MR imaging techniques that detect extremely small changes in inherently weak signal have the most to gain from high-field scanning. Diffusion tensor imaging, the measurement of water molecules' preference to move in certain directions, offers a case study.
Although DTI is possible at 1.5T, obtaining enough signal from diffusing water molecules to produce meaningful results can be difficult. Moving to 3T makes DTI much easier, according to Dr. Neerai Chepuri, a consultant neuroradiologist at Abbott Northwestern Hospital in Minneapolis.
"Manufacturers have spent a lot of time and energy tweaking radio-frequency coils or shimming magnetic fields to get relatively modest increases in signal-to-noise ratio. But when you go from 1.5T to 3T, you experience a far more dramatic increase at once," he said.
The added SNR makes a real difference to DTI because it starts from a very low signal threshold, said Dr. Gregory Sorensen, an associate professor of radiology at Harvard Medical School and codirector of the Martinos Center for Biomedical Imaging at Massachusetts General Hospital.
"Signal to noise is one of those things where a boost does make a difference in some applications, whereas in other applications it does not," he said. "DTI is signal-to-noise-hungry, so the more you can give it, the better."
Turning up the signal has its own downside, however, said Prof. Dr. Juergen Hennig, scientific director of radiology at University Hospital Freiburg in Germany. DTI is reliant on echo-planar imaging, which suffers from misregistration artifacts. Because these artifacts scale directly with signal, DTI distortion can be worse at 3T. Gradients used for diffusion encoding can introduce further distortions via eddy currents.
"Most people are currently adopting a pragmatic approach, by just accepting that images are distorted and doing some correction afterwards," Hennig said.
Hennig primarily uses 3T DTI as a research tool to investigate connectivity in the brain. Water molecules have a known preference for diffusion along neuronal fibers in white matter, so DTI can help reveal the fibers' pathways and structure. He has been scanning young children's brains, in collaboration with colleagues from the Zurich Children's Hospital, to show how fiber tracts develop and mature. A separate study, undertaken together with psychiatrists, focuses on the involvement of white matter in neurodegenerative conditions such as schizophrenia and Alzheimer's disease. DTI may show whether disease progression is associated with the thinning of fibers or chronic deterioration of fiber tracts.
"We are developing the tools necessary to merge DTI with functional MRI. I hope that next year we will have that up and running so it can be of use routinely for all kinds of scientific and clinical activities," Hennig said.
Chepuri has already put 3T DTI to work with some of his clinical patients. Because DTI is not reimbursable, he performs the technique only on an ad hoc basis. Most of his cases have required DTI to categorize brain tumors prior to neurosurgery. An aggressive cancer that is invading adjacent parenchyma may be difficult to resect, while a tumor that is only pushing healthy tissue aside should be far easier to remove.
Sorensen has been using DTI routinely in clinical practice for the past 10 years, despite the fact it is not reimbursable. One common application is to identify the location of fiber tracts in brain tumor patients before the surgeon decides where to cut, he said.
"Any time we do any diffusion imaging, even at 1.5T, we do DTI because it doesn't take any longer and we get the diffusion imaging for free," he said.
But Sorensen remains unsure of the value of DTI as a stand-alone tool, even at 3T, and doubts it will ever be reimbursable in the U.S. Yet over 100 sites worldwide have requested licenses for the DTI processing software developed by Sorensen's group at MGH, indicating widespread interest in the technique.
Perhaps signal-hungry DTI simply needs more SNR to prove its worth. Initial results from a prototype 32-channel head coil make 3T images look more like 7T images, according to Sorensen.
"It's easily a doubling of SNR going to 32 channels," he said. "Once these coils are used routinely, then we'll finally be able to see things that might help us understand whether this is going to be valuable clinically for some of the diagnostic dilemmas that we face today."