Show Summary Details
Page of

Anatomical Validation of DTI and Tractography 

Anatomical Validation of DTI and Tractography
Chapter:
Anatomical Validation of DTI and Tractography
Author(s):

Nigel I. Lawes C.

and Christopher A. Clark

DOI:
10.1093/med/9780195369779.003.0026
Page of

PRINTED FROM OXFORD MEDICINE ONLINE (www.oxfordmedicine.com). © Oxford University Press, 2016. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in Oxford Medicine Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: null; date: 18 December 2018

Summary Box

  • The factors causing problems in diffusion tensor imaging (DTI) tractography are specific to each tract and specific to each species, so transfer of results from other species to humans or from one tract to another cannot be relied upon.

  • Conventional hodological techniques cannot be used in humans for ethical reasons, although they can validate tractography in other species.

  • In human brains, comparison of blunt dissection of fixed human brains gives comparable results to DTI tractography where tracts are relatively uncomplicated, but both techniques run into similar difficulties where tracts intersect or run closely together.

  • Electrical disruption of language produces a division of language pathways similar to DTI tractography and blunt dissection.

  • Electromyography appears to identify a larger corticospinal tract than that with tractography, but this is almost certainly due to inadequate placement of the regions of interest used to generate the pathway and to the use of a high FA threshold.

  • The somatotopic organization of corticostriatal hand, foot, and mouth pathways is different for functional MRI and DTI tractography; this is a difficult pathway fraught with problems caused by neighboring tracts.

  • Pathways such as cerebellothalamic tracts, the occipitofrontal fasciculus, geniculate nuclei, and somatotopy in the internal capsule have been investigated, generally with positive results.

  • In other species, Gd-DTPA contrast has been used to enhance the signal-to-noise ratio of DTI tractography in fixed primate brains; fiber reconstructions corresponded to known anatomy.

  • Manganese-enhanced T1-weighted images have been used to compare the direction of a tract with the principle eigenvector of DTI tractography; the directions were largely parallel except where tracts crossed or terminated.

  • Post-mortem tractography has been compared with in vivo injection of tracers and overlap functions of 0.58–0.98 were computed.

As we have already seen in this book, diffusion of water in white matter is rendered anisotropic by being hindered in directions perpendicular to fiber tracts (see Chapter 7 by Beaulieu for the biophysical underpinnings). The principal eigenvector of a tensor description of this anisotropy purports to identify axonal fiber tracts within the brain. The validity of this assumption can be tested directly in experimental animals using standard hodological techniques (hodology is the study of pathways).

A major problem in the validation of tractography in humans, however, is that there is no gold standard available. While gold standards are available in other animals, most of the invalidity of tractography derives from partial-volume and crossing-fiber issues (see Chapter 27 by Alexander and Seunarine and Chapter 28 by Tournier), which will be different for different tracts in the same species and different for the same tract in different species. Validation against a gold standard in one species, therefore, will have no implications for validation in another species, nor will validation of one tract carry any implications for validation of another tract. Furthermore, the human brain has developed some regions to a far greater extent than other species. This is particularly true for language pathways, for example, and for those areas that have expanded massively in the human brain, such as the prefrontal cortex and the temporal pole; in neither case will there be good animal models of the pathways involved.

Most hodological techniques used in neuroanatomy are concerned with determining which regions of the brain project to which other regions; they are much less concerned with the exact pathway taken. If principal eigenvectors are to be compared directly to histological specimens, however, further problems arise from the processes of fixation, dissection and sectioning, all of which cause a distortion of the tissue, tarnishing the “gold” of the standard. One solution to these dilemmas may be to use functional techniques to validate diffusion tensor imaging (DTI) tractography (Thiebaut de Schotten et al., 2005; Berman et al., 2007; Duffau, 2008). Electrical and magnetic stimulation, electroencephalography (EEG), and functional magnetic resonance imaging (fMRI) are examples of techniques that could be used as surrogate standards.

Studies in Humans: Direct Comparisons

Parker et al. (2001) used a fast marching tractography method to compare the optic radiation and corticopeduncular tracts with expectation in humans and macaques. Partial volume effects led to the inclusion of the anterior part of the occipitofrontal fasciculus and some spurious pathways with putative optic radiation in the macaque. This false-positive error was not evident in the human brain as a result of higher anatomical resolution in a bigger brain. Inspection of their Figure 6, however, reveals a substantial component of the putative optic radiation deviating to the dorsolateral surface instead of to the medial surface where it would be expected to end. This deviation suggests that the posterior part of the occipitofrontal fasciculus had been included after all. This underlies the need for validation: are these dorsolaterally terminating fibers a real component of the human optic radiation, or is this an error created by the close apposition of the occipitofrontal fasciculus?

