13 Computational Biomechanics of the Brain in the Operating Theatre
327
13.2.2.6 Rendering the Information
Once the deformed configuration of the brain has been computed, this information
must be rendered to the surgeon in a clear and informative way. The most straightforward solution is to use the deformation field derived from the brain shift modelling
to map the pre-operative MRI image to the patient’s brain during surgery. To go
further, several groups have proposed augmented reality visualisation, especially
in the surgical microscope, to help fusing, understanding, and visualising complex
medical imaging data and anatomical structures [21].
13.2.3 Validation and Clinical Studies
Validating the fidelity of a computational model of the brain in itself, from a
mechanical point of view, is a difficult problem. In a surgical context, a more
relevant objective may be to evaluate the overall procedure, i.e. the outcomes of the
simulation with respect to actual data about the deformed tissues. Ideally, validation
images to measure a target registration error should be acquired after resection and
should be different than the ones used to express the loads and boundary conditions.
To our knowledge, a comprehensive study including independent input data,
acquired in real surgical conditions, and an independent post-resection iMRI,
acquired for validation only, is not yet available. However, several groups have
started acquiring such data, and results are expected soon. While the iMRI setup
will influence the measured deformation, due to its length, removal of tools, tissue
debulking, and dural closing, these images are expected to provide a much better
ground truth than any other modalities.
13.2.3.1 Practicability and Integration in the Surgical Workflow
Few studies clearly evaluate the practicability of their method. Several factors
should be considered such as integration in the surgical workflow, especially to
acquire intra-operative images, manual interactions, overall computation times, and
information rendering. In [8], our method required less than 2 min to process
the iUS images, simulate the brain deformations with a constraint-based iterative
method, and then update the pre-operative MRI (only the iUS acquisition time
is not included, usually 2–3 additional minutes). This method was evaluated in a
retrospective study only, but we showed that both the hardware and the software
could be compatible with the constraints of the operating room.
To our knowledge, almost all existing studies, including our own, are retrospective only. A notable exception is [22], in which simulations were carried out in
the operating room, using the method proposed in [19], before updated MRI was
displayed in the neuronavigation system. This study could then be considered as one
of the first computational models of the brain actually used in the operating theatre.
327
13.2.2.6 Rendering the Information
Once the deformed configuration of the brain has been computed, this information
must be rendered to the surgeon in a clear and informative way. The most straightforward solution is to use the deformation field derived from the brain shift modelling
to map the pre-operative MRI image to the patient’s brain during surgery. To go
further, several groups have proposed augmented reality visualisation, especially
in the surgical microscope, to help fusing, understanding, and visualising complex
medical imaging data and anatomical structures [21].
13.2.3 Validation and Clinical Studies
Validating the fidelity of a computational model of the brain in itself, from a
mechanical point of view, is a difficult problem. In a surgical context, a more
relevant objective may be to evaluate the overall procedure, i.e. the outcomes of the
simulation with respect to actual data about the deformed tissues. Ideally, validation
images to measure a target registration error should be acquired after resection and
should be different than the ones used to express the loads and boundary conditions.
To our knowledge, a comprehensive study including independent input data,
acquired in real surgical conditions, and an independent post-resection iMRI,
acquired for validation only, is not yet available. However, several groups have
started acquiring such data, and results are expected soon. While the iMRI setup
will influence the measured deformation, due to its length, removal of tools, tissue
debulking, and dural closing, these images are expected to provide a much better
ground truth than any other modalities.
13.2.3.1 Practicability and Integration in the Surgical Workflow
Few studies clearly evaluate the practicability of their method. Several factors
should be considered such as integration in the surgical workflow, especially to
acquire intra-operative images, manual interactions, overall computation times, and
information rendering. In [8], our method required less than 2 min to process
the iUS images, simulate the brain deformations with a constraint-based iterative
method, and then update the pre-operative MRI (only the iUS acquisition time
is not included, usually 2–3 additional minutes). This method was evaluated in a
retrospective study only, but we showed that both the hardware and the software
could be compatible with the constraints of the operating room.
To our knowledge, almost all existing studies, including our own, are retrospective only. A notable exception is [22], in which simulations were carried out in
the operating room, using the method proposed in [19], before updated MRI was
displayed in the neuronavigation system. This study could then be considered as one
of the first computational models of the brain actually used in the operating theatre.