In vivo segmentation of the human locus coeruleus
using 7T MRI
Renée San Giorgi 5929105 25-08-2013 Cognitive Science Center Amsterdam Internship supervisor: M.C. Keuken Co-assessor: B.U. Forstmann
Abstract
The Locus Coeruleus (LC) is a nucleus in the brainstem with diffuse cortical and sub-cortical connections and is involved in the modulation of attentional states. For further investigation into the functional role of the LC, there is need for a high resolution in vivo probabilistic atlas. High resolution MP2RAGE magnetic resonance imaging (resolution of 0.6 mm isotropic) data is used to map the LC manually. The inter-rater agreement was assessed and found to be too low for the MP2RAGE sequence to be useful for making a probabilistic atlas of the LC. Further implications of this finding are discussed.
Introduction
The locus coeruleus (LC) is a mammalian brainstem nucleus located in the pons, first described in 1809 (Reil, 1809) and named in 1812 (Wenzel & Wenzel, 1812). The name refers to its blue (coeruleus = cerulean) appearance in unstained tissue. It is a long, thin, ‘tube like’ shaped nucleus (German, Walker, Manaye, Smith, & Woodward, 1988). Reports of the length of the structure are between 13 mm and 17 mm in the rostrocaudal direction (Chan-Palay, 1989; German et al., 1988;
Hoogendijk, Pool, Troost, van Zwieten, & Swaab, 1995) in healthy subjects. It is anterolateral to the floor of the fourth ventricle (Sasaki et al., 2006; Shibata et al., 2006) and it extends to the inferior colliculi (Keren, Lozar, Harris, Morgan, & Eckert, 2009). The volume of the LC is reported between 10.5 and 18.5 mm³ (Hoogendijk et al., 1999, 1995).
Neuromelanin, the pigment responsible for giving the structure its colour, is produced by noradrenergic neurons (German et al., 1988; Keren et al., 2009). The LC produces approximately 50% of the total amount of norepinephrine (NE)
produced in the brain (Carter, De Lecea, & Adamantidis, 2013; Kandel, Schwartz, & Jessell, 2000). This NE is delivered throughout a wide area of the central nervous system (CNS) via two major fibre systems; a descending noradrenergic bundle and the ascending rostral limb of the dorsal periventricular pathway (Kandel et al., 2000; Klimek et al., 1999). The descending pathway reaches the cerebellum and the dorsal and ventral horns of the spinal cord. The ascending pathway reaches subcortical regions, such as the thalamus and the hypothalamus, but also all regions of the cerebral cortex. Approximately 12,000 noradrenergic neurons in the LC make very diffuse connections with almost the entire brain: just a single noradrenergic LC neuron can have more than 250,000 synapses and the axons of one neuron can branch in both the cerebral and the cerebellar cortex (Bear, Connors, & Paradiso, 2007).
There are indications that the LC is involved in the modulation of attentional states (Aston-Jones, Rajkowski, & Cohen, 2000), levels of arousal (Carter et al., 2013), and sleep-wake cycles (Bear et al., 2007) through the locus coeruleus-norepinephrine (LC-NE) system. It modulates responsiveness to novel stimuli by influencing both arousal by modulating activity in the forebrain and modulating
sensory perception in the brain stem and spinal cord (Kandel et al., 2000) The LC has been reported to be affected in several neurodegenerative diseases. For example, Alzheimer is characterized by abnormalities that selectively affect neurons in specific brain regions, including the LC. The degeneration of the LC, and with it the major source of NE, is closely linked to the progression of the symptoms of Alzheimer in mice (Heneka et al., 2010). A recent ex vivo study shows that the neuronal loss in Alzheimer patients in the LC is even greater than neuronal loss in the substantia nigra (Zarow, Lyness, Mortimer, & Chui, 2003). By investigating the role of the LC in healthy subjects, the role of the LC in diseases like Alzheimer might be understood better and might lead to better treatment. Most previous research has been done ex vivo, since the LC is not easy to find with functional magnetic resonance imaging (fMRI). In vivo exploration of the LC may shed more light on the role it plays in Alzheimer. For example, such studies could investigate whether certain morphological properties (such as size) relate to different stages of Alzheimer.
