166
K. Miller et al.
7.1.1 Background
Epilepsy is a chronic neurological disorder that affects over 70 million people, costs
over $50 billion annually and first arises during childhood for 75% of patients [18].
Surgical intervention can be curative but is rarely used because of difficulties with
localisation of epileptogenic regions of the brain and subsequent surgical planning.
When MRI-visible lesions are not present, surgical planning must rely on functional
localisation. Intracranial EEG (iEEG) is the most effective method for functional
localisation of the SOZ; invasive electrodes are implanted and monitored for several
days and then removed during a second surgery when the resection is performed
[4]. Collecting data from directly on or within the cortex significantly increases the
spatial data fidelity when compared to scalp EEG measurements and allows more
accurate identification and location of the SOZ.
Despite being the most accurate available technology for monitoring seizure
activity, iEEG does not guarantee surgical success. Unsuccessful surgeries are
performed, in part, due to difficulties in the clinical interpretation of the measured
invasive data. In current practice, electrodes are aligned to the cortex using a
combination of intra-operative photographs and diagrams to estimate the placements
of electrode grids and depth electrodes. While this allows generalised alignment,
it is not quantitative and does not provide the degree of accuracy necessary for
emerging precision surgical techniques such as radio frequency (RF) and laser
ablation [6, 24] and focused ultrasound [17]. To fully utilise the precision of these
emerging techniques, accurate alignment algorithms are needed to enable precise
identification of the SOZ with respect to both pre-operative and intra-operative
imaging.
A significant factor contributing to electrode alignment error is the physical
shifting of brain tissue during invasive measurement (see Fig. 7.1 next page).
Electrode grids in the intracranial space, and the body’s inflammatory response
to the craniotomy, displace and deform the brain from the configuration observed
in pre-surgical MR imaging [23, 27]. To ensure accurate alignment of electrode
placements and correct clinical evaluation of invasive data, brain shift must be
accurately modelled and accounted for.
In Chap. 6 we described in detail effective approaches for computing brain
deformations during neurosurgery. In this chapter we focus on applying these
modelling methods to the particularly challenging and clinically relevant problem
of the registration of pre-operative MRI to intra-operative CT (with electrodes
implanted) images.
Accurate registration is achieved using reliable computation of the deformations
within the brain due to invasive electrode placement. An accurate model of brain
shift enables us to accurately map presurgical imaging onto the deformed intraoperative space, accurately aligning these scans with X-ray CT (Fig. 7.1), thereby
providing the location of electrodes as well as the SOZ identified by them relative to
the anatomical structures of the brain as seen on pre-operative MRI ‘warped’ onto
intra-operative CT images.
K. Miller et al.
7.1.1 Background
Epilepsy is a chronic neurological disorder that affects over 70 million people, costs
over $50 billion annually and first arises during childhood for 75% of patients [18].
Surgical intervention can be curative but is rarely used because of difficulties with
localisation of epileptogenic regions of the brain and subsequent surgical planning.
When MRI-visible lesions are not present, surgical planning must rely on functional
localisation. Intracranial EEG (iEEG) is the most effective method for functional
localisation of the SOZ; invasive electrodes are implanted and monitored for several
days and then removed during a second surgery when the resection is performed
[4]. Collecting data from directly on or within the cortex significantly increases the
spatial data fidelity when compared to scalp EEG measurements and allows more
accurate identification and location of the SOZ.
Despite being the most accurate available technology for monitoring seizure
activity, iEEG does not guarantee surgical success. Unsuccessful surgeries are
performed, in part, due to difficulties in the clinical interpretation of the measured
invasive data. In current practice, electrodes are aligned to the cortex using a
combination of intra-operative photographs and diagrams to estimate the placements
of electrode grids and depth electrodes. While this allows generalised alignment,
it is not quantitative and does not provide the degree of accuracy necessary for
emerging precision surgical techniques such as radio frequency (RF) and laser
ablation [6, 24] and focused ultrasound [17]. To fully utilise the precision of these
emerging techniques, accurate alignment algorithms are needed to enable precise
identification of the SOZ with respect to both pre-operative and intra-operative
imaging.
A significant factor contributing to electrode alignment error is the physical
shifting of brain tissue during invasive measurement (see Fig. 7.1 next page).
Electrode grids in the intracranial space, and the body’s inflammatory response
to the craniotomy, displace and deform the brain from the configuration observed
in pre-surgical MR imaging [23, 27]. To ensure accurate alignment of electrode
placements and correct clinical evaluation of invasive data, brain shift must be
accurately modelled and accounted for.
In Chap. 6 we described in detail effective approaches for computing brain
deformations during neurosurgery. In this chapter we focus on applying these
modelling methods to the particularly challenging and clinically relevant problem
of the registration of pre-operative MRI to intra-operative CT (with electrodes
implanted) images.
Accurate registration is achieved using reliable computation of the deformations
within the brain due to invasive electrode placement. An accurate model of brain
shift enables us to accurately map presurgical imaging onto the deformed intraoperative space, accurately aligning these scans with X-ray CT (Fig. 7.1), thereby
providing the location of electrodes as well as the SOZ identified by them relative to
the anatomical structures of the brain as seen on pre-operative MRI ‘warped’ onto
intra-operative CT images.