diffusion parameters [39,40]. The similar results in nerve architecture between PM and in vivo suggest that PMDTI is a reproducible method to visualize the 3D architecture of nerves. The results from the traumatic case with chronic traumatic lesions showed asymmetry and disorganization of the lumbosacral nerves compared to the non-trauma FT results. This was in accordance with the clinical condition of the deceased subject (Table 1) and may be explained by the chronic changes in nervous tissue during a span of years. In future research, more pathological chronic trauma cases should be evaluated to support the hypothesis that PMDTI and FT are able to identify pathology and trauma. Furthermore, acute trauma cases should be included to investigate to what extent PMDTI and FT are able to identify acute lesions of peripheral nervous tissue. However, as this case is comparable to results found in spina bifida patients [26], we expect that PMDTI and FT may advantageously be used to detect and document nerve disorders. Although the sample size of this study was small, the matching FT results with living subjects illustrate the potential of PMDTI and FT in investigating peripheral nervous structures after death. The relatively low spatial resolution used in the DTI sequences is likely affecting the measured diffusion parameters, especially in small nerve roots [41,42]. In this study, a single-shot EPI sequence was used in both in vivo and PM studies for comparison, but the PM protocol could benefit from using multi-shot EPI sequence as it allows the possibility to increase resolution or to reduce distortions [43]. However, such sequence leads to prolonged scan time and the risk of augmented body temperature. The peripheral nervous system is a relatively neglected area in the discipline of forensic pathology which needs further attention in order to elucidate its significance. Although the nervous system is infrequently evaluated during a normal autopsy, visualization of these structures could provide important insights about nerve damage, particular in traumatic cases. Additionally, PMDTI provides the opportunity to identify other peripheral nerve pathologies [21] which can be helpful in understanding pathogenesis and disease progression as it can be compared with histology. In conclusion, this study demonstrates that PMDTI and FT are reproducible methods to investigate the architecture and microstructure of the lumbar and sacral nerves in PM subjects. We showed the potential of PMDTI and FT to detect nerve trauma in one case with chronic lesions. Therefore, PMDTI and FT may become valuable in identification and documentation of PM nerve trauma or pathologies in forensic sciences.
Grant support The research of A.L. is supported by VIDI Grant 639.072.411 from the Netherlands Organisation for Scientific Research (NWO).
145
Acknowledgements We would like to thank the hospital porters at Skejby Sygehus and Lidy Kuster for their help in performing the MRI examinations.
References [1] S. Grabherr, B.A. Stephan, U. Buck, S. Na¨ther, A. Christe, L. Oesterhelweg, et al., Virtopsy – radiology in forensic medicine, Imaging Decis. MRI 11 (2007) 2–9. [2] I.S.D. Roberts, R.E. Benamore, E.W. Benbow, S.H. Lee, J.N. Harris, A. Jackson, et al., Post-mortem imaging as an alternative to autopsy in the diagnosis of adult deaths: a validation study, Lancet 379 (2012) 136–142. [3] J. An˜on, L. Remonda, A. Spreng, E. Scheurer, G. Schroth, C. Boesch, et al., Traumatic extra-axial hemorrhage: correlation of postmortem MSCT, MRI, and forensicpathological findings, J. Magn. Reson. Imaging 28 (2008) 823–836. [4] P.M. Leth, H. Struckmann, J. Lauritsen, Interobserver agreement of the injury diagnoses obtained by postmortem computed tomography of traffic fatality victims and a comparison with autopsy results, Forensic Sci. Int. 225 (2013) 15–19. [5] R. Dirnhofer, C. Jackowski, P. Vock, K. Potter, M.J. Thali, VIRTOPSY: minimally invasive, imaging-guided virtual autopsy, Radiographics 26 (2006) 1305–1333. [6] S. Grabherr, E. Gygax, B. Sollberger, S. Ross, L. Oesterhelweg, S. Bolliger, et al., Twostep postmortem angiography with a modified heart–lung machine: preliminary results, AJR Am. J. Roentgenol. 190 (2008) 345–351. [7] C. Bruguier, P.J. Mosimann, P. Vaucher, A. Uske´, F. Doenz, C. Jackowski, et al., Multiphase postmortem CT angiography: recognizing technique-related artefacts and pitfalls, Int. J. Legal Med. 127 (2013) 639–652. [8] A. Christe, S. Ross, L. Oesterhelweg, D. Spendlove, S. Bolliger, P. Vock, et al., Abdominal trauma – sensitivity and specificity of postmortem noncontrast imaging findings compared with autopsy findings, J. Trauma 66 (2009) 1302–1307. [9] A.H. Thomsen, A.G. Jurik, L. Uhrenholt, A. Vesterby, An alternative approach to Computerized Tomography (CT) in forensic pathology, Forensic Sci. Int. 183 (2009) 87–90. [10] S. Thayyil, N.J. Sebire, L.S. Chitty, A. Wade, W. Chong, O. Olsen, et al., Post-mortem MRI versus conventional autopsy in fetuses and children: a prospective validation study, Lancet 382 (2013) 223–233. [11] D.K. Jones, A. Leemans, Diffusion tensor imaging, in: M. Modo, J.W.M. Bulte (Eds.), Magnetic Resonance Neuroimaging, Humana Press, 2011, pp. 127–144. [12] C. Beaulieu, The basis of anisotropic water diffusion in the nervous system – a technical review, NMR Biomed. 15 (2002) 435–455. [13] S. Mori, P.C.M. van Zijl, Fiber tracking: principles and strategies – a technical review, NMR Biomed. 