In order to explore the functional properties of the LC, for example using functional magnetic resonance imaging (fMRI), it is necessary that a reliable
probabilistic atlas of the structure is available. Ex vivo, the LC is easy to locate as the pigmentation makes it stand out from other brain structures. But in order to make an atlas based on in vivo MR images, there are some obstacles that make the LC a difficult structure to map. First, due to the shape and size of the nucleus, partial volume is especially problematic. When mapping larger brain structures it would only affect the borders of the structure. However, the LC is so small and narrow that, with standard MRI scanners, almost all voxels containing a LC signal will be voxels that contain both the LC and the surrounding tissue. Therefore, a larger portion of the
signal of this structure might get lost than when mapping larger brain structures. Second, many other nuclei and nerve fibres are close by. For example, the nucleus trochlearis, nucleus dorsalis tegmenti and nucleus tractus mesencephalicus trigemini surround the LC at the level of the inferior colliculi (Mai & Paxinos, 2008; Naidich et al., 2009). Moreover, most structures in the area surrounding the LC have the same longitudinal shape following the rostrocaudal axis. This makes it harder to
differentiate between the LC and surrounding structures on shape alone. Third, the LC is adjacent to the fourth ventricle. The fourth ventricle is filled with cerebrospinal fluid, and its signal causes interference with the signal of nearby structures
(McRobbie, 2007).
There are several ex vivo atlases that include the LC (Kahle, Leonhardt, & Platzer, 2005; Mai & Paxinos, 2008; Naidich et al., 2009). Disadvantages of these atlases are that they describe an individual and are therefore not probabilistic. Another problem with these atlases is that they do not contain several consecutive slices showing the LC. Finally, as these atlases are based on ex vivo tissue, they might be distorted due to shrinkage (Bobinski et al., 2000). The first in vivo atlas of the human LC was created in 2009 by (Keren et al., 2009). However, the methods used to create that atlas should be viewed carefully. (Keren et al., 2009) reported a volume estimate of 46.5 mm³, which is considerably larger than previous reports. This might be explained by how the atlas was created. They did not make the atlas probabilistic in such a way that each voxel is registered to have a certain probability of being in the LC. Instead of registering all voxels containing LC signal, they only registered a single voxel; the voxel with the highest signal intensity per participant. The authors chose this method because a 3T MRI scanner was at their disposal and the resolution was not high enough to map the whole LC reliably. However, this atlas
does not contain information about the shape or size of the LC. It only gives information about where in the brain the point with the highest signal of the LC is located. If the area in this atlas is large, it only means that the individual difference in that specific location is high.
Therefore, an atlas mapping the entire structure and not only the point with the highest signal will be more valuable since it also gives information about the size and shape of the entire LC. By using a 7T MRI scanner to acquire the required data, the resolution can be high enough to map the LC manually. Only after this high resolution probabilistic atlas of the LC is created, it is possible to link functional data to the location of the LC. In this research, 7T MRI data using several sequences are studied and used to attempt to locate and segment the LC.
Methods
Participants
30 healthy participants (14 female) with a mean age of 24.2 (SD 2.4)
participated in this study. All participants had normal or corrected-to-normal vision, and none of them had a history of neurological, major medical, or psychiatric disorders. All subjects were right-handed, as confirmed by the Edinburgh Inventory (Oldfield, 1971).The study was approved by the local Leipzig University Hospital ethics committee. All participants gave written informed consent.
MRI sequences
Structural scans were obtained using a 7T Siemens Magnetom MRI system with three sequences: MP2RAGE (Marques et al., 2010) (repetition time TR = 5000
ms; inversion times TI1, TI2 = 900/2750 ms) with 0.7 mm (240 sagittal slices) and 0.6 mm (slab of 128 slices) isotropic resolution, and FLASH. The FLASH slab (0.5 mm isotropic voxels) had 128 slices (TR = 41 ms with three echo times (TE): 11 / 20 / 30 ms. The MP2RAGE and FLASH slabs were parallel to the AC-PC line. Only the MP2RAGE slab scans were used for manual segmentation, since the LC was not visible using either of the two other sequences, either due to voxel size or tissue properties of the LC.
