High-Resolution Time-of-Flight Magnetic Resonance Angiography with a Volume Rendering Algorithm for Detection of Unruptured Aneurysms: Correlations with Digital Subtraction Angiography

Aleum Lee

doi:10.15746/sms.24.004

Soonchunhyang Med Sci > Volume 30(1); 2024 > Article

Lee: High-Resolution Time-of-Flight Magnetic Resonance Angiography with a Volume Rendering Algorithm for Detection of Unruptured Aneurysms: Correlations with Digital Subtraction Angiography

Original Article

Soonchunhyang Medical Science (SMS) 2024; 30(1): 23-31.

Published online: June 30, 2024

DOI: https://doi.org/10.15746/sms.24.004

High-Resolution Time-of-Flight Magnetic Resonance Angiography with a Volume Rendering Algorithm for Detection of Unruptured Aneurysms: Correlations with Digital Subtraction Angiography

Aleum Lee

Department of Radiology, Soonchunhyang University Bucheon Hospital, Bucheon, Korea

Correspondence to: Aleum Lee, Department of Radiology, Soonchunhyang University Bucheon Hospital, 170 Jomaru-ro, Wonmi-gu, Bucheon 14584, Korea, Tel: +82-32-621-5851, Fax: +82-32-621-5874, E-mail: aleerad@schmc.ac.kr; aleerad@gmail.com

Received May 12, 2024 Accepted June 24, 2024

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (https://creativecommons.org/licenses/by-nc/4.0/).

ABSTRACT

Objective

We investigated the diagnostic accuracy of high-resolution magnetic resonance angiography (HR-MRA) with volume rendering (VR) post-processing techniques for the detection of cerebral aneurysms compared with conventional MRA (C-MRA) using digital subtraction angiography (DSA) as the gold standard.

Methods

HR-MRA was performed for 51 possible aneurysms of 38 patients on C-MRA. For each possible aneurysm, readers recorded their level of confidence on a 5-point scale. All patients underwent DSA, which was used as the standard of reference. Receiver-operating characteristic (ROC) analysis and the area under the ROC curves (AUC) were used to determine the effectiveness of C-MRA and HR-MRA in detecting cerebral aneurysm with and without VR algorithm. The increased discriminative value was examined by calculation of Net Reclassification Index (NRI) and Integrated Discrimination Improvement (IDI) index.

Results

DSA revealed 37 aneurysms in 26 patients. In aneurysm-based analysis, HR-MRA showed higher diagnostic accuracy than C-MRA, and increased diagnostic accuracy on adding VR (C-MRA vs. HR-MRA, P<0.01; C-MRA with VR vs. HR-MRA with VR, P<0.01, respectively). The differences in AUC, IDI (19.38%, Z=3.18, P<0.01) and NRI (46.3%, Z=6.32, P<0.01), on adding VR to the HR-MRA, were also statistically significant. The application of HR-MRA with VR increased the detection rate of aneurysms less than 3 mm.

Conclusion

The application of HR-MRA with VR algorithm improves diagnostic performance for the detection of intracranial aneurysms, especially when the aneurysm is less than 3 mm. Clinically, HR-MRA with VR algorithm could facilitate the accurate differentiation of aneurysms from infundibula and prevent unnecessary, invasive procedure before aneurysmal rupture.

Key words: Angiography; Aneurysm; Magnetic resonance

INTRODUCTION

A cerebral aneurysm is an abnormal focal dilatation of a cerebral artery with attenuation of the vessel wall [1]. A subarachnoid hemorrhage caused by a ruptured cerebral aneurysm is frequently associated with a poor outcome [2,3].