15 (2002) 468–480. [14] D. Le Bihan, J.F. Mangin, C. Poupon, C.A. Clark, S. Pappata, N. Molko, et al., Diffusion tensor imaging: concepts and applications, J. Magn. Reson. Imaging 13 (2001) 534–546. [15] H. D’Arceuil, A. de Crespigny, The effects of brain tissue decomposition on diffusion tensor imaging and tractography, Neuroimage 36 (2007) 64–68. [16] A. Pfefferbaum, E.V. Sullivan, E. Adalsteinsson, T. Garrick, C. Harper, Postmortem MR imaging of formalin-fixed human brain, Neuroimage 21 (2004) 1585–1595. [17] K.L. Miller, C.J. Stagg, G. Douaud, S. Jbabdi, S.M. Smith, T.E. Behrens, et al., Diffusion imaging of whole, post-mortem human brains on a clinical MRI scanner, Neuroimage 57 (2011) 167–181. [18] H.E. D’Arceuil, S. Westmoreland, A.J. de Crespigny, An approach to high resolution diffusion tensor imaging in fixed primate brain, Neuroimage 35 (2007) 553–565. [19] S. Rane, T.Q. Duong, Comparison of in vivo and ex vivo diffusion tensor imaging in rhesus macaques at short and long diffusion times, Open Neuroimaging J. 5 (2011) 172–178. [20] E. Scheurer, K.O. Lovblad, R. Kreis, S.E. Maier, C. Boesch, R. Dirnhofer, et al., Forensic application of postmortem diffusion-weighted and diffusion tensor MR imaging of the human brain in situ, Am. J. Neuroradiol. 32 (2011) 1518–1524. [21] K. Schmierer, C.A.M. Wheeler-Kingshott, P.A. Boulby, F. Scaravilli, D.R. Altmann, G.J. Barker, et al., Diffusion tensor imaging of post mortem multiple sclerosis brain, Neuroimage 35 (2007) 467–477. [22] P.M. Flach, S. Schroth, W. Schweitzer, G. Ampanozi, J. Slotboom, C. Kiefer, et al., Deep into the fibers! Postmortem diffusion tensor imaging in forensic radiology, Am. J. Forensic Med. Pathol. 36 (2015) 153–161. [23] T.M. Schmidt, R. Fischer, S. Acar, M. Lorenzen, A. Heinemann, U. Wedega¨rtner, et al., DWI of the brain: postmortal DWI of the brain in comparison with in vivo data, Forensic Sci. Int. 220 (2012) 180–183. [24] K. Yen, J. Weis, R. Kreis, E. Aghayev, C. Jackowski, M. Thali, et al., Line-scan diffusion tensor imaging of the posttraumatic brain stem: changes with neuropathologic correlation, Am. J. Neuroradiol. 27 (2006) 70–73. [25] S. Winklhofer, C.T. Stoeck, N. Berger, M. Thali, R. Manka, S. Kozerke, et al., Postmortem cardiac diffusion tensor imaging: detection of myocardial infarction and remodeling of myofiber architecture, Eur. Radiol. 24 (2014) 2810–2818. [26] W. Haakma, P. Dik, B. Ten Haken, M. Froeling, R.A.J. Nievelstein, I. Cuppen, et al., Diffusion tensor magnetic resonance imaging and fiber tractography of the sacral plexus in children with spina bifida, J. Urol. 192 (2014) 927–933. [27] D.C. Karampinos, G. Melkus, T.M. Shepherd, S. Banerjee, E.U. Saritas, A. Shankaranarayanan, et al., Diffusion tensor imaging and T2 relaxometry of bilateral lumbar nerve roots: feasibility of in-plane imaging, NMR Biomed. 26 (2013) 630–637.
146
W. Haakma et al. / Forensic Science International 263 (2016) 139–146
[28] T. Takahara, J. Hendrikse, T.C. Kwee, T. Yamashita, M. Van Cauteren, D. Polders, et al., Diffusion-weighted MR neurography of the sacral plexus with unidirectional motion probing gradients, Eur. Radiol. 20 (2010) 1221–1226. [29] P.K.N. van der Jagt, P. Dik, M. Froeling, T.C. Kwee, R.A. Nievelstein, B. ten Haken, et al., Architectural configuration and microstructural properties of the sacral plexus: a diffusion tensor MRI and fiber tractography study, Neuroimage 62 (2012) 1792–1799. [30] A. Leemans, B. Jeurissen, J. Sijbers, D.K. Jones, ExploreDTI: a graphical toolbox for processing, analyzing, and visualizing diffusion MR data, Proc. Intl. Soc. Magn. Reson. Med. 17 (2009) 3536. [31] A. Leemans, D.K. Jones, The B-matrix must be rotated when correcting for subject motion in DTI data, Magn. Reson. Med. 61 (2009) 1336–1349. [32] M.O. Irfanoglu, L. Walker, J. Sarlls, S. Marenco, C. Pierpaoli, Effects of image distortions originating from susceptibility variations and concomitant fields on diffusion MRI tractography results, Neuroimage 61 (2012) 275–288. [33] C.M. Tax, W.M. Otte, M.A. Viergever, R.M. Dijkhuizen, A. Leemans, REKINDLE: robust extraction of kurtosis INDices with linear estimation, Magn. Reson. Med. 73 (2014) 794–808. [34] J. Veraart, J. Sijbers, S. Sunaert, A. Leemans, B. Jeurissen, Weighted linear least squares estimation of diffusion MRI parameters: strengths, limitations, and pitfalls, Neuroimage 81 (2013) 335–346.
[35] P.J. Basser, S. Pajevic, C. Pierpaoli, J. Duda, A. Aldroubi, In vivo fiber tractography using DT-MRI data, Magn. Reson. Med. 44 (2000) 625–632. [36] J.M. Bland, D.G. Altman, Statistical methods for assessing agreement between two methods of clinical measurement, Lancet 1 (1986) 307–310. [37] M. Jones, A. Dobson, S. O’Brian, A graphical method for assessing agreement with the mean between multiple observers using continuous measures, Int. J. Epidemiol. 40 (2011) 1308–1313. [38] H. Ozbek, Viscosity of aqueous sodium chloride solutions from 0 to 150 8C, in: 29th Southeast Regional Meeting of the American Chemical Society, 1971, 1–68. [39] D.L. Polders, A. Leemans, J. Hendrikse, M.J. Donahue, P.R. Luijten, J.M. Hoogduin, Signal to noise ratio and uncertainty in diffusion tensor imaging at 1.5, 3.0, and 7.0 T, J. Magn. Reson. Imaging 33 (2011) 1456–1463. [40] E.S. Hui, M.M. Cheung, K.C. Chan, E.X. Wu, B-value dependence of DTI quantitation and sensitivity in detecting neural tissue changes, Neuroimage 49 (2010) 2366– 2374. [41] Q. Hogan, Size of human lower thoracic and lumbosacral nerve roots, Anesthesiology 85 (1996) 37–42. [42] S.B. Vos, D.K. Jones, M.A. Viergever, A. Leemans, Partial volume effect as a hidden covariate in DTI analyses, Neuroimage 55 (2011) 1566–1576. [43] S. Chabert, C. Galindo, C. Tejos, S.A. Uribe, Multiple echo multi-shot diffusion sequence, J. Magn. Reson. Imaging 39 (2014) 1027–1032.
