Institute of Neuro Innovation

View Original

An Introduction to Diffusion Tensor Imaging

By Amir Vala Tavakoli

Movement and Sensation by Greg Dunn and Brian Edwards

When we look up to see the sky crossed by a white trace, we can follow the white vine of water-vapor to its seeding source, the burning engine of a jet. What if we could similarly trace the thread-like connections of the brain? Diffusion tensor imaging (DTI) is a technique for imaging the brain that allows us to describe the structures of the brain’s connections by contouring the diffusion, or directional movement, of water. While DTI builds upon standard magnetic resonance imaging (MRI) techniques, it is specifically designed to uncover the brain’s connective structure. DTI enables researchers and clinicians to image the brain with regard to the tubular paths recruited by neural communication during cognitive and behavioral functions (1). This introductory essay aims to briefly introduce some concepts that are fundamental to DTI.

We regard structural connectivity as an important neuroanatomical feature because our percepts, behaviors, and thoughts do not emerge from independently functioning neurons. Rather, even basic neurological functions such as movement (5) and vision (6) are vitally determined by dynamic processes occurring between structurally connected networks of interacting populations of neurons. 

Revealing connective structure is possible because axons, the extended and tubular appendages of neurons, make up the connective structure of the brain known as white matter.  Each axon is structurally continuous with the body of a neuron and helps pass that neuron’s information to other neurons, like an arm extending a note. White matter tracts are densely packed bundles of these connective axons. These tracts structure the brain’s connectivity by carrying dynamically oscillating chemical gradients that facilitate the propagation of neural information and communication. Neuroimaging with DTI uses these physical constraints and obstacles to the otherwise random movement of water to image the brain’s connective structure (3). 

DTI is sensitive to the preferred direction of water’s movement and the strength of this preference. Imagine that we have placed a bowl of warm noodle soup in a scanner that is sensitive to the directional movement of the salty broth. We would see that much of the broth is not moving randomly, but rather moving directionally, along paths of least resistance, within and between the long structure of noodles. More precisely, in an otherwise empty space, physically unconstrained water molecules will move in random directions, i.e., they have no directional preference. Yet, if placed in a physically constrained space that has a coherently or reliably longitudinal structure, this will cause the water molecules to diffuse in a directional manner. By contrast, if we were to image a regular kitchen sponge, soaked and suspended in a pool of water, the sponge would reveal very little directional movement of water due to the sponge’s irregular and predominantly random, porous structure. The water-logged tunnels that perforate the sponge have no coherent directional structure. Incidentally, with its irregular or non-directional constraint of water’s movement, the sponge resembles the grey matter of the brain. This is why DTI is primarily used to produce images of white matter tracts, or tractography, that connect cortical and subcortical regions. These diffusion-based measures of the brain’s anatomic and neurophysiological connectivity enable clinicians and investigators to make informed and qualified predictions about the anatomically-specific neurological functions of a given patient or population.

Using DTI to map a patient’s web of connective tracts is like a clever visitor to a new city trying to locate train stations. She visually scans a transit map and follows the convergence of train tracks to locate stations and transit hubs. Still, transit lines lack the three-dimensional complexity occupied by white matter tracts in the brain, and so, let’s use a different metaphor.  Diffusion-based tractography is like an earth-penetrating GPS map of the major roads and tunnels passing between, over, and through a mountain range. This special type of GPS won’t show you how many people travel the roads on average, but it will reveal the roads’ most significant features. These features include their breadth, length, orientation in three-dimensional space, as well as where roads appear clustered together at a common point. DTI allows us to develop powerful inferences about these features within the brain (3). Furthermore, connectivity may be analyzed for the whole brain or between specifically selected regions of interest. 