Procedure
Manual segmentation was carried out by two independent researchers using the FSL 4.1.4 viewer. Inter-rater agreement was assessed using the intra-class correlation coefficient (ICC) for brain volume (Shrout & Fleiss, 1979) and Cohen’s kappa (Cohen, 1960), a statistical measure of inter-rater agreement for qualitative items. Masks defining the LC were delineated with the aid of the T1-weighted slab MP2RAGE images. To register these masks to MNI standard space, they were first aligned to the MP2RAGE whole-brain image of the same subject, before normalizing to the 0.4 mm3 MNI template as implemented in the CBS high-resolution toolbox (http://www.nitrc.org/projects/cbs-tools/). All registration steps were done linearly using MIPAV (www.mipav.cit.nih.gov).
Before the segmentation procedure, a segmentation protocol has been set up, based on previous literature. Segmentation was done manually with FSLView by two experimenters. Afterwards, the inter-rater was calculated and only the voxels that both experimenters agreed on were used for further purposes. Then, a probability map has been registered to MNI space.
Segmentation protocol
The tube-like shape of the LC reaches roughly from the inferior colliculi to the inferior end of the pons. Its slightly tilted position – with the superior end more medial and the inferior end more lateral – combined with its thin, longitudinal form makes the LC hard to spot in the sagittal and coronal view. It might be partly visible but the longitudinal shape is interrupted because of the tilted angle of the structure. Therefore, the main view used for segmentation was the axial view. The most
important landmark is the fourth ventricle. The LC is located anterolateral to the floor of the fourth ventricle. After the researcher segments the LC from the axial view, he or she uses the sagittal and coronal view to verify if the mask is connected and creates a fluent, tube-like shape. If the LC is not visible in some axial slices, this can be solved by connecting separate masks in the sagittal and coronal view. The researcher takes care that the structure that is mapped is not too far from the fourth ventricle (namely, more than +/- 2 mm).
Results
The inter-rater values are shown in Table 1. To compare the mean volume of the LC based on the current data with the mean volume found in previous literature, a one-sampled t-test was carried out. The volumes of the current data were found to be significantly smaller than the volumes reported by (Keren et al., 2009) (5.4 mm² versus 46.5 mm²; t(29), p < 0.001) and (Hoogendijk et al., 1995) (5.4 mm² versus 10.5 mm², t(29), p < 0.001).
Table 1: Means and rage of Interrater volume, Cohen’s kappa, Dice’s coefficient & Intra Class Coefficient
Interrater volume Cohen's Kappa
Dice's
coefficient Intra Class Coefficient
Mean (±SD) left 8.9 (±6.3) 0.26 (±0.20) 26.3 (±20.3) 0.70 (±0.08) Mean (±SD) right 1.9 (±1.3) 0.38 (±0.20) 37.6 (±20.1) 0.75 (±0.08) Mean (±SD) left & right 5.4 (±3.8) 0.32 (±0.20) 31.9 (20.8) 0.72 (±0.08)
Range left 0 – 24 0 - 0.65 0 - 65.1 0.53 - 0.83
Range right 0 - 5.2 0 - 0.69 0 - 68.6 0.59 - 0.88 Range left & right 0 – 24 0 - 0.69 0 - 68.6 0.53 - 0.88
Figure 1 shows two examples of the MP2RAGE images used for manual segmentation. The LC is the bright spot circled in red, anterolateral of the fourth ventricle (indicated with a black arrow). It is poorly visible at best, and in some participants it is unrecognizable. The maximal inter-rater overlap was 13.3% and the minimal overlap was 0% (no overlap at all).
Figure 1.
A sagittal view of the pons, at the level of the LC. The LC is circled in red, the fourth ventricle (V4) is indicated with a black arrow. Note that the LC is poorly visible, as it cannot or hardly be seen on the contralateral side.
atlas of the LC are listed in table 2.
Table 2: Montreal Neurological Institute (MNI) space coordinates of peak and centre of gravity. MNI space coordinates
X y z
Peak coordinates left -6.70 -37.98 -29.73
right 6.90 -37.98 -26.13
Centre of gravity left -5.50 -36.78 -27.33
right 6.5 -36.78 -25.73
The correlation between the left and right hemisphere of the probabilistic atlas of the LC was calculated using a paired two-sample t-test. The inter-rater volumes of the left and right hemispheres correlated significantly (rho = 0.9979 p = 0.000). However, a t-test showed that the inter-rater volumes of the left (mean = 8.9 mm²) and right (mean = 1.9 mm²) differed significantly (p < 0.000).