Digital subtraction angiography (DSA) is considered the gold standard for the diagnosis of an intracranial aneurysm. However, the procedure is invasive, requires radiation exposure, and carries uncommon risks of associated neurologic complications. Consequently, noninvasive brain imaging techniques such as magnetic resonance angiography (MRA) have been pursued as potential methods for the screening and follow-up of intracranial aneurysms. Time-of-flight (TOF) MRA is a widely available noninvasive technique that requires no radiation exposure and no administration of contrast medium for the evaluation and characterization of an intracranial aneurysm [4–6]. Previous studies have suggested that TOF-MRA could identify aneurysms measuring >3 mm in size with a high sensitivity of 74% to 98% [3,7,8]. However, many lesions, including infundibula, can mimic a small aneurysm and reduce diagnostic performance for the detection of aneurysms using MRA. An infundibulum is a funnel-shaped symmetrical vascular enlargement with a vessel arising from the apex that has a maximum diameter of <3 mm at the base, and lacks an aneurysmal neck [9]. They often occur at branching sites of the posterior communicating and anterior choroidal arteries from the internal carotid artery. The incidence of infundibulum detected by angiography or at autopsy ranges from 7% to 25% [10–12].

Although infundibula are frequently considered normal anatomic variants of no pathogenic significance, some authors consider an infundibulum to be a “pre-aneurysmal” lesion because of the occasional progression of an infundibulum to aneurysm formation and rupture, the risk of which increases with age [13–16].

High-resolution TOF-MRA (HR-MRA) is a noninvasive imaging technique that provides delineation of the vascular structure in detail. This technique enables visualization of the vascular structure with sub-millimeter resolution, it has been used extensively for carotid atherosclerosis, and more recently, for carotid dissection [17,18]. Therefore, we assumed that the application of HR-MRA could improve diagnostic performance for the detection of an aneurysm. Especially for a small aneurysm less than 3 mm, which is difficult to differentiate from an infundibulum, HR-MRA would be more helpful. However, few studies have investigated aneurysm detection using HR-MRA with full correlation of DSA [3,19].

Cerebral MRA provides a three-dimensional (3D) representation of the intracranial vasculature, which is routinely captured in a two-dimensional perspective by a maximum intensity projection (MIP) algorithm or multiplanar volume reformations that might obscure an abnormality because of vascular overlay or lead to misinterpretation of tortuous vascular structures such as an aneurysm. The implementation of volume rendering (VR) algorithms to evaluate MRA data sets have been recently described by several investigators, but evidence of successful differentiation of an aneurysm from an infundibulum is lacking [15,16,20–23].

The purpose of our study was to evaluate the diagnostic accuracy of HR-MRA with VR for the detection of cerebral aneurysms compared with conventional TOF-MRA (C-MRA), using DSA as the gold standard.

MATERIALS AND METHODS

1. Patients

This retrospective study was approved by the Institutional Review Board of Soonchunhyang University Bucheon Hospital, and the requirement for informed consent was waived.

We evaluated patients with suspected or known intracranial aneurysms initially detected by C-MRA, and who underwent both DSA and HR-MRA examinations to confirm the diagnosis of an intracranial aneurysm. From September 2014 to May 2017, 38 consecutive patients (10 men and 28 women; median age, 58 years; range, 31–74 years) with 51 possible aneurysmal lesions on C-MRA, were enrolled in the present study. No patient showed a subarachnoid hemorrhage or neurologic deficit.

2. Imaging protocol

All patients underwent C-MRA, HR-MRA with MIP and VR post-processing methods, and DSA. C-MRA, which is a combined investigation of both the neck and the brain, was performed on a 1.5 T or 3 T scanner (1.5 T in 31 patients and 3 T in 20 patients; 1.5 T SIGNA EXCITE [GE Healthcare, Milwaukee, WI, USA]; 1.5 T Sonata [Siemens Medical Systems, Erlangen, Germany]; and 3 T Trio Tim and Verio [Siemens Medical System]). For 3D C-MRA, the following parameters were used: 3D fast imaging with steady precession; repetition time, 23–37 ms; echo time, 2.4–9.6 ms; flip angle, 20°–25°; effective section thickness, 0.7–1.2 mm; field of view, 180–240×215–240 mm; and a 256–640×160–336 matrix covering an area from the clivus to the genu of the corpus callosum. Rotational MIP images were available for all cases. Average of resolution in C-MRA is 0.34 (0.71, 0.21, 0.22, and 0.22, respectively).

