338
H. Courtecuisse et al.
13.3.5.1 Quantitative and Qualitative Results After Dural Opening
Our method has been evaluated and compared to the modified ICP (providing a rigid
registration) of [31, 38, 40], available in the open-source framework for imageguided therapy CustusX [41]. To obtain quantitative results, the first difficulty was
to define reliable landmarks visible in both image modalities. For each case, 5–9
landmarks were first identified in blood vessel bifurcations, by two operators, on
both the pre-operative MRA and intra-operative power Doppler US images. These
landmarks were used to define a target registration error (except for one patient
for which the quality of US was not enough to define landmarks). In addition,
anatomical structures such as sulcus were delineated by a clinician in the MRI and
B-mode US acquisitions, allowing a measurement that is completely independent
from the data used by the method. Note that the measured errors, with vessel
landmarks and sulci, are subsurface and located in the region of interest, i.e. the
tumour.
Figure 13.6a shows an example of simulation and Fig. 13.6b, c the corresponding
errors for four patients. On average 72% of the deformation occurring is corrected
with our method, with a mean error under 1.5 mm for each patient [8]. In addition,
less than 2 min are required to segment the intra-operative US images, run the
simulation, and update the MR images, which is compatible with a clinical use.
After registration, MRI is warped using the displacement field of the FE model
(see Fig. 13.7, right). Qualitative observations, especially deep around the tumour,
Fig. 13.6 Brain shift compensation using constraint-based simulation. Pre- and intra-operative
blood vessels, respectively, in blue and orange, are registered. Other constraints maintain the brain
in sliding contact with the dura mater and under the US probe. The Hausdorff distance between
landmarks and delineated anatomical structures manually segmented is shown in graphs (b, c). (a)
Registration. (b) Landmarks. (c) Structures
H. Courtecuisse et al.
13.3.5.1 Quantitative and Qualitative Results After Dural Opening
Our method has been evaluated and compared to the modified ICP (providing a rigid
registration) of [31, 38, 40], available in the open-source framework for imageguided therapy CustusX [41]. To obtain quantitative results, the first difficulty was
to define reliable landmarks visible in both image modalities. For each case, 5–9
landmarks were first identified in blood vessel bifurcations, by two operators, on
both the pre-operative MRA and intra-operative power Doppler US images. These
landmarks were used to define a target registration error (except for one patient
for which the quality of US was not enough to define landmarks). In addition,
anatomical structures such as sulcus were delineated by a clinician in the MRI and
B-mode US acquisitions, allowing a measurement that is completely independent
from the data used by the method. Note that the measured errors, with vessel
landmarks and sulci, are subsurface and located in the region of interest, i.e. the
tumour.
Figure 13.6a shows an example of simulation and Fig. 13.6b, c the corresponding
errors for four patients. On average 72% of the deformation occurring is corrected
with our method, with a mean error under 1.5 mm for each patient [8]. In addition,
less than 2 min are required to segment the intra-operative US images, run the
simulation, and update the MR images, which is compatible with a clinical use.
After registration, MRI is warped using the displacement field of the FE model
(see Fig. 13.7, right). Qualitative observations, especially deep around the tumour,
Fig. 13.6 Brain shift compensation using constraint-based simulation. Pre- and intra-operative
blood vessels, respectively, in blue and orange, are registered. Other constraints maintain the brain
in sliding contact with the dura mater and under the US probe. The Hausdorff distance between
landmarks and delineated anatomical structures manually segmented is shown in graphs (b, c). (a)
Registration. (b) Landmarks. (c) Structures