When doctors try to understand how body fat affects health, they need to know exactly where that fat is located. This means distinguishing between visceral fat, which sits deep inside the belly, and subcutaneous fat, which sits just under the skin. Doing this manually is slow and difficult, but a new look at data from over 12,000 women shows that AI can do the heavy lifting.
Researchers found that an AI system called 3D-Unet was incredibly accurate at identifying these different fat types. In fact, the AI was over 99% faster than manual methods. While a human might take about 46 minutes to process one patient's scan, the AI finished the task in about 3 minutes. This speed allows doctors to get results much faster without sacrificing precision.
However, there are some hurdles to clear before this becomes a standard in every clinic. Because different studies used different tools and methods, the results aren't perfectly consistent across the board. While the AI is highly accurate on an individual level, the lack of a single standard rule for how to use these tools means it may not work the same way in every hospital just yet.