Yoshiura et al. (2008) have used a three-dimensional directional diffusion function defined by a probability function of the local diffusion tensor to study the somatotopic organisation of the corticospinal tract within the posterior limb of the internal capsule. They refer to “cephalad to caudal” segments of the precentral gyrus, which appear to be dorsomedial and ventrolateral, respectively. These segments projected in a posterior-to-anterior distribution within the posterior limb of the internal capsule, but only in a proportion of patients. Although Yoshiura et al. claim that this corresponds to the known somatotopic organization of the corticospinal tract within the internal capsule, it is in reverse order to that given in standard neuroanatomical textbooks (Brodal, 1981; Nadeau et al., 2004). Given this discrepancy, the somatotopic organization found by them cannot be considered to validate their method. Oddly, the region in the precentral gyrus designated “caudal” contains the most cephalad part of the homunculus, whereas the region designated as “cephalad” contains the most caudal part of the homunculus present in the precentral gyrus. Perhaps their claim that their findings accord with the literature involves a misunderstanding along these lines.

Devlin et al. (2006) used probabilistic tractography to differentiate lateral from medial geniculate nuclei. They classified voxels from the diencephalic region based on a probability of connection exceeding 0.1%. A cross-correlation matrix of similarity between the connections of each voxel was then computed and voxels with high correlations forced to the diagonal. Clusters with high correlation were mapped back into anatomical space. The results identified both nuclei, which were displaced approximately one voxel from the locations given by high-resolution proton density weighted scanning optimized for subcortical gray–white contrast. Similar results were reported by Behrens et al. (2006) for differentiation of the supplementary motor area from pre-supplementary motor area. The DTI-based method was considerably faster in terms of acquisition times, a major issue if academic research is to be translated into clinical practice.

One approach to validation of tractography is to compare the results to dissection of formalin-fixed brains (Lawes et al., 2008). White matter can be teased apart with blunt dissection as sheets of axons have natural planes of cleavage parallel to the axons (see Chapter 3by Axer). Simple tracts in which axons remain parallel are easily dissected and correspond closely to the results of tractography. A dissection of the inferior occipitofrontal fasciculus corresponded closely to tractography obtained from the averaged DTI data from 15 male brains (Lawes et al., 2008) (see Fig. 26.1). Where the dissected tract curved, the tractography tract curved, and where the fibers changed direction, the direction-encoding color (DEC) assignment scheme of Pajevic and Pierpaoli (1999) showed a corresponding color.

Figure 26.1 Left inferior occipitofrontal fasciculus. i) Track representation of the separated components of the fasciculus from the mean DTI data set. Posteriorly, the fasciculus terminates in the inferior part of the middle occipital gyrus (LOMI), the lingual gyrus (LOLN), and the inferior occipital gyrus (LOIN). Anteriorly, the fasciculus terminates in the lateral part of the orbitofrontal cortex (LFOL) and the marginal gyrus of the frontal lobe (LFMG). ii) Tract representation of the whole fasciculus from the mean DTI. iii) The dissected occipitofrontal fasciculus. Letters in (ii) and (iii) highlight areas of similarity in the two images. iv) Location of the occipitofrontal fasciculus as determined from the mean DTI. v) Intersubject tract variability maps of the occipitofrontal fasciculus for image slices through standard space (all images are illustrated using the neurological convention). Variability maps are colored using a hot color map (yellow/opaque = 1.0 [all subjects have streamlines that pass through image voxel]; black/transparent = 0.0 [no subjects have streamlines that pass through image voxel]). From Lawes et al. (2008) NeuroImage 39:62–79 with permission from Elsevier.

Figure 26.1
i)ii)iii)iv)v)Lawes et al. (2008) NeuroImage 39:62–79 with permission from Elsevier.

Where the shape of the dissected tract changed, so too did the tractography-reconstructed tract (Fig. 26.1). As such, the similarities between the blunt dissection and the DTI reconstructions suggest that the fiber tractography accurately reconstructs the pathways of the inferior–occipital–frontal pathway. Appearances can be deceptive, however: where two tracts run closely together, both techniques run into similar difficulties (Fig. 26.1). The inferior occipitofrontal fasciculus runs alongside the lateral aspect of the optic radiation. In the dissection, some of the temporal fibers of the optic tract belonging to Meyer's loop came off with the occipitofrontal fasciculus. Interestingly, this same region of the optic tract shared voxels with the inferior occipitofrontal fasciculus and appeared in the DTI-reconstructed tract as well, in an identical location.

