Hello! I’ve been using total segmentator for body composition segmentation, from my understanding, total segmentator doesn’t use HU to separate different body tissues and please correct me if I’m wrong.
What I’ve noticed is that when I’m manual painting the paraspinal muscles using the paint function with a HU range from -29 to 150 (this is to pick up paraspinal muscle that is missed by total segmentator), I can see that a lot of the intermuscular adipose tissue segmented by total segmentator (which should have a HU range from -150 to -30) will get overwritten (which means they are actually skeletal muscle tissue instead of intermuscular adipose tissue if it gets picked up). I foresee this is going to cause a great amount of issue given that I’m comparing body composition changes with time… The only solution I can think of to minimize error in my case is to either abandon the semi-automated approach to accept the imperfect skeletal muscle segmentation/to not touch whatever segmentation outcome that’s generated by total segmentator, or to pain over every bits of the intermuscular adipose tissue to ensure skeletal muscle is not falsely being segmented as intermuscular adipose tissue.
Does this make total segmentator not a reliable tool in body composition segmentation if structures are mistaken, has someone encountered this, any suggestions and ideas on how to approach this issue would be very much appreciated