Discussion
Existing atlases of the LC are not reliable tools if one wishes to investigate the functional role of the LC using fMRI. This is due to the fact that these atlases are created using ex vivo material, or low resolution in vivo data. The aim of this research was to create a probabilistic atlas of the LC based on high resolution MRI data. However, since the maximal inter-rater overlap was 13.3 %, it can be concluded that using a MP2RAGE sequence is not useful for making the LC visible. The resolution (0.6 mm² isotropic) should be high enough, based on previous reports on the volume of the LC.
Two previous studies (Keren et al., 2009; Sasaki et al., 2006) used the T1-weighted turbo spin-echo (T1-TSE) sequence using a 3T MRI scanner and received results on the location of the LC that matched ex vivo coordinates. This is the only
reported MRI sequence that is known to be reliably used to locate the LC. If this same sequence is used with a 7T MRI scanner with in vivo subjects, it is expected to be possible to create a reliable probabilistic atlas using manual segmentation.
Although the T1-TSE sequence is very useful for locating the LC, some questions still need to be answered. It is reported that it exploits the presence of neuromelanin, which is produced by the noradrenergic neurons (Keren et al., 2009). However, it is not entirely clear which exact physical properties of the LC, and its neuromelanin, result in the T1-TSE being a useful sequence. Further research should therefore also be directed to answer the question why the LC can only be viewed using the T1-TSE sequence and, for example, not by using a T2* sequence that could focus on the ferrous properties of the LC.
Several difficulties are encountered when locating the LC using in vivo MRI data, and they have been described in previous sections in this report. When taking these difficulties into account, it becomes clear that previous published in vivo studies concerning the LC should be viewed with extreme caution. Currently, there are no reliable probabilistic atlases of the LC available. So before any functional or
connectivity studies can be executed, the probabilistic location of the LC should be explored first.
Literature
Aston-Jones, G., Rajkowski, J., & Cohen, J. (2000). Locus coeruleus and regulation of behavioral flexibility and attention. Progress in brain research, 126(215), 165– 82. doi:10.1016/S0079-6123(00)26013-5
Bear, M. F., Connors, B. W., & Paradiso, M. A. (2007). Neuroscience: Exploring the
Bobinski, M., Leon, M. J. D. E., Wegiel, J., Desanti, S., Convit, A., Louis, L. A. Saint, & Rusinek, H. (2000). THE HISTOLOGICAL VALIDATION OF POST MORTEM MAGNETIC RESONANCE IMAGING-DETERMINED HIPPOCAMPAL VOLUME IN ALZHEIMER ’ S DISEASE, 95(3), 721–725.
Carter, M. E., De Lecea, L., & Adamantidis, A. (2013). Functional wiring of hypocretin and LC-NE neurons: implications for arousal. Frontiers in behavioral
neuroscience, 7(May), 43. doi:10.3389/fnbeh.2013.00043
Chan-Palay, V. (1989). Alterations in catecholamine neurons in the locus coeruleus in dementias of Alzheimer ’ s and Parkinson ' s diseases, 19–20.
Cohen, J. (1960). A Coefficient of Agreement for Nominal Scales. Educational and
Psychological Measurement, 20(1), 37–46.
Forstmann, B. U., Keuken, M. C., Jahfari, S., Bazin, P.-L., Neumann, J., Schäfer, A., Anwander, A., et al. (2012). Cortico-subthalamic white matter tract strength predicts interindividual efficacy in stopping a motor response. NeuroImage,
60(1), 370–5. doi:10.1016/j.neuroimage.2011.12.044
German, D. C., Walker, B. S., Manaye, K., Smith, W. K., & Woodward, D. J. (1988). The Human Distribution Locus Coeruleus : Computer Reconstruction of Cellular,
8(May).
Heneka, M. T., Nadrigny, F., Regen, T., Martinez-Hernandez, A., Dumitrescu-Ozimek, L., Terwel, D., Jardanhazi-Kurutz, D., et al. (2010). Locus ceruleus controls Alzheimer’s disease pathology by modulating microglial functions through norepinephrine. Proceedings of the National Academy of Sciences of
the United States of America, 107(13), 6058–63. doi:10.1073/pnas.0909586107
Hoogendijk, W. J., Pool, C. W., Troost, D., Van Zwieten, E., & Swaab, D. F. (1995). Image analyser-assisted morphometry of the locus coeruleus in Alzheimer’s disease, Parkinson's disease and amyotrophic lateral sclerosis. Brain, 118(1), 131–143.