Contents lists available at ScienceDirect
Forensic Science International journal homepage: www.elsevier.com/locate/forsciint
Diffusion tensor imaging of peripheral nerves in non-fixed post-mortem subjects Wieke Haakma a,b,c,*, Michael Pedersen b,d, Martijn Froeling c, Lars Uhrenholt a, Alexander Leemans e, Lene Warner Thorup Boel a a
Department of Forensic Medicine, Aarhus University, Aarhus, Denmark Comparative Medicine Lab, Department of Clinical Medicine, Aarhus University, Aarhus, Denmark Department of Radiology, University Medical Center Utrecht, Utrecht, The Netherlands d MR Research Center, Department of Clinical Medicine, Aarhus University, Aarhus, Denmark e Image Sciences Institute, University Medical Center Utrecht, Utrecht, The Netherlands b c
A R T I C L E I N F O
A B S T R A C T
Article history: Received 26 November 2015 Received in revised form 10 February 2016 Accepted 1 April 2016 Available online 11 April 2016
Purpose: While standard magnetic resonance imaging (MRI) sequences are increasingly employed in post-mortem (PM) examinations, more advanced techniques such as diffusion tensor imaging (DTI) remain unexplored in forensic sciences. Therefore, we studied the temporal stability and reproducibility of DTI and fiber tractography (FT) in non-fixed PM subjects. In addition, we investigated the lumbosacral nerves with PMDTI and compared their tissue characteristics to in vivo findings. Methods: MRI data were acquired on a 1.5 T MRI scanner in seven PM subjects, consisting of six nontrauma deaths and one chronic trauma death, and in six living subjects. Inter-scan (within one session) and inter-session (between days) reproducibility of diffusion parameters, fractional anisotropy (FA), and mean diffusivity (MD), were evaluated for the lumbosacral nerves using Bland–Altman and Jones plots. Diffusion parameters in nerves L3–S2 were compared to living subjects using the non-parametric Mann– Whitney U test. Results: Reproducibility of diffusion values of inter-scan 95% limits of agreement ranged from 0.058 to 0.062 for FA, and ( 0.037 to 0.052) 10 3 mm2/s for MD. For the inter-session this was 0.0423 to 0.0423, and ( 0.0442 to 0.0442) 10 3 mm2/s for FA, and MD, respectively. Although PM subjects showed approximately four-fold lower diffusivity values compared to living subjects, FT results were comparable. The chronic trauma case showed disorganization and asymmetry of the nerves. Conclusion: We demonstrated that DTI was reproducible in characterizing nervous tissue properties and FT in reconstructing the architecture of lumbosacral nerves in PM subjects. We showed differences in diffusion values between PM and in vivo and showed the ability of PMDTI and FT to reconstruct nerve lesions in a chronic trauma case. We expect that PMDTI and FT may become valuable in identification and documentation of PM nerve trauma or pathologies in forensic sciences. ß 2016 Elsevier Ireland Ltd. All rights reserved.
Keywords: Post-mortem imaging Magnetic resonance imaging Diffusion tensor imaging Fiber tractography Peripheral nerves
1. Introduction Imaging techniques are becoming increasingly important in forensic examinations and may contribute to, or partially replace,
Abbreviations: AD, axial diffusivity; DTI, diffusion tensor imaging; EPI, echo planar imaging; FA, fractional anisotropy; FOV, field of view; FT, fiber tractography; MD, mean diffusivity; MRI, magnetic resonance imaging; PM, post-mortem; PMDTI, post-mortem diffusion tensor imaging; PMMRI, post-mortem magnetic resonance imaging; RD, radial diffusivity; TSE, turbo spin echo. * Corresponding author at: Department of Forensic Medicine, Aarhus University, Palle Juul-Jensens Boulevard 99, 8200 Aarhus N, Denmark. Tel.: +45 8716 7500. E-mail address: [email protected] (W. Haakma). http://dx.doi.org/10.1016/j.forsciint.2016.04.001 0379-0738/ß 2016 Elsevier Ireland Ltd. All rights reserved.
conventional autopsy [1–4]. With computed tomography (CT) it is possible to identify fractures, pathologic gas development, large hemorrhages, and metal or other foreign bodies [5–9]. Magnetic resonance imaging (MRI) on the other hand is particularly valuable when soft tissue injuries or organ trauma need to be identified [5,8,10]. Regarding some of the more advanced MRI techniques, such as diffusion tensor imaging (DTI) [11–14], their potential value in forensic sciences remains unknown. DTI is an MRI technique which can measure the random movement of water molecules, known as Brownian motion [11]. In nervous tissue this movement is more pronounced along the nerve, so-called anisotropy [12]. Especially in the identification of peripheral nerves, DTI might be valuable, since nerves are difficult to dissect
W. Haakma et al. / Forensic Science International 263 (2016) 139–146
140
during conventional autopsy and are therefore typically neglected in forensic pathology. DTI can be used to characterize the structure of nervous tissue in detail and microstructural properties can be evaluated. DTI may serve as a valuable tool in investigations of accidents, homicides and other traumatic cases. Important differences between post-mortem (PM) and living tissue should be considered. Shortly after death, processes such as autolysis occur. Combined with bacterial degradation, which facilitates tissue decomposition, nervous tissue will deteriorate [15]. Therefore, PMDTI has been performed predominantly on formaldehyde fixed tissue [16]. However, molecular diffusion characteristics are known to change markedly following the fixation process due to formation of intra- and inter-molecular cross-links [17]. Moreover, temperature regulation of PM subjects can be challenging during PMMRI, which can influence DTI results since water diffusion is highly temperature sensitive [18]. Previous PMDTI studies have focused on fixed human brains and demonstrated preserved diffusion anisotropy in nerves [15,17,19], allowing for fiber tractography (FT) and quantitative investigation of diffusion parameters such as fractional anisotropy (FA), which represents the degree in which diffusion is oriented in one direction, and mean diffusivity (MD), which is the average of all eigenvalues (the magnitude of diffusion in a direction). The FA and MD are based on the axial diffusivity (AD), which is the largest of the eigenvalues and in healthy nervous tissue diffusion is mostly represented along the nerve, and radial diffusivity (RD), which is the average of the second and third eigenvalue and in healthy nervous tissue diffusion is perpendicular to the nerve. Only a few studies have employed PMDTI in non-fixed human tissue, including the brain [20–23], brain trauma [24] and cardiac infarction [25]. The aim of this study was to demonstrate that PMDTI can be used for identification and quantification of peripheral nervous tissue in a reproducible way and to identify peripheral nerve trauma. Specifically, we investigated the lumbosacral nerves with PMDTI, because they already have been described in detail in several in vivo DTI studies [26–29], and we compared their tissue characteristics to in vivo findings. In addition, we studied the temporal stability and reproducibility of DTI and FT in non-fixed PM subjects.