A novel tract between the temporal lobe and the occipital and parietal lobes, found only in the right hemisphere, was demonstrable in both the dissection and the tractography (Lawes et al., 2008). A fan-shaped tract between superior frontal gyrus and the pars triangularis of the inferior frontal gyrus also showed good correspondence between dissection and tractography, as far as the averaged DTI data set was concerned. An intersubject variability map was constructed, indicating differences and similarities between the individual brains. The intersubject variability map picked up a variable component that left the main tract on route to the pars triangularis and ascended to the middle frontal gyrus instead. Neither the dissection nor the tractography on the averaged DTI detected this variable component. Both dissection and tractography confirmed the posterior component of the arcuate fasciculus first described by Catani et al. (2005) and unpublished data confirmed the other components identified by the same authors as shown in Figure 26.2.

Figure 26.2 Left posterior segment of the arcuate fasciculus. i) Track representation from the mean DTI. Streamlines run from the posterior part of the middle temporal gyrus (LTMP) to the supramarginal gyrus and the angular gyrus (LPAN). LFIC, left inferior frontal gyrus; LTMM, middle part of the middle temporal gyrus. ii) Tract representation of the arcuate fasciculus terminating in the inferior part of the precentral gyrus (LFPI). iii) Location of the pathway as determined from the mean DTI. iv) Intersubject tract variability maps of the posterior segment of the arcuate fasciculus for image slices through standard space. From Lawes et al. (2008) NeuroImage 39:62–79 with permission from Elsevier.

Figure 26.2
i)ii)iii)iv)Lawes et al. (2008) NeuroImage 39:62–79 with permission from Elsevier.

Lawes et al. (2008) showed that when using tractography to reconstruct lemniscal sensory pathways, one reconstructs a single ascending trajectory from the medulla, through the thalamus, all the way to the cortex. However, strictly speaking, the tracts should terminate in the thalamus, and then a second-order connection should continue from the thalamus to the cerebral cortex. Thus, while tractography has the advantage that the whole medulla–thalamus–cortex system is displayed, it has the disadvantage that the individual components are not differentiated. In many ways, tractography shares the same combination of advantages and disadvantages as hodological techniques that involve transneuronal transport of a tracer (see Chapter 3), i.e., such methods will also not exclusively reveal first-order connections, and will naturally reveal second-order connections. Hence, one has to think carefully about the characteristics of standard hodological techniques before using them as a tool to “validate” DTI tractography.

All these tracts demonstrate that where the anatomy is simple, with parallel axons and no shared voxels or crossing fibers, tractography gives a close representation of the anatomy. Where fiber tracts are at least comparable in size to the dimensions of voxels, tractography corresponds to fiber orientation (Lazar et al., 2003). This is typically the case with association fibers passing anteroposteriorly within one hemisphere, such as the occipitofrontal fasciculus.

Difficulties arise, however, where the anatomy is not simple. As we have described above, if tracts run closely together but in different directions, both tractography and blunt dissection are prone to errors easily detected with sufficient a priori knowledge. This was the case with Meyer's loop and the occipitofrontal fasciculus (Lawes et al., 2008). Electrophysiology and hodological techniques give good evidence that the hand and face regions of sensorimotor cortex have strong descending connections with the brainstem and spinal cord. There appears to be a lack of evidence in the literature for these connections to be displayed by blunt dissection, although motor pathway architecture has been demonstrated using quantitative cytoarchitectonic and myeloarchitectonic image analysis (Rademacher et al., 2001). Streamline tractography, based on a single tensor model, however, consistently fails to reveal these connections. One assumption is that other pathways running anteroposteriorly conceal the pathways. Where fibers with different orientations occupy the same voxel, directional information is corrupted (Wiegell et al., 2000; Alexander et al., 2001; see also Chapters 27 and 28 in this book). In regions where sheets of axons running in different directions are interleaved, streamline tractography on DTI data fails to give an accurate representation of the anatomy. Thus almost all connections to the brainstem are confounded by shared voxel and crossing-fiber problems (see Chapters 27 and 28). Although Lawes et al. (2008) were able to demonstrate parts of the conventional corticopontocerebellar pathways using simple streamline tractography, there were as many artifactual connections that did not decussate in the pons. Strong but erroneous ipsilateral corticocerebellar “tracts” were visible in the tractography. The corticopontine tracts merged with pontocerebellar tracts running ipsilaterally instead of contralaterally. Another decussating pathway, the dentatothalamic tract through the superior cerebellar peduncle, could not be displayed by streamline tractography using a single-tensor model (Lawes et al., 2008).

Studies in Humans: Indirect Comparisons

DTI tractography has been compared to known anatomical pathways without necessarily revisiting the anatomical data. In tractography, cumulative errors due to noise cause deviations from the actual pathway. This divergence depends on the shape of the tract, step size, voxel size, method of interpolation, and degree of anisotropy (Basser and Pajevic, 2000; see also Chapter 22 by Alexander). Chen et al. (2006) developed a multipass approach to fiber tracking to minimize the cumulative effect of noise. Fiber tracking is initiated from a region of interest (ROI) and followed until a region of ambiguity is reached. Semi-random directions of tracking are generated until the fiber tract meets stopping criteria or a second designated ROI. This is done in both directions. Fiber tracts are merged, the center line is computed, and fiber tracking is then guided by this center line. Cerebellothalamic tracts were successfully reconstructed using this approach, despite the crossing of axons in the midbrain. Chen et al. warned that this method can generate false tracts, so a priori knowledge of the existence of valid tracts is required. Despite the deficiencies of the single-tensor model of streamline tractography, other methods are capable of producing results consistent with known anatomy.

