322
H. Courtecuisse et al.
of the craniectomy, the opening of the dura, and the actions of the surgeon such
as retraction or tissue resection [1]. Several studies [2–5] have reported cortical
surface and subcortical displacements larger than 20 and 7 mm, respectively, that
must be taken into account given the fact that the desired accuracy of tumour
resection is generally below 2 mm.
While neuronavigation has become standard in neurosurgery, intra-operative
imaging has proven to be useful in many situations [6] and should be of
growing importance in the future years. Among intra-operative imaging modalities,
ultrasound (US) is common because it can easily be integrated in the clinical
workflow and is widely available in medical centres. However, US images are of
limited quality, entailing significant difficulties to identify deep internal structures.
Contrary to ultrasound, intra-operative magnetic resonance imaging (iMRI) is a
means of obtaining detailed images of patient’s brain during the surgery. However,
such installation is expensive, and the acquisition of MRI images during surgery is
a complex process that significantly increases the duration of the procedure.
To fill the gap between comprehensive pre-operative data and sparse and
incomplete or low-quality intra-operative images, the general trend is to develop
nonrigid registration method. This enables to visualise inner structures, surgical
targets, or planning at key moments of the surgery where assistance is needed. Many
image-to-image registration algorithms were proposed in the literature [1, 7]. In this
paper, we will only focus on biomechanical model-based solutions. One of their
main advantages is the ease to define the mechanical actions at the origin of tissue
deformation as well as the boundary conditions and contacts between structures. In
addition, it has proven to be consistent and robust when dealing with (very) sparse
intra-operative data.
This article is organised as follows. In Sect. 13.2 we will introduce the main steps
and criteria of model-based registration approaches. Section 13.3 is dedicated to the
presentation of our method [8], which relies on finite element modelling and vesselbased constraints from intra-operative US. We then provide detailed discussion and
comparison regarding the criteria introduced previously as well as the constraints in
the operating theatre.
13.2 Key Steps for Computational Modelling
in the Operating Theatre
This section summarises key steps and modelling choices that could or should be
taken into account before and during surgery. In the following section, we will
position our retrospective study [8] with respect to these criteria and highlight
the necessary steps that are still missing for intra-operative usage of the method
(Table 13.1).
H. Courtecuisse et al.
of the craniectomy, the opening of the dura, and the actions of the surgeon such
as retraction or tissue resection [1]. Several studies [2–5] have reported cortical
surface and subcortical displacements larger than 20 and 7 mm, respectively, that
must be taken into account given the fact that the desired accuracy of tumour
resection is generally below 2 mm.
While neuronavigation has become standard in neurosurgery, intra-operative
imaging has proven to be useful in many situations [6] and should be of
growing importance in the future years. Among intra-operative imaging modalities,
ultrasound (US) is common because it can easily be integrated in the clinical
workflow and is widely available in medical centres. However, US images are of
limited quality, entailing significant difficulties to identify deep internal structures.
Contrary to ultrasound, intra-operative magnetic resonance imaging (iMRI) is a
means of obtaining detailed images of patient’s brain during the surgery. However,
such installation is expensive, and the acquisition of MRI images during surgery is
a complex process that significantly increases the duration of the procedure.
To fill the gap between comprehensive pre-operative data and sparse and
incomplete or low-quality intra-operative images, the general trend is to develop
nonrigid registration method. This enables to visualise inner structures, surgical
targets, or planning at key moments of the surgery where assistance is needed. Many
image-to-image registration algorithms were proposed in the literature [1, 7]. In this
paper, we will only focus on biomechanical model-based solutions. One of their
main advantages is the ease to define the mechanical actions at the origin of tissue
deformation as well as the boundary conditions and contacts between structures. In
addition, it has proven to be consistent and robust when dealing with (very) sparse
intra-operative data.
This article is organised as follows. In Sect. 13.2 we will introduce the main steps
and criteria of model-based registration approaches. Section 13.3 is dedicated to the
presentation of our method [8], which relies on finite element modelling and vesselbased constraints from intra-operative US. We then provide detailed discussion and
comparison regarding the criteria introduced previously as well as the constraints in
the operating theatre.
13.2 Key Steps for Computational Modelling
in the Operating Theatre
This section summarises key steps and modelling choices that could or should be
taken into account before and during surgery. In the following section, we will
position our retrospective study [8] with respect to these criteria and highlight
the necessary steps that are still missing for intra-operative usage of the method
(Table 13.1).