Keywords
intracranial aneurysm - microsurgery - magnetic resonance imaging - software
Introduction
Giant intracranial aneurysms (GIAs) are defined as aneurysms of >25 mm in diameter
and constitute 2.5 to 5.0% of all intracranial aneurysms.[1]
[2]
[3] GIAs are more tumorlike than small aneurysms. The most common clinical presentation
of a GIA is a local mass effect resulting in cranial nerve palsy, visual field defects,
or limb weakness.[2]
[4] Endovascular treatment for GIAs is limited due to its failure to relieve such mass
effects, high cost for more coils, and high complication rate.[5] Thus, in certain cases, open surgery may be the best choice. However, GIAs have
always been one of the most difficult cerebrovascular lesions to surgically manage.
Despite the advances in microsurgical techniques, the corresponding surgical morbidity
and mortality rates have remained as high as 20%.[5]
[6] Therefore, meticulous imaging studies and individualized planning are required before
surgery.
In the clinical setting, preoperative planning mainly involves digital subtraction
angiography (DSA) and computed tomography angiography (CTA), which can always provide
precise information about the circulating parts of GIAs and their patent arteries.
However, neither examination technique can reveal the noncirculating parts of GIAs,
such as thrombi. Spontaneous thrombotic aneurysms can constitute up to 60% of all
GIAs[3]
[7]; thus, the diameters and real morphology of GIAs are often likely to be underestimated
by DSA or CTA.[8] Additionally, because they behave like tumors, GIAs can compress the surrounding
parenchyma and shift the brain structures from their normal locations. These changes
also cannot be recognized only by DSA or CTA; therefore, other techniques are required
to provide more information about GIAs for preoperative planning.
The utility of magnetic resonance imaging (MRI), a useful noninvasive examination
technique, has been increasing in the evaluation of GIAs.[8] The time-of-flight (TOF) sequence, mainly referred to as magnetic resonance angiography
(MRA), has an ability to visualize circulating parts of aneurysms similar to the abilities
of DSA and CTA.[9]
[10] Motion-sensitized driven equilibrium (MSDE) is another established sequence that
can distinguish GIA thrombi from artifacts,[11]
[12] many of which are caused by turbulent blood flow.[8]
[13] 3D Slicer (http://www.slicer.org; [Fig. 1]) is a free software for medical image visualization and analysis. This software
supports image segmentation, registration, and multimodal visualization. In the present
study, we reconstructed the circulating and noncirculating parts of GIAs with TOF
and MSDE sequences by 3D Slicer (Version 4.6.2 for Mac OS X) to reveal an integrated
presentation of GIAs and validate its usefulness for preoperative planning.
Fig. 1 The (a) time of flight (TOF) and (b) motion-sensitized driven-equilibrium (MSDE) sequence data were imported into 3D
Slicer in Digital Imaging and Communications in Medicine (DICOM) format. Three compositions
were identified after comparison at the same anatomical slides between the two sequences:
blood flow, thrombi (purple arrows), and artifacts (black arrow). (c) The giant intracranial aneurysm (GIA) was first automatically marked with the Threshold
Effect tool. (d) Next, the Paint Effect tool was used to manually prune and modify the structures.
Finally, (e) the segmentation was converted to a 3D model and (f) this was similar to digital subtraction angiography (DSA) at the same angle. Integration
of the thrombotic and circulatory parts together produced a whole model of the GIA
and allowed for extraction of its (g) diameter and (h) volume. The reconstructed GIA was integrated (i) with slide images and (j) into a head model to obtain more information. (k) The GIA model could be rotated at any angle, and two arteries (black and white arrows) could be verified in the surgical view angle. (l) The arteries (black and white arrows) could be confirmed during surgery in the same view.
Material and Methods
Patients
After being approved by the institutional ethics committee of Chinese PLA general
hospital, we retrospectively reviewed patients with GIAs who were treated surgically
in our department between July 2015 and July 2018. Only the patients with preoperative
TOF and MSDE MRI sequence and DSA imaging were included. Thrombi were also counted
as parts of GIAs when measuring diameters. The maximum diameters were determined by
both DSA and MRI. The exclusion criteria were (1) ruptured aneurysms, regardless of
whether the GIAs were responsible for the hemorrhage and (2) previous treatment of
the GIAs by microsurgery or endovascular intervention. At last, 31 consecutive patients
with 31 GIAs were included in the study.
