Biomedical Engineering Reference
In-Depth Information
The aim of the present study was to construct a finite element model that
represented the normal LA muscle morphology in nulliparous females. Statistical
shape analysis techniques were used to extract the LA shape variation across the
sample population. A range of PF models constructed to span this shape variation
were subsequently used in childbirth simulations to investigate the effects of PF
shape on the mechanics of the second stage of labour.
2 Methods
Statistical shape analysis was performed on 3D finite element models constructed
from MRI data of the PF region collected from a previous study [
17
]. To ensure
consistence in image quality and minimise variations in subject imaging conditions,
all 12 datasets were acquired with a consistent resolution of 1 mm
1 mm,
using a 3D turbo spin echo Siemens SPACE sequence at the same acquisition angle.
In the T2-weighted MR images, the muscles were normally represented by hypo-
intense pixels, in contrast to the surrounding connective and fatty tissues with
hyper-intense signals. The shape of the PF muscles could generally be outlined
without ambiguity.
Each MRI dataset contained 120 angled axial slices, covering the pelvic region
up to the abdominal cavity. The pelvic bony structures were completely included in
the dataset for identification of repeatable bony landmarks. The segmentation was
performed using web-based visualisation software “Zinc”, developed by the
Auckland Bioengineering Institute. The bony pelvis, LA muscles and external
sphincter (ES) muscles were traced manually to yield data clouds that outline the
structures on each axial image, where inner surfaces of the muscles were digitised
separately from the outer surfaces. However, some muscle fibres of the LA
(puborectalis) blended into that of the ES, which made it difficult to identify
some of the borders between them on MR images. As a result, the boundaries
between the LA and the ES muscles (collectively named the PF muscles) were
approximated during the segmentations. Readily identifiable bony landmarks and
muscle attachment points were also digitised to align the pelvis and describe the
anatomical features.
1mm
2.1 Pelvic Floor Muscle Model Construction
Twelve individualised finite element models of the LA and ES were generated from
the digitisation surface data of the muscles using a non-linear least squares
optimisation algorithm implemented in the CMISS software [
18
]. The two muscle
surfaces were interpolated by a bicubic Hermite scheme where continuity of model
derivatives was enforced, in addition to that of the nodal positions. A linear
Lagrange scheme was adopted in the model across the thickness of the muscles