Another example of comparison to known anatomical pathways without necessarily revisiting the anatomical data concerns the anterior parts of the fornix and the medial parts of the anterior commissure; the anatomy of this region can be obtained from routine dissections. This region provides a clear example of where tractography failed to display known anatomy. Postcommissural fibers between the hippocampus and the mamillary bodies are visible in Figure 26.3 from Lawes et al. (2008), but an artifactual component ran on into the amygdala. This was almost certainly a part of the anterior commissure that blended into the fornix. No precomissural fibers could be detected. More strikingly, the anterior commissure could be traced from periamygdaloid regions toward the midline but did not cross it, disappearing at precisely the location of the missing precomissural fornix. This may depend on exactly where the image plane intersects the anterior commissure, as Catani et al. (2002) obtained a more complete image. Clearly, these two tracts shared the same voxels and thereby corrupted the tractography.

Figure 26.3 Fornix and anterior commissure. i) Tract representation of the postcommissural fornix from the mean DTI. The tract passes from hippocampal (LLHP) and parahippocampal (LLPH) regions through the fornix (LLFN) to the mammillary bodies (LLMB). ii) A similar artifactual tract passing from the periamygdaloid region (LLAM) to the mammillary bodies (LLMB). iii) Representation of the anterior commissure. The track passes from the periamygdaloid regions (LLAM, RLAM) to the anterior commissure (LLAC, RLAC) but fails to cross the midline. iv) Simultaneous representation of the postcommissural fornix, the artifactual tract, and the anterior commissure. The continuation of the fornix into the periamygdaloid region (red arrow) by the artifactual tract resembles the path of the anterior commissure (blue arrow) to the same region. The deficiency in the anterior commissure (yellow arrow) is close to the expected location of the precommissural fornix. v) Intersubject tract variability maps of the fornix (two left images) and the anterior commissure (two right images) in standard space. From Lawes et al. (2008) NeuroImage 39:62–79 with permission from Elsevier.

Figure 26.3
i)ii)iii)iv)v)Lawes et al. (2008) NeuroImage 39:62–79 with permission from Elsevier.

Where a tract branches or merges with another, or where two tracts approach each other then diverge, the direction of the principle eigenvector no longer represents the direction of the tracts (see Chapters 27 and 28). If a tract shares a voxel with cerebrospinal fluid (CSF) or gray matter, the FA may drop below the threshold required to continue tracking. Areas where tracts take a highly curved course will also cause problems, particularly if an angle threshold has been artificially imposed. Intersubject agreement is often mistaken for validation. Intersubject agreement may indicate consistency in production of an artifactual tract and cannot be taken as evidence of validity of the putative tract.

Studies in Humans: Comparison with Functional Techniques

Staempfli et al. (2008) used fMRI to verify DTI-based tractography of connections between motor cortex and putamen. Subjects were asked to flex and extend fingers and toes, and elevate the corner of the mouth. An advanced fast marching algorithm was used to generate tracts and a likelihood index was derived to estimate the probability of a connection. The fMRI results depicted an anteroinferior-to-posterosuperior somatotopic organization, with the face anteroinferiorly, the foot posterosuperiorly, and the hand in between. Tractography, however, showed a different pattern: the face and hand preserved the same relative, though displaced, location as in the fMRI data, but the foot area clearly moved inferiorly and foot-related voxels were present both anteriorly and posteriorly. Only 14.3% of hand voxels and 6.3% of face voxels shared fMRI and DTI somatotopy. As the authors deliberately chose a pathway that has several other pathways passing close by, the discrepancies are not surprising. The approach is clearly promising, despite the difficulties encountered.

Kinoshita et al. (2005) compared the size of “pyramidal” tracts in two subjects with the area from which electrical excitation of tracts evoked electromyographic responses during operations to remove tumors. After removal of tumor in one subject, including areas from which motor evoked responses could be elicited by electrical stimulation but which were outside the tracts displayed by DTI, the patient's left hemiparesis was worse. Inspection of their Figures 3 and 4, however, suggests that the “pyramidal” tracts identified by their DTI procedure were unusually narrow.