The full process of producing tractography tends to be computationally intensive, requiring multiple steps and significant computation time to produce modeled tracts from the output of the MRI. To represent the brain’s structure in three dimensions, DTI models the brain’s data in a three-dimensional grid of fixed voxels. Voxels are like the common pixels that make up a standard digital image, except that voxels represent the imaged brain in three dimensions, not just two. DTI combines the MRI-derived values of these voxels with mathematical tools to measure the apparent, directional quality of the diffusion of water, producing an estimate of diffusion known as a tensor. When water’s diffusion is coherently directional, the tensor is often shaped like an elongated grain of rice or cigar and this helps produce the most important measure of diffusion known as anisotropy. If at a single voxel the orientations of the contained anatomic tracts appear coherent, as if their anatomic orientations are nearly parallel, the cigar-shaped tensor modeled within this voxel is seen as having anisotropy. If, however, tract orientations within a voxel are not coherent, the average diffusion estimated would not reveal a coherent direction. The technique, deterministic tractography, takes a sample of the diffusion-data at each voxel to estimate tract structure from the direction of water’s diffusion. On the other hand, even within directionally coherent tracts, water’s directional diffusion can vary, making a single sample of diffusion unreliable (3). For example, when dropping two separate beads of ink into two identical bowls of motionless water, the expanding contours of swirling ink will still differ. This variability in the movement of water molecules will similarly vary the direction in which water diffuses at a given location within the brain over time. To address this concern, a computational method called probabilistic tractography is applied. Probabilistic tractography takes many samples of diffusion-orientation at each voxel. Still, even this more computationally intensive and time-consuming method of tractography is limited in its ability to model more complex white-matter structure within a voxel. Geometric precision about the direction of tract structure is known as angular resolution. If an imaged voxel within the brain contains tracts that run in dissimilar directions, both deterministic and probabilistic tractography may lack sufficiently high angular resolution. To address this issue, more sophisticated methods of diffusion-weighted imaging can more precisely model each voxel’s complex tract structure by acquiring more data, which requires longer scan times. Once high angular resolution diffusion images are acquired, certain computational advances in modeling, such as the q-ball method, increase the precision of tractography, allowing us to estimate the presence and orientations of crossing fibers within a voxel (7). 

DTI uses a specific combination of statistical tools that model, project, or draw streamlines that curve through the brain’s voxels, building an informed estimate of tract structure. Investigators use DTI to trace connections between different parts of the brain by selecting a source, known as a seed, from which these projections extend. Like stepping stones sequenced to form a garden path, a streamline is an individual estimate of sequential voxels that originate at a selected seed. Importantly, the reliability of streamlines is constrained by certain factors. Major anatomic features are one such example because the brain contains several sizeable pockets where water’s movement is random, where cerebrospinal fluid pools in ventricles. Also, the reliability of a streamline as an inference of connectivity between voxels tends to decrease with increasing distance from the seed, much as the water-vapor lagging behind the skyward jet expands with distance. The tracts closest to the seed may be the strongest and the most reliable connections. Eventually, after the brain’s many distributed tracts have been estimated, informative tracts must be parsed from uninformative tracts (3,8). After tractography has been modeled, a specifically identified region of interest can be targeted either for research, surgical planning, or both. 

There is a wide range of clinical applications for the spatially precise targeting of neuroanatomy. Surgically, tractography has been used to help target subcortical structures for deep brain stimulation (DBS), a therapy that electrically stimulates neural circuits associated with a range of neurological functions such as the production of smooth movement (9). Like an electrician who knows to avoid actively functioning power lines, tractography helps neurosurgeons localize, avoid, and preserve the white matter tracts that underlie diverse functions such as vision (10), language (11), and movement (12). That is, the clinical significance of preserving deep brain structures reflects their role in the integration of distributed neural information. The deep brain structures of the basal ganglia, for example, function as a clustered hub for neural communication, underlying both basic and higher-order neural processes (13,14). To alleviate the motor symptoms associated with Parkinson’s disease and essential tremor, tractography has been used to segment and target the basal ganglia for DBS (9). Tractography-guided DBS has also been used to stimulate other regions of the brain involved in pain management (15), as well as mood-related areas of the brain in patients with medication-resistant depression (16).  

Continuing advances in neurosurgical therapies for movement disorders leverage tractography as a useful means of uncovering neuroanatomy. While patients are often painlessly awake during neurosurgical procedures, recent growth of “asleep” DBS, during which patients are fully induced under anesthesia, has increased reliance on imaging-guided targeting, including DTI. Beyond invasive intracranial stimulation, non-invasive interventions, such as focused ultrasound, also benefit from tractography. In essential tremor, tractography was used to target and lesion the thalamus with focused ultrasound by first mapping thalamic regions with the highest connectivity to the cortical regions associated with movement (5). In addition to functional lesioning of dysfunctional deep brain structures, current developments in focused ultrasound methods yield steady demand for the identification of deep brain tracts. From non-destructive neuromodulation to the targeted delivery of drugs across the blood-brain barrier, long an intractable problem, DTI provides focused ultrasound with important spatial information about the location of targets. 

