RATIONALE AND OBJECTIVES. Automated liver surface determination in abd
ominal computed tomography scans, currently difficult to achieve, is o
f interest to determine liver location and size for various medical ap
plications, including radiation therapy treatment planning, surgical p
lanning, and oncologic monitoring. The authors propose to facilitate a
utomation by the addition of a priori shape information in the form of
a liver model. METHODS. The normalized geometric liver model is gener
ated by averaging outlines from a set of normal liver studies previous
ly registered using thin-plate spline warping. The model consists of a
n averaged liver surface, a set of anatomic landmarks, and a deformati
on function. RESULTS. A liver model is presented and its ability to re
present normal liver shapes is demonstrated. CONCLUSIONS. Liver surfac
e warping provides a means of data normalization for model constructio
n and a means of model deformation for representation of liver organ s
hapes.