Synopsis

This lecture will cover the basics of diffusion MRI acquisitions, raise awareness of how imaging protocol parameters can affect the study outcome, provide an overview of recently developed diffusion techniques, and explain the relationship between representations, biophysical models, and imaging biomarkers.

A lecture in the sunrise sessions on ”Imaging Without Gadolinium: Diffusion MRI”

Presenter: Markus Nilsson, Associate Professor at Diagnostic Radiology, Lund University, Sweden

Target audience: Scientists planning or analyzing a study involving diffusion MRI

Learning outcomes

• Understand the basics of the diffusion MRI acquisition
• Awareness of how acquisition parameters can affect study results
• Acquaintance with diffusion MRI techniques going beyond diffusion tensor imaging
• Knowledge of the difference between “representations” and “biophysical models”

Synopsis

Diffusion MRI can be used to obtain quantitative imaging biomarkers sensitive to changes in normal or pathogenic biological processes on the microscopic level. Examples of such imaging biomarkers are the Apparent Diffusion Coefficient (ADC), which is sensitive to the cellularity of tumors (Chen et al., 2013)or the anisotropic mean kurtosis which is sensitive to the presence of elongated cell structures (Szczepankiewicz et al., 2016).

Today diffusion MRI can be performed at most clinical MRI scanners. Adjustable parameters of the imaging protocol are the strength of the diffusion encoding (the b-value) and possibly the direction of the diffusion encoding, as well as the echo and repetition times. Acquisitions performed with few directions and a single b-value can be used to map the ADC. By acquiring in multiple directions, diffusion tensor imaging (DTI) can be performed to produce maps of the fractional anisotropy (FA) which is sensitive to white matter changes (Assaf, Johansen-Berg, & Thiebaut de Schotten, 2017). By acquiring with multiple b-values it becomes possible to use diffusional kurtosis imaging (DKI) (Jensen, Helpern, Ramani, Lu, & Kaczynski, 2005), and to produce maps of the mean kurtosis (MK), which tends to be more sensitive to e.g. white matter maturation than DTI-based parameters (Grinberg et al., 2017).

The diffusion MRI techniques mentioned so far (ADC mapping, DTI, DKI) yields parameters that are merely representations of the MR signal intensity and how it change with the diffusion encoding (Novikov, Kiselev, & Jespersen, 2018). Although the parameters provided by these techniques are sensitive to changes in tissue microstructure, they do not by default reveal much about the underlying biology. The repeatability of the parameters are generally high, but there are reproducibility concerns because the parameters are sensitive to changes in the imaging protocol (e.g. b-value or echo time) (Chuhutin, Hansen, & Jespersen, 2017; Jones & Basser, 2004). This problem has been addressed by two complimentary approaches: biophysical models that connects microstructure properties of the tissue to the observable signal (Alexander, Dyrby, Nilsson, & Zhang, 2017), and multidimensional diffusion encoding strategies (Topgaard, 2017).

Biophysical models can be used to elucidate what aspects of tissue microstructure that can be accurately measured with diffusion MRI. For example, see (Veraart, Fieremans, & Novikov, 2019). They can also be used to guide the design of new multidimensional imaging protocols to be more specific to certain microstructure properties such as cell shapes, axon density, blood volume, cell membrane permeability, and more (Nilsson, Englund, Szczepankiewicz, van Westen, & Sundgren, 2018). Such imaging protocols requires control of not only the b-value, but over the full gradient waveform used for the diffusion encoding . Examples include double diffusion encoding (Lasič, Nilsson, Lätt, Ståhlberg, & Topgaard, 2011; Shemesh et al., 2016), and q-space trajectory encoding (Westin et al., 2016).

How can advanced diffusion protocols contribute to scientific discoveries of clinical practice? Regular diffusion-weighted imaging has been used clinically for decades due to its unique ability to reveal ischemic events minutes after onset (Moseley et al., 1990), and to visualize cell-dense parts of tumors. More advanced strategies for diffusion MRI have found their way into research studies (Assaf et al., 2017; Lebel, Treit, & Beaulieu, 2017), but clinical adaption is lacking potentially due to the lack of “killer applications”. New methods for diffusion encoding may change that, and we review promising examples.

Acknowledgements

References