The fundamental purpose of tractography with DTI is to allow reliable approximation of the structural connectivity that is essential to neural communication, cognitive function, and behavior. Understanding that our behaviors and neural functions emerge from networks of connections demands tools to reveal the underlying structural connectivity of these networks. Thus tractography continues to show its utility across a range of investigations in both the clinic and basic research. Given the breadth and depth of this field relative to this brief introduction, more detailed introductions to DTI research are available (2–4).

 

References

  1. Skudlarski, P. et al. Measuring Brain Connectivity: Diffusion Tensor Imaging Validates Resting State Temporal Correlations. NeuroImage 43, 554–561 (2008).

  2. Introduction to diffusion tensor imaging and higher order models. (Elsevier/Academic Press, 2014).

  3. Diffusion MRI: theory, methods, and application. (Oxford University Press, 2010).

  4. Soares, J. M., Marques, P., Alves, V. & Sousa, N. A hitchhiker’s guide to diffusion tensor imaging. Front. Neurosci. 7, (2013).

  5. Tsolaki, E., Downes, A., Speier, W., Elias, W. J. & Pouratian, N. The potential value of probabilistic tractography-based for MR-guided focused ultrasound thalamotomy for essential tremor. NeuroImage Clin. 17, 1019–1027 (2018).

  6. Saalmann, Y. B., Pinsk, M. A., Wang, L., Li, X. & Kastner, S. The Pulvinar Regulates Information Transmission Between Cortical Areas Based on Attention Demands. Science 337, 753–756 (2012).

  7. Caiazzo, G., Trojsi, F., Cirillo, M., Tedeschi, G. & Esposito, F. Q-ball imaging models: comparison between high and low angular resolution diffusion-weighted MRI protocols for investigation of brain white matter integrity. Neuroradiology 58, 209–215 (2016).

  8. Tian, Q. et al. Diffusion MRI tractography for improved transcranial MRI-guided focused ultrasound thalamotomy targeting for essential tremor. NeuroImage Clin. 19, 572–580 (2018).

  9. Pouratian, N. et al. Multi-institutional evaluation of deep brain stimulation targeting using probabilistic connectivity-based thalamic segmentation. J. Neurosurg. 115, 995–1004 (2011).

  10. Kamada, K. et al. Functional Monitoring for Visual Pathway Using Real-time Visual Evoked Potentials and Optic-radiation Tractography: Oper. Neurosurg. 57, 121–127 (2005).

  11. Kamada, K. et al. Visualization of the frontotemporal language fibers by tractography combined with functional magnetic resonance imaging and magnetoencephalography. J. Neurosurg. 106, 90–98 (2007).

  12. Martin, A. R. et al. Translating state-of-the-art spinal cord MRI techniques to clinical use: A systematic review of clinical studies utilizing DTI, MT, MWF, MRS, and fMRI. NeuroImage Clin. 10, 192–238 (2016).

  13. Sherman, S. M. Thalamus plays a central role in ongoing cortical functioning. Nat. Neurosci. 19, 533–541 (2016).

  14. Sherman, S. M. Functioning of Circuits Connecting Thalamus and Cortex. in Comprehensive Physiology (ed. Terjung, R.) 713–739 (John Wiley & Sons, Inc., 2017). doi:10.1002/cphy.c160032

  15. Kim, W., Chivukula, S., Hauptman, J. & Pouratian, N. Diffusion Tensor Imaging-Based Thalamic Segmentation in Deep Brain Stimulation for Chronic Pain Conditions. Stereotact. Funct. Neurosurg. 94, 225–234 (2016).

  16. Tsolaki, E., Espinoza, R. & Pouratian, N. Using probabilistic tractography to target the subcallosal cingulate cortex in patients with treatment resistant depression. Psychiatry Res. 261, 72–74 (2017).