Segmentation and Measurement
Both the TOF and MSDE sequence data for every case were loaded into 3D Slicer, which
was installed in a personal computer (MacBook Pro, 2.6 GHz Intel Core i5; Apple Inc.,
Cupertino, California, United States). GIA reconstructions were independently performed
by two neurosurgeons who received the same guided training on the reconstruction procedure
until they attained adequate proficiency. The reconstructions were mainly based on
TOF sequences. Three compositions were identified in the TOF sequences: blood flow,
thrombi, and artifacts. The MSDE sequences were used as a reference for distinguishing
thrombi from artifacts in the TOF sequences.
The Threshold Effect tool was applied to automatically mark the circulatory parts
of the GIAs and the cerebrovascular circulation ([Figs. 1] and [2]). Manual drawing was then performed with the Paint Effect tool to prune or modify
the colored structures slice by slice. All of the manual procedures covered ∼25 to
50 slices according to the range of thrombi or artifacts in the GIAs. Finally, the
resulting segmentation was converted to a 3D model and rendered with a smooth contour
by the Merge and Build tool. Information on the volume of the thrombi and aneurysms
could be instantly obtained from the model. The maximum diameters were measured by
the two points that were farthest apart from each other within the reconstructed GIAs.
Both values measured by the two independent neurosurgeons were recorded, and the average
was calculated as the final result. The time required for every reconstruction was
recorded.
Fig. 2 Magnetic resonance imaging (MRI) examination of (a) time of flight (TOF) and (b) motion-sensitized driven-equilibrium (MSDE) sequences confirmed a nonthrombotic
giant intracranial aneurysm (GIA) affected by artifacts (arrows). (c) First, the GIA was automatically marked with the Threshold Effect tool. (d) We then manually pruned and painted the whole aneurysm with Structure 17 using the
Paint Effect tool. (e) Finally, the segmentation was converted to a 3D model with a smooth contour. (f) The morphology on 2D digital subtraction angiography (DSA) also confirmed the nonthrombotic
GIA. (g) We then integrated the model into the reconstructed skull with semitransparency
and rotated it to the neurosurgeon's view. The view could be adjusted to any angle
with the computer mouse. Both the (g) reconstructed model and (h) intraoperative fluorescence contrast image confirmed a surrounding perforator artery
(arrows), proving the accuracy and usefulness of the reconstruction by 3D Slicer for preoperative
planning.
For the 3D DSA, images were reconstructed by virtual reality on an Advantage Workstation
(GE Healthcare). DSA data were stored in the workstation, and the GIA parameters were
measured by a neuroradiologist who did not participate in the parameter measurement
by 3D Slicer.
Statistical Analysis
After confirmation of the data distribution, the diameters of the GIAs were determined
(median or mean ± standard deviation), and a paired t-test was used for comparison between DSA and 3D reconstruction. To test for interobserver
agreement, the intraclass correlation coefficient (ICC) for DSA and 3D reconstruction
was calculated using a one- and two-way random-effects model, respectively, and was
used as evidence of reconstructive stability. The coincidence rate for a thrombus
verified between the preoperative and postoperative periods was expressed as a percentage.
All statistical analyses were performed with SPSS Statistics 22 (IBM Corp., Armonk,
New York, United States). A value of p < 0.05 was considered statistically significant.
Results
The patients comprised 17 women (55%) and 14 men (45%), with a mean age of 46.7 years
(range, 23–67 years). Among all GIAs, 15 (48.4%) were identified as thrombotic aneurysms.
The mean diameter of all GIAs was 28.7 ± 1.5 mm (range, 25.9–31.9 mm). Most GIAs (93.5%)
were located in the anterior circulation; only two were located in the posterior circulation.
In general, most clinical presentations of GIAs were attributed to a mass effect.