All HR-MRA examinations were performed on a 3 T system (SIGNA EXCITE; GE Healthcare) with an 8-channel head coil. At first, we obtained 3D TOF-MRA with the following parameters: a spoiled gradient-echo sequence with repetition time 29 ms, echo time 4.7 ms, flip angle 20°, field of view 150×150 mm, matrix 512×512, and resolution 0.29×0.29×0.5 mm. The HR-MRA acquisition time was 6 minutes 31 seconds. The MIP reconstructions were performed at the time of imaging. The data were reconstructed around both the head-to-foot axis and the right-to-left axis, and target MIP reconstructions of vessels of interest were made. Furthermore, 3D reconstructed MRA images, for which a VR technique was used, were produced on a workstation (Advantage Windows ver. 2.0; GE Medical Systems, Chicago, IL, USA) by using the source MRA images.

Conventional cerebral angiography was performed using a biplane system (Integris Allura; Philips Medical Systems, Best, the Netherlands) via transfemoral catheterization and the selective injection of contrast media (Pamiray 300; Dongkook Pharmaceutical, Seoul, Korea) at a rate of 3 mL/sec into the carotid and vertebral arteries. Imaging was performed in standard and adequately oblique projections, as required. 3D image reconstruction following rotational angiography was performed. Image data were transferred to a workstation (Octane; Silicon Graphics, Mountain Hill, CA, USA) to reconstruct the images into a 3D volume (Integris 3D-RA Software release 3.2 [Octane, Silicon Graphics]).

3. Assessment of aneurysm probability

Two experienced neuroradiologists (T.J.Y. and A.L.L., with 10 and 6 years of neuroradiology experience, respectively) reviewed the MRA images; disagreements were resolved by consensus, and both clinicians were unaware of the information concerning the clinical history, the findings of other modalities, and final diagnoses. Sets of each modality including C-MRA and HR-MRA, with/or without application of VR algorithm, were evaluated separately with regular intervals. For the C-MRA and HR-MRA, assessment was performed based on MIP and source image and, for the C-MRA and HR-MRA in which VR algorithm had been applied, images derived from the VR algorithm were additionally included for the evaluation. Diagnostic confidence in the presence of an aneurysm on MRA images was assessed using a 5-point scale: no lesion, definite infundibulum, suspicious infundibulum, suspicious aneurysm, and definite aneurysm. The lesions identified as suspicious or definite aneurysm were considered as presence of aneurysm; all others were considered as absence of aneurysm. The size, number, and location of aneurysmal lesions were recorded.

We divided the data into two groups according to the size of aneurysm (≥3 mm or less) and evaluated the sensitivity for detection of aneurysm on C-MRA and HR-MRA with or without VR algorithm.

4. Statistical analysis

Receiver-operating characteristic (ROC) analysis was conducted to determine the effectiveness of C-MRA and HR-MRA in detecting cerebral aneurysm with and without VR. The sensitivity, specificity, and positive and negative predictive values for each outcome category (presence versus absence of an aneurysm) were calculated. An aneurysm detected on MRA and not detected on DSA was considered a false-positive case. An aneurysm not detected on MRA and detected on DSA was considered a false-negative case. Accuracy, sensitivity, specificity, positive predictive value, and negative predictive value (NPV) were evaluated on aneurysm-based and levels. The area under the ROC curves (AUC) and the 95% confidence interval of the area were computed to evaluate the detective ability. The AUC could distinguish between less predictive (0.5–0.7), moderately predictive (0.7–0.9), and highly predictive (0.9–1). The AUCs of the three modalities were compared by a non-parametric method by Delong. The increased discriminative value of the biomarkers was further examined by calculation of the Net Reclassification Index (NRI) and Integrated Discrimination Improvement (IDI) indices. NRI indicates the net increase versus decrease in risk categories among aneurysms minus the decrease among non-aneurysms. IDI indicates the difference in Yates slopes between models, in which the Yates slope is the mean difference in predicted probabilities between aneurysms and non-aneurysms. The differences in the sensitivity between the two groups (aneurysms sized ≥3 mm and <3 mm) were evaluated on C-MRA and HR-MRA with and without a VR algorithm.