of 46 years, range 30–55 years all non-trauma deaths, and one female trauma death; 35 years). PMMRI was performed 1–8 days after the estimated date of death. Specific subject information is listed in Table 1. Acquired data were included in four experiments (described below); data of subject 1–5 were included in experiment 1, data of subject 6 were included in experiment 2, data of subject 1– 5 were included in experiment 3, and data of subject 7 were included in experiment 4. All subjects were cooled in the morgue at 4 8C before and after autopsy without fixative solution preparation. The routine autopsy was performed before PMMRI (except for experiment 2) and did not affect the lumbosacral area. Six healthy living subjects (mean age of 30 years, range 25–42 years) were included as reference subjects for experiment 3. 2.2. Data acquisition
2. Materials and methods
MRI of the PM subjects was performed on a 1.5 T MRI scanner (Achieva; Philips Healthcare, Best, The Netherlands) using a 16-channel phased-array surface coil. PMDTI was performed with diffusion-weighted spin echo single-shot echo planar imaging (EPI) using the following parameters: TE = 82 ms, TR = 13,538 ms, SENSE factor = 2, number of excitations = 8, field of view (FOV) = 384 mm 216 mm, matrix size = 128 72, 72, slice thickness = 3.0 mm, resulting in a voxel size of 3.0 mm 3.0 mm 3.0 mm, EPI factor = 35, b-values of 0 and 2000 s/mm2, and 15 gradient directions. For three PM subjects the protocol was repeated two times within the scan session, and for four PM subjects it was repeated four times within the scan session, resulting in total scan times of 58 min and 1 h 56 min, respectively. MRI of the six living subjects was performed on a 3 T MRI scanner (Achieva; Philips Healthcare, Best, The Netherlands) using a 16channel phased-array surface coil. DTI was performed with diffusion-weighted spin echo single-shot EPI using the following parameters: TE = 58 ms, TR = 4209 ms, SENSE factor = 2, number of excitations = 2, FOV = 336 mm 336 mm, matrix size = 112 112, slice thickness = 3.0 mm, resulting in a voxel size of 3.0 mm 3.0 mm 3.0 mm, EPI factor = 35, b-values of 0 and 800 s/mm2, and 15 gradient directions. The total acquisition time was 4.5 min. For anatomical reference, a 3D turbo spin echo (TSE) sequence was acquired based on the protocol described by [29] for both PM and living subjects.
2.1. Subjects
2.3. Data analysis
In this study, seven non-fixed PM subjects with normal anatomy of the lower spine were included (five male, one female; median age
DTI data from the PM and living subjects were processed identically, using the ExploreDTI diffusion MRI toolbox
Table 1 Subject specifications. Identification number
Sex
Age in years
Approximate number of days between death and scan
Number of scans within one session
History
Cause of death
1 2 3 4
Male Male Male Male
55 46 38 30
3 1 5 5
2 2 4 4
Alcohol abuse Depression, posttraumatic stress disorder Drug abuse Schizophrenia
5 6
Male Female
46 49
3 5
4 4
Drug abuse Schizophrenia, drug and alcohol abuse
Cardiac death Gunshot through head Overdose methadone Overdose of potassium tablets General physical decline Pancreatic cancer
7
Female
35
2
2
Traffic accident during life, amputation right leg, injury to the lower lumbar vertebra years before death, incontinence and intestinal problems, hypoesthesia in the legs
Gunshot through head
Extra information
Scanned three days before autopsy and once after autopsy
W. Haakma et al. / Forensic Science International 263 (2016) 139–146
(www.ExploreDTI.com) [30], and processing the data comprised the following steps. First, the acquired PMDTI scans were corrected for eddy current induced geometrical distortions and EPI deformations [31,32]. Secondly, diffusion tensors were estimated using the REKINDLE procedure with the iteratively reweighted linear least squares approach [33,34]. Finally, a deterministic streamline tractography approach was used to reconstruct the fiber pathways [35]. Diffusion parameters were calculated for each nerve bundle from L3 to S2 using a tract based analysis. Per nerve, ‘SEED’ and ‘AND’ ROIs were defined to select the fiber tracts belonging to the bundle of interest. ‘NOT’ ROIs were used to remove spurious tracts that did not correspond with known anatomy. Next, a segment of 3 cm was selected starting at the level where the nerve branched with the spine. Tractography threshold parameters were set to: FA = 0.10, MD = 0.0001 mm2/s, minimum fiber length = 15 mm, and angle deviation = 208 per integration step (step size = 1 mm). In addition to the lumbar and sacral nerves, pathways of the spine fiber bundle were reconstructed. Because the MD is higher in the spine compared to surrounding tissue, the MD threshold was increased to 0.00055 mm2/s. The diffusion parameters for each structure of interest were derived from the FT results. Statistical analysis of the diffusion parameters was performed with SPSS version 20.0 (SPSS Inc., Chicago, IL, USA). The specific statistical tests for each of the four experiments are described below. 2.4. Experimental procedures This study focused on different aspects of the PMDTI temporal stability and reproducibility and consisted of four experiments: inter-scan reproducibility and effect of temperature, inter-session reproducibility and autopsy, PM versus living subjects, and the examination of a pathological case. 2.4.1. Experiment 1: inter-scan reproducibility and effect of temperature Inter-scan reproducibility of the diffusion properties for all nerve levels of the first two consecutive scans of PM subjects 1–5 were investigated with a Bland–Altman plot [36]. The 95% limits of inter-scan agreement per diffusion parameter were defined as the mean of paired differences 1.96 the standard deviation (SD). The potential effect of temperature change on the diffusion parameters was investigated by recording the body temperature in the abdomen before and directly after every MRI session. Changes in temperature were correlated to mean values of FA, MD, AD and RD calculated from two ROIs drawn in the erector spinae muscle (lower back muscle). The diffusion parameters of the first and last scan were then compared using the Wilcoxon signed rank-sum test. FT results between the consecutive scans were visually compared. 2.4.2. Experiment 2: inter-session reproducibility and autopsy To determine the inter-session reproducibility PM subject 6 was scanned three times with intervals of 24 h (scan time point 1, 2, and 3, respectively) after which autopsy was performed. Reproducibility of the diffusion parameters of these three scans was investigated with a Jones plot [37]. The 95% limits of intersession agreement per diffusion parameter were defined as the mean of the differences 1.96 the standard deviation (SD). During the subsequent autopsy L1–L4 was removed to expose the cauda equina. Twenty-four hours after the third scan, the subject was scanned again and the FT results were then visually compared with the FT results obtained from the three scans before autopsy. 2.4.3. Experiment 3: post-mortem versus living subjects The diffusion parameters of the lumbosacral nerves of PM subjects were compared to those of living subjects using the
141
Mann–Whitney U test. The FT results of the lumbosacral nerves of PM subjects were then visually compared to the living subjects. 2.4.4. Experiment 4: pathological case Subject 7, a female PM subject of 35 years, with previous medical history of injury of the lower lumbar vertebra years before death, was included. She suffered from incontinence, intestinal problems and hypoesthesia in the legs. The FT results were visually compared to those of the PM subjects with normal anatomy.