Among all patients, 12 presented with a vision defect, 8 with headache, 2 with ptosis,
2 with limb weakness, 2 with epilepsy, and 1 with hypomnesia. The remaining four patients
had no obvious symptoms, and their GIAs were incidentally found. The patient and aneurysm
characteristics are summarized in [Table 1].
Table 1
Patient and aneurysm characteristics
|
Characteristics
|
|
Number of GIAs
|
31
|
|
Patient sex, F/M
|
17/14
|
|
Patient age (y), mean (min, max)
|
46.7 (23, 67)
|
|
GIA location, number (%)
|
|
Anterior circulation
|
29 (93.5%)
|
|
Posterior circulation
|
2 (6.5%)
|
|
Aneurysmal character, number (%)
|
|
Thrombotic
|
15 (48.4%)
|
|
Nonthrombotic
|
16 (51.6%)
|
|
GIA diameter, mean ± SD (mm)
|
|
DSA
|
24.46 ± 5.25
|
|
3D Slicer reconstruction
|
28.67 ± 1.48
|
|
Volume, mean ± SD (mm3)
|
|
Thrombus
|
3,590 ± 1,003.7
|
|
GIAs
|
8292.6 ± 1,175.1
|
|
Clinical presentation, number (%)
|
|
Vision defect
|
12 (38.7%)
|
|
Headache
|
8 (25.8%)
|
|
Ptosis
|
2 (6.5%)
|
|
Limb weakness
|
2 (6.5%)
|
|
Epilepsy
|
2 (6.5%)
|
|
Hypomnesia
|
1 (3.2%)
|
|
Negative
|
4 (12.9%)
|
Abbreviations: GIA, giant intracranial aneurysm; F, female; M, male; DSA, digital
subtraction angiography.
The interobserver agreement was excellent for aneurysmal diameter and volume measurement
by DSA or 3D Slicer reconstruction. The ICC for diameter by DSA, diameter by 3D Slicer,
and volume by 3D Slicer was 0.98, 0.87, and 0.98, respectively, and all corresponding
p values were <0.001 ([Table 2]). The mean diameter for all GIAs measured by DSA and 3D Slicer was 24.46 ± 5.25
and 28.66 ± 1.48 mm, respectively (t = 4.948, p < 0.001). When only nonthrombotic GIAs were included, the mean diameter measured
by DSA and 3D Slicer was 28.69 ± 2.03 and 28.97 ± 1.79 mm, respectively (t = 1.023, p = 0.323; [Table 3]). The mean aneurysmal volume was 8,292.6 ± 1,175.1 mm3 and the mean thrombotic volume was 3,590.0 ± 1,003.7 mm3. For thrombotic GIAs, the thrombus volume constituted 46.89 ± 14.58% of the total
aneurysm volume. The average time for all reconstruction procedures was ∼30 minutes.
Table 2
The estimation of ICC between measurer
|
Type
|
ICC
|
p
|
|
Diameter
|
|
|
|
|
DSA
|
One-way random
|
0.98
|
< 0.001
|
|
3D Slicer reconstruction
|
Two-way random
|
0.87
|
< 0.001
|
|
Aneurysmal volume
|
Two-way random
|
0.98
|
< 0.001
|
Abbreviations: ICC, intraclass correlation coefficient; DSA, digital subtraction angiography
Table 3
Paired t-test between two methods for diameter measurement
|
DSA (mm)
|
3D Slicer (mm)
|
t
|
p
|
|
All GIAs
|
24.46 ± 5.25
|
28.66 ± 1.48
|
4.948
|
< 0.001
|
|
GIAs without thrombus
|
28.69 ± 2.03
|
28.97 ± 1.79
|
1.023
|
0.323
|
Abbreviations: DSA, digital subtraction angiography; GIAs, giant intracranial aneurysms.