A two-sided P-value of less than 0.05 was considered to indicate statistical significance. All the statistical analyses were performed using SPSS ver. 14.0 Korean version for Windows (SPSS Inc., Chicago, IL, USA) and R ver. 3.1.3 freely available on the web (http://cran.r-project.org/). Statistical analyses were performed with the MedCalc software package ver. 11.2.1.0 (MedCalc Software Ltd., Ostend, Belgium).

RESULTS

1. Clinical results

A total of 38 patients (10 men and 28 women; mean age, 58 years; range, 31–74 years) were included in our study. DSA revealed 37 aneurysms in 28 patients and infundibular dilatations or no lesion for 14 lesions in 10 patients. The location of the suspicious aneurysmal lesion was the internal carotid artery in 39.2% (n=20), middle cerebral artery in 25.5% (n=13), anterior cerebral artery (including the anterior communicating artery) in 23.5% (n=12), and the vertebrobasilar artery in 11.8% (n=6). DSA confirmed multiple aneurysms in nine patients with 20 lesions. Three aneurysms were detected in two patients, two aneurysms in seven patients, and one in 17 patients. Twenty-two (59.5%) of the 37 aneurysms were ≥3 mm and 15 aneurysms (40.5%) were <3 mm (range, 1–11 mm; mean, 4 mm).

2. Diagnostic performance of 3D TOF-MRA

The results concerning diagnostic confidence of the presence of an aneurysm on MRA and DSA images are summarized in Table 1. The application of HR-MRA with and without a VR algorithm improved sensitivity and specificity.

The performance was shown using the sensitivity, specificity, positive and negative predictive value; the area under the ROC curve in Table 2 and Fig. 1, and the IDI and NRI in Table 3 also demonstrate the diagnostic performance. When using categorized MRA findings, C-MRA showed a worse performance than that shown by HR-MRA regardless of VR use. The differences of AUC were also statistically significant (C-MRA vs. HR-MRA, P<0.01; C-MRA with VR vs. HR-MRA with VR, P<0.01). Although the addition of VR to the HR-MRA did not improve the AUC (0.8031 vs. 0.8658, P=0.16), statistical significance was observed for the IDI (19.38%, Z=3.18, P<0.01) and NRI (46.3%, Z=6.32, P<0.01). When using the grade of MRA finding, C-MRA showed better performance than that with categorized MRA finding, which was still worse than the performance of HR-MRA regardless of VR use. However, the differences of AUC were not statistically significant (C-MRA vs. HR-MRA, P=0.27; C-MRA with VR vs. HR-MRA with VR, P=0.24). The addition of VR to the HR-MRA did not improve the AUC (0.9228 vs. 0.9188, P=0.14), but the IDI (29.5%, Z=4.3, P<0.01) and NRI (73%, Z=15.39, P<0.01) were statistically significant.

When the size of the aneurysm was ≥3 mm, detection rates were equal for the four different MRA techniques. However, for small aneurysms (<3 mm), when HR-MRA with and without VR were applied, the detection rate increased (Table 4).

Figs. 2 –5 show representative images of patients with an infundibulum, no lesion, and aneurysm on C-MRA and HR-MRA.

DISCUSSION

In the present study, we demonstrated that the application of HR-MRA improves diagnostic performance for the detection of unruptured aneurysms. The degrees of improvement in terms of diagnostic accuracy for HR-MRA were 13.8% and 15.7% when the VR algorithm was or was not applied, respectively. The high degree of accuracy obtained in the present study seems to be the result of several factors; a large acquisition matrix of 512×512 pixels in the frequency encoding direction was used to achieve higher spatial resolution.

