Hi all,
Why do we acquire different volume samples of a brain in diffusion MRI and what is the sequence of direction?
A DWI sequence acquires many volumes, each with a specific b-vector. The direction of the b-vector selects which fibers will generate signal. A single b-vector will only detect fibers of specific orientations, and will not detect others. Specifically, you can think of each b-vector as being sensitive to fibers pointing orthogonal to the direction of the vector. This means that the gradient [1, 0, 0] will be sensitive to a fiber pointing [0, 1, 0] or [0, 0, 1], or any other direction along the orthogonal disc. To detect all the fibers, and to discriminate fibers we need to acquire volumes with different b-vectors. To correct for eddy currents, it is useful to sample a whole sphere. General advice is here. Some DWI scans are multi-shell, acquiring volumes where the b-value varies. Higher b-values are more selective for fiber directions, but have lower signal to noise. Varying b-values can help estimate scalar properties like kurtosis and also help develop a more accurate orientation distribution function, which can help distinguish regions where fibers are crossing.
Hi @Chris_Rorden thanks for the response. Are you aware of any instance where the identification of fibers have been used in a clinical setting
Hi @Chris_Rorden in above you have mentioned about fODF(fiber orientation distribution function). In the papers I have read that to estimate fODF they have used DTI, q ball and so on. But in DTI there is the problem of crossing fibers. In one paper I read that using deep learning an RNN based model they have been able to generate a fODF. What are you opinions on using deep learning techniques to estimate fODF and what will be the advantage of using this method compared to DTI, q ball?
I would look at published articles like this one rather than one user’s anecdotal comments to make decisions. When I developed diffusion sequences for my center, I described the trade-offs I considered, described an objective metric, and acquired sample data. My own work is with chronic stroke patients who often experience kyphosis and where obesity is a risk factor. Therefore, we used a larger head coil with fewer channels (16 for the head) than others. Higher spatial resolution provides better ability to detect fiber bundles with reduced partial volume effects, but have inherently lower SNR that impacts the ODF calculation. Likewise, higher b-values are more sensitive to fiber direction, but have lower SNR. If one is only interested in tractography, you will choose different parameters than if you are interested in scalar measures (MD, MK, etc).
When setting up my sequences, I asked for advice from MRI physicists from 3 different centers. I would urge others to do the same. A large amount of resources are devoted to acquiring data, so consulting with experts at the planning stage is worthwhile. If your instrument is a Siemens, GE or Philips instrument with a research agreement, the vendor should have a research collaboration manager who is tasked with providing you advice. These individuals are a great resource. They know a lot of details about your equipment that is not publicly disclosed. They can help you tune your sequence.
1 Like