Discussion
Because they are tumorlike, GIAs often present with paroxysmal neurologic symptoms
that are associated with a mass effect,[2]
[4]
[14] and this may limit the ability to treat GIAs with endovascular intervention. In
certain cases, open surgery may be the best choice. However, GIAs are always challenging
to manage microsurgically. Therefore, meticulous imaging studies and individualized
planning are required before surgery. Therapeutic planning for GIAs requires comprehensive
images. DSA and CTA, both standard techniques, cannot reveal a thrombus in a GIA,[8]
[13] which influences the accuracy of preoperative planning. The aneurysmal reconstruction
with TOF and MSDE sequences by 3D Slicer in our series is a supplemental technique
for surgical planning.
3D Slicer is a free open-source, extensible software application for medical image
computing and segmentation.[15] It can be used for quantitative assessment of the aneurysmal and thrombotic volume,
which may be an independent risk factor for a poor outcome.[16]
[17] The software allows us to not only measure the whole volume of the GIA but also
split the volume into thrombotic and nonthrombotic portions ([Table 1]; [Fig. 1], [2]). Because they are compressed by peripheral brain tissue, progressive GIAs do not
always present as an increase in size but instead by a change in volume.[18]
[19] In such cases, the change in aneurysmal volume over time as measured by 3D Slicer
can provide evidence for a high rupture risk, which is especially important for conservatively
managed GIAs. In our series, all volumes were measured with the TOF sequence.
The TOF MRA imaging is often affected by artifacts resulting from slow and swirling
blood flow in GIAs[8]
[13]
[20] ([Fig. 2]). MSDE was recently proposed for fast vessel wall imaging with accurate blood signal
suppression and is mostly used for vessel wall imaging at the carotid bifurcation,[12]
[21] where swirling flow is also present. Several reports have demonstrated the ability
to use MSDE techniques for imaging of vessels in the intracranial region.[22] In the present study, we could have reconstructed the thrombus with the MSDE sequence
and integrated the data together with the circulatory portion reconstruction using
the TOF sequence. However, the different parameters for the layer thickness between
the two sequences could result in an inconsistent volume measurement; therefore, we
only used the MSDE sequence as a reference for identification during the reconstructions
in this series.
As a free open-source software accessible from the internet, 3D Slicer can run on
a personal computer and does not require specialized equipment. Thus, it is easy to
access at home or in the operating room. This feature is very important for neurosurgeons.
Medical imaging, especially the MRI, is performed by radiologists. Although radiologists
can offer neurosurgeons the best images with which to diagnose GIAs, they cannot consider
the imaging findings from the neurosurgeon's intraoperative viewpoint. Additionally,
traditional 2D images may be adequate for very experienced neurosurgeons, but a less
experienced neurosurgeon may have difficulty extracting useful information. Therefore,
3D Slicer helps us to seize the initiative of processing image data to some extent.
The process of reconstruction is also a method with which to learn the anatomical
relationship between GIAs and peripheral tissue before surgery, and the reconstructed
model can be infused into a brain model and rotated at any angle ([Figs. 1], [2]). From this aspect, 3D Slicer reconstruction can serve as a teaching tool for less
experienced neurosurgeons. On the other hand, combined with the popular augmented
reality (AR) technology,[23] the reconstructed 3D images may optimize the workflow for surgical procedures by
providing essential anatomical information that is entirely integrated into the surgical
field. This is really worthy of a further exploration.
This study has some limitations. First, it may have been relatively time-consuming;
the average time required for all reconstruction procedures was ∼30 minutes. However,
this time decreased as the learning curve progressed. Moreover, the time was dependent
on the desired quality of the model. A coarser but adequately usable model for preoperative
evaluation could be reconstructed even more rapidly. Second, the 3D Slicer is not
approved by the Food and Drug Administration for certain clinical tasks and could
only be used as a reference. However, this limitation is double-sided; the packages
provide users with plenty of useful tools without any restrictions, as in commercial
software. Third, no aneurysms of <25 mm in diameter were included in our series, and
the sample size (31 GIAs) was small.
Conclusion
In conclusion, GIAs are challenging to manage microsurgically, and meticulous imaging
studies and individualized surgical planning are required before surgery. The MSDE
sequence brings diagnostic benefit as comparison to other MRI sequences. GIA reconstruction
with 3D Slicer is a low-cost, dependable, and useful supplemental technique for surgical
planning.