There are no specific data on where they placed their ROIs, apart from the comment that they were in motor cortex. As only a small proportion of the “pyramidal” tract originates in motor cortex, it is not surprising that DTI based on an ROI in motor cortex underestimated the extent of the tract. Typical estimates of the origin of the corticospinal tract (a.k.a. the “pyramidal” tract) are 22%–31% from BA4 (motor cortex), 29% from BA6 (premotor cortex), and 40%–79% from BA3, 1, 2, 5, 7 (sensory cortex and superior parietal lobule). A significant part of the corticospinal tract originates on the medial surface of the cerebral hemispheres, from paracentral lobule, premotor cortex, and supplementary motor cortex, so it would be a mistake to restrict an ROI to the precentral gyrus. It would appear that the problem may not be so much the failure of DTI to delineate the full extent of the tract as the placement of ROIs based on a limited view of the anatomy of the corticospinal system. A more realistic representation of the extent of the corticospinal system delineated by DTI is shown in Figure 10iii in Lawes et al. (2008). A valid comment made by Kinoshita et al. is that the contribution to the “pyramidal” tract made by fibers passing through the centrum semiovale from more lateral cortical regions is usually not visible in DTI tractography. The decision to terminate tracking when the FA dropped below 0.3 may also have excluded a significant proportion of the tracts.

Duffau (2008) reviewed comparisons of intraoperative electrostimulation of subcortical language pathways to DTI tractography. Electrostimulation of the inferior occipitofrontal fasciculus delineated by Catani et al. (2002) induced errors concerning the meaning of words, or semantic paraphasias. Electrostimulation of the arcuate fasciculus, delineated in tractography by Catani et al. (2005), induced errors in the phonological form of words, or phonemic paraphasias. Combined electrostimulation of the anterior perisylvian language area and an electrocorticogram of the posterior language area confirmed that this connection is bidirectional (Matsumoto et al., 2004). The posterior limb of the indirect component of the arcuate fasciculus (Catani et al., 2005) connecting temporal cortex with the inferior parietal lobule was implicated in syllable discrimination and identification (Parker et al., 2005). Electrostimulation of the anterior component of this pathway connecting the inferior parietal lobule with Broca's area produced articulatory disorders, including complete anarthria. In short, a partitioning of the arcuate fasciculus into direct and anterior and posterior indirect components on the basis of DTI tractography corresponds closely to a functional subdivision based on intraoperative electrostimulation. This is perhaps the strongest validation currently available in human brains.

Duffau (2008) also reviewed the corticostriatal control stream of the subcallosal fasciculus. The subcallosal fasciculus was delineated with DTI tractography by Catani at el. (2002) and intraoperative electrostimulation of it induced a motor aphasia of spontaneous speech, leaving repetition intact (Duffau et al., 2002). Electrostimulation of the head of the caudate induced perseveration of the previous test item instead of the current item. Electrostimulation and DTI tractography concurred in tracing pathways from the lower motor area medially, bifurcating before a sharp turn into the “pyramidal” tract (presumably corticobulbar tract) and cerebral peduncle (Henry et al., 2004).

Studies in Other Species

A major problem with diffusion tensor imaging is the time necessary to acquire high-resolution images. With fixed brains, time is no longer a problem, so if tractography can be applied to fixed brains, many of the problems associated with large voxels can be circumvented.

D’Arceuil et al. (2007) investigated conditions affecting the use of Gd-DTPA contrast to optimize the signal-to-noise ratio of diffusion tensor imaging scans of formalin-fixed primate brains. They found that diffusion anisotropy was preserved after tissue fixation. Fractional anisotropy was slightly elevated in white matter but the apparent diffusion coefficient values were greatly decreased, an effect partly attributed to tissue death and reduced temperature. They confirmed that there was significant diffusion anisotropy in cortical gray matter (see Chapter 39 by D’Arceuil and De Crespigny for more details). Fiber reconstructions seeded from the forceps major, forceps minor, and posterior limb of the internal capsule of one hemisphere were compared with high-resolution FLASH scans and were apparently consistent with known anatomy (Fig. 26.4). One slightly puzzling comment indicated that tracts from the posterior limb of the internal capsule coursed through the thalamus and medullary pyramids. There is no known tract that does this, so perhaps they were referring to two separate tracts, one passing through the thalamus and a separate one passing through the medullary pyramids.

Figure 26.4 Orthogonal planes through high-resolution (425 µm) trace ADC maps, direction-encoded FA maps, and gradient echo FLASH (175 µm) images of a fixed macaque brain acquired at 4.7 T. Voxel color in the FA maps encodes principle eigenvector directions in the usual way (red = left–right, green = anterposterior, blue = superior–inferior). From D’Arceuil et al. (2007) NeuroImage 35:553–565 with permission from Elsevier.

Figure 26.4
D’Arceuil et al. (2007) NeuroImage 35:553–565 with permission from Elsevier.

