Protocol optimization for quantitative MRI studies in radiation oncology: II. diffusion MRI.
Authors
Affiliations (7)
Affiliations (7)
- Cancer Therapy Centre, SWSLHD, 1 Elizabeth Street, Liverpool, 2170, Australia.
- Ingham Institute for Applied Medical Research, 1 Campbell street, Liverpool, 2170, Australia.
- South Western Sydney Clinical School, University of New South Wales, 1 Elizabeth Street, Sydney, New South Wales, 2170, Australia.
- Cancer Therapy Centre, SWSLHD, 1 Elizabeth Street, Liverpool, New South Wales, 2170, Australia.
- Cancer Therapy Centre, Liverpool Cancer Therapy Centre, 1 Elizabeth Street, Liverpool, 2170, Australia.
- Siemens Healthcare GmbH, ., Erlangen, ., Germany.
- School of Science, Western Sydney University, Locked bag 1797, Penrith, New South Wales, 2751, Australia.
Abstract
One of the most used bio-functional MRI techniques deployed in radiation therapy is diffusion-weighted MRI (DWI) for both the delineation of the gross target volume (GTV) and for treatment response assessment. While the diffusion mechanism is in principle inherently independent of the magnetic field strength and scanner type, there exist significant discrepancies in quantitative ADC maps between published studies, which can be largely attributed to the use of sub-optimal scanning and oversimplified diffusion models, including inline vendor-implemented ADC map reconstructions. Robust quality assurance (QA) and the use of adequate fitting models are essential to guarantee the accuracy of quantitative DWI. Different QA methods are proposed and common pitfalls in clinical DWI studies are discussed. While standard commercial QA phantoms provide a first and necessary step in QA of clinical trials, it is shown that some methodological errors may remain undetected. Significant deviations from the mono-exponential diffusion model, a dependence on the diffusion time and directional dependence are illustrated in different human tissues. Different diffusion models are discussed. It is also illustrated how organ motion can severely compromise the accuracy of quantitative parametric diffusion maps and is not always recognized as obvious imaging artifacts.