Interestingly, there was a notable increase in the specificity and NPV, which means that HR-MRA can also exclude an infundibulum or no lesion. These results are clinically relevant because unnecessary and invasive DSA procedures, performed to exclude an infundibulum or no lesion, may be avoided by using HR-MRA.

From the sub-analysis grouped by aneurysm size, our results show that when HR-MRA was performed the sensitivity for an unruptured aneurysm increased for small aneurysms of less than 3 mm. As expected, small aneurysms of less than 3 mm would generate difficulties when differentiating them from an infundibulum. Therefore, for these small aneurysms, HR-MRA seems to be more useful in terms of an accurate diagnosis.

The usefulness of higher field strength HR-MRA has not been established. Some authors compared the delineation ability of 3 T MRA with 1.5 T MRA in patients known to have an intracranial aneurysm and reported that 3 T MRA had a greater ability to delineate an intracranial aneurysm than that of a 1.5 T magnetic resonance scanner [24]. In contrast, Mönninghoff et al. [25] reported that 7 T magnetic resonance imaging (MRI) was superior in spatial resolution, but that all aneurysms were detected with 1.5 T MRI; however, the study was limited to only 10 patients with a known intracranial aneurysm.

Regarding the morphology of the aneurysm, the introduction of a VR algorithm contributed by providing excellent identification of the shape of the aneurysm [26]. The VR algorithm is based on the percentage classification technique, which provides an accurate determination of the amounts of materials when the voxel consists of two or more materials being volume-averaged. The volume-averaged voxels are included in the final image because the VR algorithm calculates a weighted sum from all voxels along a ray projected through the data set. Therefore, VR improves the delineation of small-caliber vessels (parent and branch vessels), the perception of a lobulated aneurysm surface, and contributes to precision in assessing aneurysm size [27]. However, the usefulness of the VR algorithm in a clinical setting has not been established. In this study, our results show that the application of a VR algorithm can improve diagnostic accuracy for differentiating an aneurysm from an infundibulum or no lesion. Virtual views of the vasculature and intracranial vessels, appearing more conspicuously after applying the VR algorithm, seem to have contributed to the higher diagnostic accuracy when the VR algorithm was applied. In addition, easier recognition of anatomic variants and anomalies might result [28]. The results of C-MRA with a VR algorithm and HR-MRA with a VR algorithm suggested that a lesion less than 3 mm in size could be more accurately detected with the implementation of a VR algorithm. This technique does not require any additional image acquisition and can be utilized with previously acquired images. Therefore, the technique is feasible in a clinical setting to increase diagnostic accuracy.

We acknowledge the following limitations of our study. Considering the clinical flow of an aneurysm, we performed the analysis for lesions that were suspected on C-MRA. Therefore, the application of these results in the general population would be limited. In addition, our data for the initial acquisition of C-MRA were acquired using machines with different magnetic fields (1.5 T or 3 T). In summary, we demonstrated that HR-MRA and a VR algorithm improved diagnostic performance for the detection of intracranial aneurysms, especially when the aneurysm was less than 3 mm. In a clinical setting, the application of HR-MRA and a VR algorithm would facilitate the accurate differentiation of aneurysms from infundibula or no lesion and avoid unnecessary invasive procedures.

CONFLICT OF INTEREST

No potential conflict of interest relevant to this article was reported.

Fig. 1

(A, B) Diagnostic Performance between conventional magnetic resonance angiography (C-MRA) and high-resolution magnetic resonance angiography (HR-MRA). The performance was shown using the sensitivity, specificity; the area under the receiver operating characteristic curve (AUC) between C-MRA and HR-MRA. When using categorized MRA findings (A) and grade of MRA finding (B), C-MRA showed a worse performance than that shown by HR-MRA regardless of volume rendering (VR) use. The differences of AUC were also statistically significant.