Lin et al., (2001) used manganese-enhanced T1-weighted images as a gold standard for estimating of DTI tractography. Manganese chloride was injected into the vitreous humor of the eye and the rats were scanned 10 hours later. The deviation between principal diffusion eigenvectors and tangents to the manganese-enhanced images was calculated. Figure 26.5 shows the methods used to determine the tract orientation from the manganese-enhanced MRI. Comparison of the manganese-enhanced MRI and eigenvector map derived from DTI data is shown Figure 26.6. Except at the optic chiasma and the lateral geniculate nucleus, the directions were largely parallel. The rms deviation was 13.27°, 5.11° attributed to the manganese-enhanced T1-weighted images and 12.25° to DTI tractography. Lin et al. chose a pathway where technical considerations were optimal, in contrast to Staempfli et al. (2008), for example, who investigated the more challenging pathway between motor cortex and putamen. Thus Lin et al. were concerned with whether the principal eigenvector accurately predicted the direction of a tract or not, rather than with the validity of tractography in situations where crossing fibers, partial volumes, and kissing and other factors had a potentially negative impact. Lin et al. give several reasons why the technique would not be applicable to longer pathways requiring greater transport times, one being that the manganese enhancement fades after 48 hours.

Figure 26.5 Procedures of determining vectors tangential to manganese (Mn)-enhanced tracts and computation of deviation angles. The process started with an image of Mn-enhanced MRI with bright optic nerves as shown in the left panel. Using an appropriate magnitude threshold, the enhanced tracts were isolated (a). Sixth-order least-square polynomials were fit to the enhanced pixels (b). The tangential vector T of any point on the tract was determined by taking spatial derivatives of the polynomials (c). The deviation angle was then computed by subtracting the polar angle of the principal diffusion eigenvector θ‎d 1 from the polar angle of the tangential vector θ‎T at each corresponding position (d, e). Taken from Lin et al. (2001) NeuroImage 14:1035–1047 with permission from Elsevier.

Figure 26.5
(a)(b)(c)θd1θT(d, e)Lin et al. (2001) NeuroImage 14:1035–1047 with permission from Elsevier.

Figure 26.6 Images of Mn-enhanced optic tracts superimposed with principal eigenvector maps of the diffusion tensors. The magnified images are the zoom-in regions of interest enclosed by rectangles in the images on top. The superimposed images show that (a, b), at corresponding positions, the principal diffusion eigenvectors (indicated by yellow segments) are mostly parallel to the enhanced tracts. To distinguish the tract structures from the adjacent tissues, the length of each yellow segment was rescaled according to the fractional anisotropy of the diffusion tensor at that position. Having registered Mn-enhanced MRI with the images of the principal diffusion eigenvectors, deviation angles can be computed by direct comparison between tract orientations and the principal diffusion eigenvectors at each pixel as described in the legend of Figure 26.5. Taken from Lin et al. (2001) NeuroImage 14:1035–1047 with permission from Elsevier.

Figure 26.6
(a, b)Figure 26.5Lin et al. (2001) NeuroImage 14:1035–1047 with permission from Elsevier.

Dyrby et al. (2007) compared postmortem multifiber probabilistic tractography to in vivo injection of tracers in the gyrencephalic minipig brain. Biotinylated dextran amine and manganese chloride were injected into three pig brains. They were scanned 2 days before and after injection (see Fig. 26.7) then sacrificed 2 weeks later. After perfusion with paraformaldehyde, the brains were postfixed in paraformaldehyde then scanned again and processed histochemically. An overlap function compared termination sites obtained by tractography with those obtained with manganese chloride–enhanced MRI. Connections between somatosensory cortex and cortical, thalamic, and nigral sites produced overlap functions ranging from 0.63 to 0.93. Connections between prefrontal cortex and cortical, thalamic, and nigral sites produced overlap functions ranging from 0.58 to 0.98, although one brain lacked any thalamic connection demonstrated by tractography. Connections between motor cortex and cortical and nigral sites produced overlap functions ranging from 0.59 to 0.89, although a different brain lacked any thalamic connection demonstrated by tractography. The biotinylated dextran amine results were discrete, in line with standard hodological techniques. The tracts obtained with manganese chloride, by contrast, appeared to show evidence of considerable diffusion out of recognized axonal pathways into the surrounding extracellular fluid. It is unclear, therefore, whether the lower overlap functions represent an issue with tractography or one with manganese chloride–enhanced MRI. Nevertheless, the two techniques gave broadly comparable results.