Hoogendijk, W. J., Sommer, I. E., Pool, C. W., Kamphorst, W., Hofman, M. a, Eikelenboom, P., & Swaab, D. F. (1999). Lack of association between depression and loss of neurons in the locus coeruleus in Alzheimer disease.
Archives of general psychiatry, 56(1), 45–51. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/9892255
Kahle, W., Leonhardt, H., & Platzer, W. (2005). Sesam atlas van de anatomie. Deel
3. Zenuwstelsel en zintuigen (18th ed.). Baarn: Bosch & Kreuningen.
Kandel, E. R., Schwartz, J. H., & Jessell, T. M. (2000). Principles of neural science (4th ed.). McGraw-Hill.
Keren, N. I., Lozar, C. T., Harris, K. C., Morgan, P. S., & Eckert, M. A. (2009). In vivo mapping of the human locus coeruleus. NeuroImage, 47(4), 1261–7.
Klimek, V., Rajkowska, G., Luker, S. N., Dilley, G., Meltzer, H. Y., Overholser, J. C., Stockmeier, C. a, et al. (1999). Brain noradrenergic receptors in major
depression and schizophrenia. Neuropsychopharmacology : official publication
of the American College of Neuropsychopharmacology, 21(1), 69–81.
doi:10.1016/S0893-133X(98)00134-1
Mai, J. K., & Paxinos, G. (2008). Atlas of the Human Brain (3rd ed.). Academic Press. Marques, J. P., Kober, T., Krueger, G., Van der Zwaag, W., Van de Moortele, P. F., &
Gruetter, R. (2010). MP2RAGE, a self bias-field corrected sequence for improved segmentation and T1-mapping at high field. NeuroImage2, 49(2), 1271–1281.
McRobbie, D. (2007). MRI from Picture to Proton (2nd ed.). Cambridge University Press. Retrieved from http://books.google.com/books?hl=en&lr=&id=cw8_zC-Ula4C&oi=fnd&pg=PR11&dq=MRI+From+Picture+to+Proton&ots=m3UNTO4TG A&sig=cpimY6wk8KLJ6VI9TcNpKXWn84s
Naidich, T. P., Duvernoy, H. M., Delman, B. N., Sorensen, A. G., Kollias, S. S., & Haacke, E. M. (2009). Duvernoy’s atlas of the human brain stem and
cerebellum. Wien: SpringerWienNewYork.
Oldfield, R. C. (1971). The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia, 9(1), 97–113. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/5146491
Reil, J. C. (1809). Untersuchungen über den Bau des grossen Gehirn im Menschen.
Arch Physiol, 9, 136–208.
Sasaki, M., Shibata, E., Tohyama, K., Takahashi, J., Otsuka, K., Tsuchiya, K.,
Takahashi, S., et al. (2006). Neuromelanin magnetic resonance imaging of locus ceruleus and substantia nigra in Parkinson’s disease. Neuroreport, 17(11), 1215–8. doi:10.1097/01.wnr.0000227984.84927.a7
Shibata, E., Sasaki, M., Tohyama, K., Kanbara, Y., Otsuka, K., Ehara, S., & Sakai, A. (2006). Age-related changes in locus ceruleus on neuromelanin magnetic
resonance imaging at 3 Tesla. Magnetic resonance in medical sciences :
MRMS : an official journal of Japan Society of Magnetic Resonance in Medicine, 5(4), 197–200. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/17332710
Shrout, P. E., & Fleiss, J. L. (1979). Intraclass Correlations: Uses in Assessing Rater Reliability. Psychological Bulletin, 86(2), 420–428.
Wenzel, J., & Wenzel, C. (1812). De Penitiori Structura Cerebri Hominis et Brutorum. Tübingen.
Zarow, C., Lyness, S. a, Mortimer, J. a, & Chui, H. C. (2003). Neuronal loss is greater in the locus coeruleus than nucleus basalis and substantia nigra in Alzheimer and Parkinson diseases. Archives of neurology, 60(3), 337–41. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/12633144