3. Results In this PMDTI study, we were able to reconstruct the 3D architecture of the lumbar and sacral nerves in all seven PM subjects and all six living subjects, detailing the individual pathway trajectories and the microstructural properties of L3–S2. 3.1. Experiment 1: inter-scan reproducibility and effect of temperature Fig. 1 shows Bland–Altman plots of the first two consecutive scans of PM subjects 1–5 with limits of agreement of 0.058 to 0.062, ( 0.037 to 0.052) 10 3 mm2/s, ( 0.051 to 3 2 0.069) 10 mm /s, and ( 0.038 to 0.048) 10 3 mm2/s for FA, MD, AD, and RD, respectively. The measured temperature of the PM subjects was on average 15.5 8C (range 11.9–18.2 8C) at the start of the examination, and 18.2 8C (range 14.0–21.5 8C) at the end of the examination. There was a small but significant increase in RD (0.0117 10 3 mm2/s, p = 0.037) in the erector spinae muscle between the first and last scan. The architecture of lumbosacral nerves visualized with FT was comparable between the consecutive scans for all PM subjects (Fig. 2). 3.2. Experiment 2: inter-session reproducibility and autopsy Fig. 3 shows Jones plots of the first three scans (before autopsy) of subject 6 with limit of agreement of 0.0423 to 0.0423, ( 0.0442 to 0.0442) 10 3 mm2/s, ( 0.0564 to 0.0564) 10 3 mm2/s, and ( 0.0407 to 0.0407) 10 3 mm2/s for FA, MD, AD and RD, respectively. Mean diffusion values over all nerve levels of FA, MD, AD, and RD remained constant (FA = 0.37 0.05, 0.36 0.03, and 0.36 0.04, MD = 0.35 0.03 10 3 mm2/s, 0.35 0.03 10 3 mm2/ s, and 0.34 0.04 10 3 mm2/s, AD = 0.50 0.06 10 3 mm2/s, 0.49 0.05 10 3 mm2/s, 0.47 0.05 10 3 mm2/s, and RD = 0.27 0.02 10 3 mm2/s, 0.28 0.03 10 3 mm2/s, 0.27 0.04 10 3 mm2/s for scan time points 1, 2, and 3, respectively). The architecture of lumbosacral nerves visualized with FT of the first three scans obtained before autopsy showed no qualitative changes in FT results (Fig. 4A–C). After the autopsy, where the vertebrae at the level of L1–L4 were removed (Fig. 4E and I), changes in FT of the nerves at that level and at lower levels (L4–S2) were identified (Fig. 4D). A decrease of FA and an increase of MD, AD, and RD were observed (0.28 0.03, 0.41 0.04 10 3 mm2/s, 0.53 0.02 10 3 mm2/s, 0.34 0.03 10 3 mm2/s, respectively). 3.3. Experiment 3: post-mortem versus living subjects Nerves of PM subjects showed approximately four-fold lower MD, AD, and RD compared to those of living subjects (Table 2). The FT results of PM subjects were comparable to living subjects (Fig. 5A and C). It was possible to track the nerves at each level. The lumbosacral nerves visualized with FT were similar in architecture for both PM and living subjects.
142
W. Haakma et al. / Forensic Science International 263 (2016) 139–146
Fig. 1. Bland–Altman plots for each diffusion parameter (fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD)) in all postmortem subjects. The y-axis is the difference of the diffusion parameter between the first and second scan and the x-axis is the mean value of the first and second scan. Agreement is shown between the first two scans within one MRI session for all diffusion parameters. The upper dashed line shows the upper 95% limit of agreement, the bottom dashed line shows the lower 95% limit of agreement, and the red line shows the mean of the differences.
3.4. Experiment 4: pathological case
4. Discussion
At the level of the damaged cauda equina (L5, S1, S2), the reconstructed nerves with FT showed disorganization and asymmetry (Fig. 5B) compared to the non-trauma FT results (Fig. 5A). At the non-affected levels (L3 and L4) it was possible to reconstruct the nerves up to the spine with FT which was also possible in non-trauma PM subjects.
In this study, we demonstrated the ability of PMDTI to reconstruct the 3D architecture of the lumbar and sacral nerves in PM subjects, detailing the individual pathway trajectories and the microstructural properties at the level of L3–S2. Diffusion parameters and FT were found to be reproducible over time. The FT results in PM subjects were comparable to those obtained in living
Fig. 2. Post-mortem fiber tractography results of the lumbosacral nerves in a post-mortem subject from the consecutive scans within one MRI session (A: first scan, B: second scan, C: third scan, D: fourth scan). The color-encoding is according to the mean diffusivity (in units mm2/s). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
W. Haakma et al. / Forensic Science International 263 (2016) 139–146
143
Fig. 3. Jones plots of the inter-session measurements for each diffusion parameter (fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD)). The y-axis is the difference between an individual diffusion measurement (of scan time point 1, 2, and 3, respectively) and the mean of the three diffusion measurements (of scan time point 1, 2, and 3). The x-axis is the mean of the three diffusion measurements (of scan time point 1, 2, and 3). Agreement is shown between the three scans acquired in three consecutive days. The upper dashed line shows the upper 95% limit of agreement, the bottom dashed line shows the lower 95% limit of agreement, and the red line shows the mean of the differences (zero).
subjects in terms of architectural configuration. We demonstrated that it was possible to identify nerve damage in a chronic trauma case using PMDTI. Bland–Altman plots showed good agreement between the first two scans within one MRI session indicating a good inter-scan
Table 2 Diffusion parameters (fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD)) with standard deviation (SD) of postmortem (PM) and living subjects.
FA Living PM
L3
L4
L5
S1
S2
Mean SD
Mean SD
Mean SD
Mean SD
Mean SD
0.27 0.05 0.29 0.07
0.27 0.04 0.29 0.06
0.29 0.03 0.33 0.07
0.26 0.04 0.29 0.08
0.25 0.03 0.29 0.07
1.43 0.17* 0.37 0.04*
1.41 0.17* 0.36 0.03*
1.47 0.17* 0.39 0.07*
1.49 0.10* 0.36 0.08*
1.84 0.18* 0.48 0.04*
1.87 0.19* 0.49 0.03*
1.89 0.16* 0.52 0.07*
1.88 0.12* 0.48 0.08*
1.22 0.17* 0.31 0.05*
1.19 0.16* 0.29 0.04*
1.26 0.17* 0.33 0.08*
1.29 0.10* 0.31 0.08*
MD (10 3 mm2/s) Living 1.47 0.16* PM 0.38 0.05* AD (10 Living PM
3
mm2/s) 1.90 0.17* 0.50 0.05*
RD (10 Living PM
3
mm2/s) 1.25 0.16* 0.31 0.05*
* p < 0.0005, significant difference at each nerve level between living and PM in MD, AD, RD.
reproducibility. The effect of temperature was measured in the muscle rather than in nerves as it was expected that the muscle provided a more robust measurement. Due to the larger area, the variability of the diffusion parameters will be smaller and will suffer less from partial volume effects in comparison to nervous tissue. The observed differences in RD between the first and last scan are in line with the fact that an increase in temperature leads to an increased diffusion. However, this difference was relatively low and had a negligible effect on the diffusion parameter estimates and FT results. Despite the temperature differences between the PM subjects, FT and diffusion parameters were reproducible as reflected by the low range of limits of agreement (Fig. 3) and the unchanged architecture of the nerves over days (Fig. 4). This implies that degeneration processes were limited due to cooling of the PM subjects shortly after death, retarding the cellular breakdown. In general, diffusion parameters are affected by the breakdown of cellular membranes, but this is limited the first few days after death when the body is preserved in a cooled environment. A previous report showed that leaving tissue unfixed led to deleterious effects on diffusion properties in brain tissue over time [15]. Diffusion values may decrease further when the PM subject is preserved in the morgue for a longer period of time. However, the same authors also reported that diffusion parameters in the brain decreased in the first three days when cooled immediately after death [15]. We were unable to confirm this finding based on the data of our study.
144
W. Haakma et al. / Forensic Science International 263 (2016) 139–146
Fig. 4. Post-mortem fiber tractography (FT) results of a subject which was scanned four days in a row. A, B, and C show the FT results obtained on day 1, 2 and 3, respectively. F, G, and H show the matching T2-weighted images of the spinal cord. D shows the FT results after autopsy on the fourth day. I shows the cauda equina exposed during autopsy where L1–L4 were removed (E and I). Note that there are small air cavities in the spinal cord (I) indicated with red arrows which were not present during the first three scans (F, G, and H). Color-encoding is the same as in Fig. 2. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
There were differences in PM FT and diffusion parameters between post and pre autopsy. However, the exact mechanism explaining these differences is unknown. Possible explanations can be related to different processes of which we discuss three examples. First, mechanical (e.g. due to force applied to remove L1–L4) and morphological (e.g. reduction in cerebrospinal fluid due to leakage) processes might change FT and diffusion parameters. Secondly, these differences may be related to small air cavities observed in the spine and lower part of the abdomen after the autopsy, due to leakage of the cerebrospinal fluid from the spinal
cord. Fig. 4F–H shows the spinal cord fully intact, where Fig. 4I shows small air cavities in the spinal cord. These air cavities lead to susceptibility artifacts with subsequent adverse effects on the FT results. Lastly, it is likely that the observed temperature changes in the subjects have an effect on the diffusion parameters. PM FA values were comparable to in vivo results. FA values describe the ratio between AD and RD values. As the ratio between AD and RD remains relatively unchanged in a PM setting, the anisotropic properties of the nervous tissue PM are still intact. Schmidt et al. investigated the brain with diffusion weighted
Fig. 5. Fiber tractography results of (A) a post-mortem (PM) subject, (B) a PM case with injury to the lower vertebra, and (C) a living subject. The architecture of the lumbosacral nerves PM is comparable to living subjects. The traumatic case (B) shows assymetry and disorganization at the level of the damaged cauda equina. Colorencoding is the same as in Fig. 2. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
W. Haakma et al. / Forensic Science International 263 (2016) 139–146
imaging (DWI) and found MD values which were approximately 3–4 times lower than in vivo [23], which was similar to our results. Part of the observed differences in MD, AD and RD between PM subjects and in living nervous tissue are presumed to be caused by temperature differences [18]. A drop in temperature also influences the viscosity of liquids and could therefore be another explanation for our results [38]. However, a previous study showed that when correcting for temperature between PM and in vivo results, differences were still present [20]. Therefore, there might be other processes which could explain these differences. For practical reasons the PM subjects were scanned on a 1.5 T scanner and the living subjects on a 3 T scanner. Furthermore, different b-values were used in the in vivo and PM protocol. A different field strength and b-value may induce differences in calculated diffusion parameters [39,40]. The similar results in nerve architecture between PM and in vivo suggest that PMDTI is a reproducible method to visualize the 3D architecture of nerves. The results from the traumatic case with chronic traumatic lesions showed asymmetry and disorganization of the lumbosacral nerves compared to the non-trauma FT results. This was in accordance with the clinical condition of the deceased subject (Table 1) and may be explained by the chronic changes in nervous tissue during a span of years. In future research, more pathological chronic trauma cases should be evaluated to support the hypothesis that PMDTI and FT are able to identify pathology and trauma. Furthermore, acute trauma cases should be included to investigate to what extent PMDTI and FT are able to identify acute lesions of peripheral nervous tissue. However, as this case is comparable to results found in spina bifida patients [26], we expect that PMDTI and FT may advantageously be used to detect and document nerve disorders. Although the sample size of this study was small, the matching FT results with living subjects illustrate the potential of PMDTI and FT in investigating peripheral nervous structures after death. The relatively low spatial resolution used in the DTI sequences is likely affecting the measured diffusion parameters, especially in small nerve roots [41,42]. In this study, a single-shot EPI sequence was used in both in vivo and PM studies for comparison, but the PM protocol could benefit from using multi-shot EPI sequence as it allows the possibility to increase resolution or to reduce distortions [43]. However, such sequence leads to prolonged scan time and the risk of augmented body temperature. The peripheral nervous system is a relatively neglected area in the discipline of forensic pathology which needs further attention in order to elucidate its significance. Although the nervous system is infrequently evaluated during a normal autopsy, visualization of these structures could provide important insights about nerve damage, particular in traumatic cases. Additionally, PMDTI provides the opportunity to identify other peripheral nerve pathologies [21] which can be helpful in understanding pathogenesis and disease progression as it can be compared with histology. In conclusion, this study demonstrates that PMDTI and FT are reproducible methods to investigate the architecture and microstructure of the lumbar and sacral nerves in PM subjects. We showed the potential of PMDTI and FT to detect nerve trauma in one case with chronic lesions. Therefore, PMDTI and FT may become valuable in identification and documentation of PM nerve trauma or pathologies in forensic sciences.
Grant support The research of A.L. is supported by VIDI Grant 639.072.411 from the Netherlands Organisation for Scientific Research (NWO).
145
Acknowledgements We would like to thank the hospital porters at Skejby Sygehus and Lidy Kuster for their help in performing the MRI examinations.
References [1] S. Grabherr, B.A. Stephan, U. Buck, S. Na¨ther, A. Christe, L. Oesterhelweg, et al., Virtopsy – radiology in forensic medicine, Imaging Decis. MRI 11 (2007) 2–9. [2] I.S.D. Roberts, R.E. Benamore, E.W. Benbow, S.H. Lee, J.N. Harris, A. Jackson, et al., Post-mortem imaging as an alternative to autopsy in the diagnosis of adult deaths: a validation study, Lancet 379 (2012) 136–142. [3] J. An˜on, L. Remonda, A. Spreng, E. Scheurer, G. Schroth, C. Boesch, et al., Traumatic extra-axial hemorrhage: correlation of postmortem MSCT, MRI, and forensicpathological findings, J. Magn. Reson. Imaging 28 (2008) 823–836. [4] P.M. Leth, H. Struckmann, J. Lauritsen, Interobserver agreement of the injury diagnoses obtained by postmortem computed tomography of traffic fatality victims and a comparison with autopsy results, Forensic Sci. Int. 225 (2013) 15–19. [5] R. Dirnhofer, C. Jackowski, P. Vock, K. Potter, M.J. Thali, VIRTOPSY: minimally invasive, imaging-guided virtual autopsy, Radiographics 26 (2006) 1305–1333. [6] S. Grabherr, E. Gygax, B. Sollberger, S. Ross, L. Oesterhelweg, S. Bolliger, et al., Twostep postmortem angiography with a modified heart–lung machine: preliminary results, AJR Am. J. Roentgenol. 190 (2008) 345–351. [7] C. Bruguier, P.J. Mosimann, P. Vaucher, A. Uske´, F. Doenz, C. Jackowski, et al., Multiphase postmortem CT angiography: recognizing technique-related artefacts and pitfalls, Int. J. Legal Med. 127 (2013) 639–652. [8] A. Christe, S. Ross, L. Oesterhelweg, D. Spendlove, S. Bolliger, P. Vock, et al., Abdominal trauma – sensitivity and specificity of postmortem noncontrast imaging findings compared with autopsy findings, J. Trauma 66 (2009) 1302–1307. [9] A.H. Thomsen, A.G. Jurik, L. Uhrenholt, A. Vesterby, An alternative approach to Computerized Tomography (CT) in forensic pathology, Forensic Sci. Int. 183 (2009) 87–90. [10] S. Thayyil, N.J. Sebire, L.S. Chitty, A. Wade, W. Chong, O. Olsen, et al., Post-mortem MRI versus conventional autopsy in fetuses and children: a prospective validation study, Lancet 382 (2013) 223–233. [11] D.K. Jones, A. Leemans, Diffusion tensor imaging, in: M. Modo, J.W.M. Bulte (Eds.), Magnetic Resonance Neuroimaging, Humana Press, 2011, pp. 127–144. [12] C. Beaulieu, The basis of anisotropic water diffusion in the nervous system – a technical review, NMR Biomed. 15 (2002) 435–455. [13] S. Mori, P.C.M. van Zijl, Fiber tracking: principles and strategies – a technical review, NMR Biomed. 15 (2002) 468–480. [14] D. Le Bihan, J.F. Mangin, C. Poupon, C.A. Clark, S. Pappata, N. Molko, et al., Diffusion tensor imaging: concepts and applications, J. Magn. Reson. Imaging 13 (2001) 534–546. [15] H. D’Arceuil, A. de Crespigny, The effects of brain tissue decomposition on diffusion tensor imaging and tractography, Neuroimage 36 (2007) 64–68. [16] A. Pfefferbaum, E.V. Sullivan, E. Adalsteinsson, T. Garrick, C. Harper, Postmortem MR imaging of formalin-fixed human brain, Neuroimage 21 (2004) 1585–1595. [17] K.L. Miller, C.J. Stagg, G. Douaud, S. Jbabdi, S.M. Smith, T.E. Behrens, et al., Diffusion imaging of whole, post-mortem human brains on a clinical MRI scanner, Neuroimage 57 (2011) 167–181. [18] H.E. D’Arceuil, S. Westmoreland, A.J. de Crespigny, An approach to high resolution diffusion tensor imaging in fixed primate brain, Neuroimage 35 (2007) 553–565. [19] S. Rane, T.Q. Duong, Comparison of in vivo and ex vivo diffusion tensor imaging in rhesus macaques at short and long diffusion times, Open Neuroimaging J. 5 (2011) 172–178. [20] E. Scheurer, K.O. Lovblad, R. Kreis, S.E. Maier, C. Boesch, R. Dirnhofer, et al., Forensic application of postmortem diffusion-weighted and diffusion tensor MR imaging of the human brain in situ, Am. J. Neuroradiol. 32 (2011) 1518–1524. [21] K. Schmierer, C.A.M. Wheeler-Kingshott, P.A. Boulby, F. Scaravilli, D.R. Altmann, G.J. Barker, et al., Diffusion tensor imaging of post mortem multiple sclerosis brain, Neuroimage 35 (2007) 467–477. [22] P.M. Flach, S. Schroth, W. Schweitzer, G. Ampanozi, J. Slotboom, C. Kiefer, et al., Deep into the fibers! Postmortem diffusion tensor imaging in forensic radiology, Am. J. Forensic Med. Pathol. 36 (2015) 153–161. [23] T.M. Schmidt, R. Fischer, S. Acar, M. Lorenzen, A. Heinemann, U. Wedega¨rtner, et al., DWI of the brain: postmortal DWI of the brain in comparison with in vivo data, Forensic Sci. Int. 220 (2012) 180–183. [24] K. Yen, J. Weis, R. Kreis, E. Aghayev, C. Jackowski, M. Thali, et al., Line-scan diffusion tensor imaging of the posttraumatic brain stem: changes with neuropathologic correlation, Am. J. Neuroradiol. 27 (2006) 70–73. [25] S. Winklhofer, C.T. Stoeck, N. Berger, M. Thali, R. Manka, S. Kozerke, et al., Postmortem cardiac diffusion tensor imaging: detection of myocardial infarction and remodeling of myofiber architecture, Eur. Radiol. 24 (2014) 2810–2818. [26] W. Haakma, P. Dik, B. Ten Haken, M. Froeling, R.A.J. Nievelstein, I. Cuppen, et al., Diffusion tensor magnetic resonance imaging and fiber tractography of the sacral plexus in children with spina bifida, J. Urol. 192 (2014) 927–933. [27] D.C. Karampinos, G. Melkus, T.M. Shepherd, S. Banerjee, E.U. Saritas, A. Shankaranarayanan, et al., Diffusion tensor imaging and T2 relaxometry of bilateral lumbar nerve roots: feasibility of in-plane imaging, NMR Biomed. 26 (2013) 630–637.
146
W. Haakma et al. / Forensic Science International 263 (2016) 139–146
[28] T. Takahara, J. Hendrikse, T.C. Kwee, T. Yamashita, M. Van Cauteren, D. Polders, et al., Diffusion-weighted MR neurography of the sacral plexus with unidirectional motion probing gradients, Eur. Radiol. 20 (2010) 1221–1226. [29] P.K.N. van der Jagt, P. Dik, M. Froeling, T.C. Kwee, R.A. Nievelstein, B. ten Haken, et al., Architectural configuration and microstructural properties of the sacral plexus: a diffusion tensor MRI and fiber tractography study, Neuroimage 62 (2012) 1792–1799. [30] A. Leemans, B. Jeurissen, J. Sijbers, D.K. Jones, ExploreDTI: a graphical toolbox for processing, analyzing, and visualizing diffusion MR data, Proc. Intl. Soc. Magn. Reson. Med. 17 (2009) 3536. [31] A. Leemans, D.K. Jones, The B-matrix must be rotated when correcting for subject motion in DTI data, Magn. Reson. Med. 61 (2009) 1336–1349. [32] M.O. Irfanoglu, L. Walker, J. Sarlls, S. Marenco, C. Pierpaoli, Effects of image distortions originating from susceptibility variations and concomitant fields on diffusion MRI tractography results, Neuroimage 61 (2012) 275–288. [33] C.M. Tax, W.M. Otte, M.A. Viergever, R.M. Dijkhuizen, A. Leemans, REKINDLE: robust extraction of kurtosis INDices with linear estimation, Magn. Reson. Med. 73 (2014) 794–808. [34] J. Veraart, J. Sijbers, S. Sunaert, A. Leemans, B. Jeurissen, Weighted linear least squares estimation of diffusion MRI parameters: strengths, limitations, and pitfalls, Neuroimage 81 (2013) 335–346.
[35] P.J. Basser, S. Pajevic, C. Pierpaoli, J. Duda, A. Aldroubi, In vivo fiber tractography using DT-MRI data, Magn. Reson. Med. 44 (2000) 625–632. [36] J.M. Bland, D.G. Altman, Statistical methods for assessing agreement between two methods of clinical measurement, Lancet 1 (1986) 307–310. [37] M. Jones, A. Dobson, S. O’Brian, A graphical method for assessing agreement with the mean between multiple observers using continuous measures, Int. J. Epidemiol. 40 (2011) 1308–1313. [38] H. Ozbek, Viscosity of aqueous sodium chloride solutions from 0 to 150 8C, in: 29th Southeast Regional Meeting of the American Chemical Society, 1971, 1–68. [39] D.L. Polders, A. Leemans, J. Hendrikse, M.J. Donahue, P.R. Luijten, J.M. Hoogduin, Signal to noise ratio and uncertainty in diffusion tensor imaging at 1.5, 3.0, and 7.0 T, J. Magn. Reson. Imaging 33 (2011) 1456–1463. [40] E.S. Hui, M.M. Cheung, K.C. Chan, E.X. Wu, B-value dependence of DTI quantitation and sensitivity in detecting neural tissue changes, Neuroimage 49 (2010) 2366– 2374. [41] Q. Hogan, Size of human lower thoracic and lumbosacral nerve roots, Anesthesiology 85 (1996) 37–42. [42] S.B. Vos, D.K. Jones, M.A. Viergever, A. Leemans, Partial volume effect as a hidden covariate in DTI analyses, Neuroimage 55 (2011) 1566–1576. [43] S. Chabert, C. Galindo, C. Tejos, S.A. Uribe, Multiple echo multi-shot diffusion sequence, J. Magn. Reson. Imaging 39 (2014) 1027–1032.