Figure 26.7 In vivo tracing using manganese injected into the right prefrontal cortex (PFC) of brain 2 (A–G) and the left motor cortex (MC) of brain 3 (H–J). Pathways of the manganese labeling are represented by the statistical t-maps (color bar) visualized on the high-resolution in vitro MPRAGE MRI seen in coronal (A–D, H, I), sagittal (E, F, J), or horizontal section plane (G). The right side of the brain is depicted on the left side of the figures. Injections into the PFC (A, arrow) revealed a corticocortical pathway to the contralateral PFC (A, G, double arrow) crossing through the rostrum of the corpus callosum (E, double arrow). Ipsilateral labeling of manganese was observed passing through the internal capsule depositing ventrally in a high-intensity ventral area (B, F, arrow) as well as contralaterally (B, double arrow). More caudally, this region splits into at least two distinct pathways: one passes through the ventral thalamus eventually entering the mediodorsal nucleus (C, E, G, arrow), while another pathway is directed ventrally toward the substantia nigra (D, F, double arrow). Manganese injected into the MC (H, arrow) projects to the contralateral MC (H, double arrow) via the body of corpus callosum, as well as to the caudate (J, arrow) and the ventral anterior and ventral lateral nuclei of the thalamus (I, J, double arrow). Scale bar = 5 mm. Taken from Dyrby et al. (2007) NeuroImage 37:1267–1277 with permission from Elsevier.

Figure 26.7
(A–G)(H–J)Dyrby et al. (2007) NeuroImage 37:1267–1277 with permission from Elsevier.

Conclusion

In conclusion, many attempts have been made to validate diffusion tensor imaging and tractography. Comparison of tractography with conventional hodological methods in experimental animals will ultimately become the gold standard, but only for the specific tracts investigated in the particular species studied. The precise anatomical relationships between any tract and its confounding factors will differ from tract to tract and from species to species, rendering a universal gold standard an unattainable goal. Conventional hodological techniques are, of course, inapplicable to humans.

Use of functional techniques in humans offers a more realistic approach. Functional MRI has its own limitations, but offers one potential avenue. Intraoperative electrostimulation is also a powerful strategy but relies on the pathway being sufficiently eloquent to be tested in an operative setting. The study of language is particularly relevant in this context, as no other species has the same capacity.

Ultimately, higher-quality imaging data in terms of both resolution and signal-to-noise ratio together with improved descriptions of complex fiber architecture including crossing and kissing fibers will render much of the current difficulties and their prohibitively expensive solutions obsolete. Despite this requirement, significant progress has been made in the quest to validate DTI and tractography. These techniques now represent the primary modality for investigation of connectivity in the living human brain.

References

Alexander AL, Hasan KM, Lazar M, Tsuruda JS, Parker DL (2001). Analysis of partial volume effects in diffusion-tensor MRI. Magn Reson Med 45:770–780.Find this resource:

Basser PJ, Pajevic S (2000). Statistical artefacts in diffusion tensors MRI (DT-MRI) caused by background noise. Magn Reson Med 44(1):41–50.Find this resource:

Behrens TEJ, Jenkinson M, Robson MD, Smith SM, Johansen-Berg H (2006). A consistent relationship between local white matter architecture and functional specialisation in medial frontal cortex. Neuroimage 30:220–227.Find this resource:

Berman JI, Berger MS, Chung SW, Nagarajan SS, Henry RG (2007). Accuracy of diffusion tensor magnetic resonance imaging tractography assessed using intraoperative subcortical stimulation mapping and magnetic source imaging. J Neurosurg 107(3):488–494.Find this resource:

Brodal A (1981). Neurological Anatomy in Relation to Clinical Medicine. New York: Oxford University Press.Find this resource:

    Catani M, Howard RJ, Pajevic S, Jones DK (2002). Virtual in vivo interactive dissection of white matter fasciculi in the human brain. Neuroimage 17:77–94.Find this resource:

    Catani M, Jones DK, Ffytche DH (2005). Perisylvian language networks of the human brain. Ann Neurol 57:8–16.Find this resource:

    Chen P, Magnotta VA, Wu D, Nopoulos P, Moser DJ, Paulsen J, Jorge R, Andreasen NC (2006). Evaluation of the GTRACT diffusion tensor tractography algorithm: a validation and reliability study. Neuroimage 32:1075–1085.Find this resource:

    D’Arceuil HE, Westmoreland S, de Crespigny AJ (2007). An approach to high resolution diffusion tensor imaging in fixed primate brain. Neuroimage 35:553–565.Find this resource:

    Devlin JT, Sillery EL, Hall, DA, Hobden P, Behrens TEJ, Nunes RG, Clare S, Matthews PM, Moore DR, Johansen-Berg H (2006). Reliable identification of the auditory thalamus using multi-modal structural analysis. Neuroimage 30:1112–1120.Find this resource:

    Duffau H (2008). The anatomo-functional connectivity of language revisited. New insights provided by electrostimulation and tractography. Neuropsychologia 46(4):927–934.Find this resource:

    Duffau H, Capelle L, Sichez N, Denvil D, Lopes M, et al. (2002). Intraoperative mapping of the subcortical language pathways using direct stimulation. An anatomo-functional study. Brain 125:199–214.Find this resource:

    Dyrby TB, Søgaard LV, Parker GJ, Alexander DC, Lind NM, Baaré WFC, Hay-Schmidt A, Eriksen N, Pakkenberg B, Paulson OB, Jelsing J (2007). Validation of in vitro probabilistic tractography. Neuroimage 37:1267–1277.Find this resource:

    Henry RG, Berman JI, Nagarajan SS, Mukherjee P, Berger M (2004). Subcortical pathways serving cortical language sites: initial experience with diffusion tensor imaging fiber tracking combined with intraoperative language mapping. Neuroimage 21:616–622.Find this resource:

    Kinoshita M, Yamada K, Hashimoto N, Kato A, Izumoto S, Baba T, Maruno M, Nishimura T, Yoshimine T (2005). Fibre-tracking does not accurately estimate size of fibber bundle in pathological condition: initial neurosurgical experience using neuronavigation and subcortical white matter stimulation. Neuroimage 25:424–429.Find this resource:

    Lawes INC, Barrick TR, Murugam V, Spierings N, Evans DR, Marie Song M, Clark CA (2008). Atlas based segmentation of white matter tracts of the human brain using diffusion tensor tractography and comparison with classical dissection. Neuroimage 39:62–79.Find this resource:

    Lazar M, Weinstein DM, Tsuruda JS, Hasan KM, Arfanakis K, Meyerand ME, Badie B, Rowley HA, Haughton V, Field A, Alexander AL (2003). White matter tractography using diffusion tensor deflection. Hum Brain Mapp 18(4):306–321.Find this resource:

    Lin C-P, Tseng W-YI, Cheng H-C, Chen J-H (2001). Validation of diffusion tensor magnetic resonance axonal fibre imaging with registered manganese-enhanced optic tracts. Neuroimage 14:1035–1047.Find this resource:

    Matsumoto R, Nair DR, LaPresto E, Najm I, Bingaman W, Shibasaki H, et al. (2004). Functional connectivity in the human language system: a cortico-cortical evoked potential study. Brain 127:2316–2330.Find this resource:

    Mori S, van Zijl PC (2002). Fiber tracking: principles and strategies—a technical review. NMR Biomed 15(7-8):468–480.Find this resource:

    Nadeau SE, Ferguson TS, Valenstein E, Vierck CJ, Petruska JC, Streit WJ, Ritz LA (2004). Medical Neuroscience. Philadelphia: WB Saunders Elsevier.Find this resource:

      Pajevic S, Pierpaoli C (1999). Color schemes to represent the orientation of anisotropic tissues from diffusion tensor data: application to white matter fiber tract mapping in the human brain. Magn Reson Med 42:526–540.Find this resource:

      Parker GJM, Luzzi S, Alexander DC, Wheeler-Kingshott CAM, Ciccarelli O, Ralph MAL (2005). Lateralisation of ventral and dorsal auditory–language pathways in the human brain. Neuroimage 24:656–666.Find this resource:

      Parker GJM, Stephan KE, Barker GJ, Rowe JB, MacManus DG, Wheeler-Kingshott CAM, Ciccarelli O, Passingham RE, Spinks RL, Lemon RN, Turner R (2001). Initial demonstration of in vivo tracing of axonal projections in the macaque brain and comparison with the human brain using diffusion tensor imaging and fast marching tractography. Neuroimage 15:797–809.Find this resource:

      Rademacher J, Bürgel U, Geyer S, Schormann T, Schleicher A, Freund HJ, Zilles K (2001). Variability and asymmetry in the human precentral motor system. A cytoarchitectonic and myeloarchitectonic brain mapping study. Brain 124:2232–2258.Find this resource:

      Staempfli P, Reascher C, Jaeermann T, Valavanis A, Kollias S, Boesiger P (2008). Combining fMRI and DTI: a framework for exploring the limits of fMRI-guided DTI fibre tracking and for verifying DTI-based fibre tractography results. Neuroimage 39:119–126.Find this resource:

      Thiebaut de Schotten M, Urbanski M, Duffau H, Volle E, Lévy R, Dubois B, Bartolomeo P (2005). Direct evidence for a parietal-frontal pathway subserving spatial awareness in humans. Science 309:2226–2228.Find this resource:

      Wiegell MR, Larsson HB, Wedeen VJ (2000). Fiber crossing in human brain depicted with diffusion tensor MR imaging. Radiology 217:897–903.Find this resource:

      Yoshiura T, Kumazawa S, Noguchi T, Hiwatashi A, Togao O, Yamashita K, Arimura H, Higashida Y, Toyofuku F, Mihara F, Honda H (2008). MR tractography based on directional diffusion function: validation in somatotopic organization of the pyramidal tract. Acad Radiol 15:186–192.Find this resource: