P. Wust et al., EVALUATION OF SEGMENTATION ALGORITHMS FOR GENERATION OF PATIENT MODELS IN RADIOFREQUENCY HYPERTHERMIA, Physics in medicine and biology (Print), 43(11), 1998, pp. 3295-3307
Time-efficient and easy-to-use segmentation algorithms (contour genera
tion) are a precondition for various applications in radiation oncolog
y, especially for planning purposes in hyperthermia. We have developed
the three following algorithms for contour generation and implemented
them in an editor of the HyperPlan hyperthermia planning system. Firs
tly, a manual contour input with numerous correction and editing optio
ns. Secondly, a volume growing algorithm with adjustable threshold ran
ge and minimal region size. Thirdly, a watershed transformation in two
and three dimensions. In addition, the region input function of the H
elax(TM) commercial radiation therapy planning system was available fo
r comparison. All four approaches were applied under routine condition
s to two-dimensional computed tomographic slices of the superior thora
cic aperture, mid-chest, upper abdomen, mid-abdomen, pelvis and thigh;
they were also applied to a 3D CT sequence of 72 slices using the thr
ee-dimensional extension of the algorithms. Time to generate the conto
urs and their quality with respect to a reference model were determine
d. Manual input for a complete patient model required approximately 5
to 6 h for 72 CT slices (4.5 min/slice). If slight irregularities at o
bject boundaries are accepted, this time can be reduced to 3.5 min/sli
ce using the volume growing algorithm. However, generating a tetrahedr
on mesh from such a contour sequence for hyperthermia planning (the ba
sis for finite-element algorithms) requires a significant amount of po
stediting. With the watershed algorithm extended to three dimensions,
processing time can be further reduced to 3 min/slice while achieving
satisfactory contour quality. Therefore, this method is currently rega
rded as offering some potential for efficient automated model generati
on in hyperthermia. In summary, the 3D volume growing algorithm and wa
tershed transformation are both suitable for segmentation of even low-
contrast objects. However, they are not always superior to user-friend
ly manual programs for contour generation. When the volume growing alg
orithm is used, the contours have to be postprocessed with suitable fi
lters. The watershed transformation has a large potential if appropria
tely developed to 3D sequences and 3D interaction features. After all,
the practicality and feasibility of every segmentation method critica
lly depend on various details of the user software as pointed out in t
his article.