A few months ago I was approached with a wonderful idea for MR Imaging from a different group on campus. The basic premise is that there are two types of fat: brown and white. Traditionally we think of fat as a bad thing (white) but brown adipose tissue can actually be a good thing. Rich in mitochondria it can help keep the body warm and actually aids in metabolism ( fat burning fat!). It turns out babys have a lot of it, and lose most of it as they age. No one is entirely sure of the dynamics of it, but that’s what we intend to find out using a cohort of babies.

Some preliminary animal models have been done at USC on a GE scanner, our task at hand uses a different sequence for quantification, the Dixon method. The Dixon method takes advantage of the differing resonant frequencies in adipocytes and water molecules. Two images are acquired, one where the two are fully out of phase (water – fat) and one where the two are fully in phase (water + fat). Simple addition/subtraction yields either a water image, or a fat image. Things are more complicated than this, but the gist is there. Since brown fat is a little “leakier”, meaning it’s filled with blood/watery tissue it can be segmented apart from white fat by it’s fat fraction (fat/(fat+water)). In theory. To test the imaging we constructed some phantoms on the fly by emulsifying different concentrations of oil/water with detergent (breaking down lipids). The concentrations from top-right, clockwise: Sterile water with a syringe of pure oil as a control, 22% oil (this mixture would not emulsify properly, I suspect it’s closer to 1-2% oil in these slices), ~70% oil (this emulsified with a lot of bubbles, meaning a lot of water was extracted, I’d estimate the oil concentration to be a bit higher in reality) and ~86% oil. Keep in mind these were done on the fly so they are measured a few % off at best.

The opp/in phase images:

Fat/Water images:

Fat/(Fat+Water) image (scale is cold=0, hot=1):

The trends are clearly there. Hot tube of oil in the top right, two bottles of mostly water on the right, two bottles of mostly oil on the left. You can also see the intensity of the top left bottle being greater than the bottom left bottle. It is a little less of a contrast than expected (measured 80% and 74% concentrations respectively). The pure oil only measured ~85% as well, likely due to the susceptibility differences in the oil/plastic/water interface. You can see a lot of this in the phase and fat/water images (it roughly cancels in the ratios), apparent is the higher fat content in the oil. It’s tough to comment on the bottom right bottle due to a poor emulsion, but it did have a significantly higher concentration than pure water (7% vs. 2%). This was a reservedly successful measurement, the next step is to mimic concentrations closer to that expected in white and brown adipose tissue to see if we can get the separation necessary for segmentation.