Estimation of Error in Volume Measurement for 100 Intracranial Aneurysms Imaged Serially with CE-MRA at 1.5T and 3T
Farshid Faraji1, Alastair Martin1, and David Saloner1

1Department of Radiology, UCSF, San Francisco, CA, United States

Synopsis

Intracranial aneurysms are localized dilations in blood vessels occurring in 1-6% of the population which can have devastating consequences in the event of rupture. Many aneurysms are asymptomatic, found incidentally, and fall below the surgical size threshold. For this reason, physicians choose to follow aneurysms with imaging rather than opting for surgical or endo-vascular intervention. Here we further our analysis of a previously presented image processing technique using contrast-enhanced MRA to follow intracranial aneurysms longitudinally. We investigate the measurement error of this post-processing technique in 100 intracranial aneurysms, and evaluate the differential effects of imaging at 1.5T and 3T.

Introduction

We report on an ongoing longitudinal study of unruptured intracranial aneurysms. Intracranial aneurysms pose a considerable risk for subsequent neurological events, such as aneurysmal rupture resulting in hemorrhagic stroke or mass effect from the growing aneurysm pressing on critical brain structures. This study reports on measurement error of 3D MRI/MRA methods used to evaluate the progression of disease in patients with intracranial aneurysms where no intervention was planned, either because of size, unfavorable treatment options, or patient choice. This analysis extends our previously reported findings by evaluating changes in measurement error when increasing field strength from 1.5T to 3T.

Methods

87 patients with 100 aneurysms of the intracranial circulation were recruited for serial imaging using an IRB-approved protocol. Patients were imaged at baseline and followed in intervals ranging between 6 months and 1 year. All patients were imaged on a Philips 1.5T Achieva (n=85 patients with 365 studies), or Siemens 3T Skyra (n=60 patients with 175 studies). 50 patients were imaged on both MRI scanners. Patients were categorized by territory, with 62 aneurysms of the Internal Carotid Artery, 12 of the Middle Cerebral Artery, 8 of the Anterior Cerebral Artery, and 18 found to be aneurysms of the posterior circulation. Of 100 aneurysms, 2 had 12 follow-up studies, 2 had 11 follow-up studies, 2 had 10 follow-up studies, 3 had 9 follow-up studies, 8 had 8 follow-up studies, 7 had 7 follow-up studies, 9 had 6 follow-up studies, 10 had 5 follow-up studies, 11 had 4 follow-up studies, 9 had 3 follow-up studies, 17 had 2 follow-up studies, and the remaining 18 had one follow-up study. This results in a total of 540 interval measurements. At each imaging session MRI and MRA were conducted to assess lumenal volume and presence of thrombus1,2. A 3D balanced steady-state free precession pulse sequence was run (bFFE on Philips and TrueFISP on Siemens) with orientation and resolution matching the CE-MRA study to check for the presence of thrombus. If thrombus was present, the patient was excluded from this analysis, as longitudinal changes could be due to lumen reduction/expansion from thrombus. The MRA study used was a contrast-enhanced 3D acquisition with a parallel acceleration factor of 2 resulting in high-resolution CE-MRA images of the intracranial vessels (0.6 x 0.63 x 1.2mm at 1.5T and 0.7mm isotropic resolution at 3T). Serial MR studies were co-registered using anatomical features. Consistent intensity-based threshholding was imposed, requiring that a reference segment of undiseased vessel maintain the same lumenal volume over time. The lumenal volume of the aneurysmal segment was subsequently assessed on the follow-up CE-MRA studies for regional and global changes. Changes in volume of the aneurysmal segment were calculated as a percentage of the baseline volume and were normalized on an annualized basis. Measurement error was calculated as the residual standard deviation after performing a mixed-effects REML regression.

Results

Measurement error was calculated as 4.6% for studies performed at 1.5T. Measurement error was calculated as 3.5% for studies performed at 3T. Of the aneurysms that were followed, 11 of 100 showed growth that was beyond measurement error. By territory, MCA aneurysms showed the greatest measurement error, with 15.7% error. This was likely due to the small volume of aneurysms in this territory, as well as the difficulty in performing consistent segmentations at the trifurcation of the middle cerebral artery where these aneurysms are primarily located. Additionally, we note that small aneurysms demonstrate an appreciable size increase upon first imaging at 3T, most likely due to a reduction in partial volume effect resulting from the increase in spatial resolution.

Discussion

MRI provides a minimally invasive method to monitor intracranial aneurysms, allowing for the investigation of their natural history in ways that have not been possible before. In particular, 3D analyses remove the limitations of traditional methods utilizing linear measurements. Here we further analyze methodology that allows for longitudinal assessment of growth using contrast-enhanced MRA. Our results demonstrate less than 5% measurement error at 1.5T. Additionally, we found that imaging at 3T allows for more than 1% reduction in measurement error, which gives greater sensitivity and specificity to detect aneurysm growth. This is important considering that growing aneurysms are at greater risk for rupture, and will better inform physicians as to which patients should go on to have an intervention due to re-stratification of risk. Measurement error reduction also allows for significant sample size reduction to detect statistical significance in longitudinal studies.

Acknowledgements

Special thanks to Dr. Charles McCulloch for formulation, discussion, and interpretation of statistical analyses.

References

[1] Dispensa, BP et al JMRI 2007; 26:177–183

[2] Boussel et al. JVIR 2011; 22(7):1007-11

Figures

Iso-surface rendering of an ICA aneurysm (red) and an un-diseased reference vessel segment (blue)

Saccular ACA aneurysm showing rapid growth from baseline (blue) to 24 months (red)



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
2678