Fig. 2

(A, B) An infundibulum (arrow) at the proximal left posterior cerebral artery (PCA) in a 31-year-old man on conventional magnetic resonance angiography (C-MRA) and high-resolution magnetic resonance angiography (HR-MRA) at 3 T. (A) C-MRA shows a probable aneurysmal lesion (arrow) at the proximal left PCA. (B) HR-MRA demonstrates junctional dilatation clearly and even visualizes a segment of the branching artery (arrow).

Fig. 3

(A–E) A 51-year-old woman with one small infundibulum located at the orifice of the left superior cerebellar artery (SCA). (A, B) Maximum intensity projection (MIP) and volume rendering (VR) images on conventional magnetic resonance angiography (MRA) show one small suspicious aneurysmal lesion at the left side of the basilar top (arrow). (C, D) MIP and VR images on high-resolution MRA reveal a branching left SCA at the lesion (arrow). (E) Digital subtraction angiography confirms an infundibulum located at the orifice of the left SCA (arrow).

Fig. 4

(A–C) A 57-year-old woman with an anterior communicating artery fenestration. (A) Conventional magnetic resonance angiography (MRA) shows a suspicious aneurysmal lesion in the anterior communicating artery (arrow). (B, C) Maximum intensity projection and volume rendering (VR) images on high-resolution MRA reveal duplication of the vascular structure, not an aneurysm, and the VR image clearly demonstrates an anterior communicating artery fenestration (arrow).

Fig. 5

(A–D) Multiple aneurysms at the left internal carotid artery (ICA) bifurcation site and anterior choroidal artery orifice in a 44-year-old woman. (A) Conventional magnetic resonance angiography (MRA) shows two suspicious aneurysmal lesions at the left ICA bifurcation (yellow arrow) and at the terminal segment (open arrow). (B) High-resolution MRA faintly shows a branching vascular structure (white arrow) adjacent to a bulging contour at the terminal segment of the left ICA, representing an infundibulum. (C, D) Digital subtraction angiography confirms two aneurysms located at the left ICA bifurcation (yellow arrow) and terminal segment of the left ICA (open arrow).

Table 1

Results obtained on assessment of lesions for cerebral aneurysm detection on C-MRA and HR-MRA (total=51 possible aneurysms)

Variable	C-MRA	HR-MRA	DSA
Without VR
Aneurysm	43	37	37
Definite	23	25	16
Suspicious	20	12	11
Infundibulum	6	12	12
Definite	3	7	6
Suspicious	3	5	6
No lesion	2	2	2
Sensitivity (%)	83.8	89.2
Specificity (%)	14.3	64.3

With VR
Aneurysm	43	38	37
Definite	30	29	16
Suspicious	13	9	11
Infundibulum	7	11	12
Definite	1	4	6
Suspicious	6	7	6
No lesion	1	2	2
Sensitivity (%)	89.2	94.6
Specificity (%)	21.4	71.4

Values are presented as number or % unless otherwise stated.

C-MRA, conventional magnetic resonance angiography; HR-MRA, high-resolution magnetic resonance angiography; DSA, digital subtraction angiography; VR, volume rendering.

Table 2

Diagnostic accuracy of various MRA techniques for the detection of cerebral aneurysm by using categorization or grading of MRA findings (aneurysm-based analysis of 51 possible aneurysms)

Variable	Sensitivity	Specificity	Accuracy	PPV	NPV	AUC (95% CI of AUC)	Comparison (P-value)^a)

							C-MRA vs. HR-MRA	C-MRA with VR vs. HR-MRA with VR	HR-MRA with VR vs. without VR
Using categorized MRA							<0.01	<0.01	0.16
C-MRA	6/37 (0.16)	12/14 (0.86)	18/51 (0.35)	6/8 (0.75)	12/43 (0.28)	0.5097 (0.3971–0.6222)
HR-MRA	33/37 (0.89)	10/14 (0.71)	43/51 (0.84)	33/37 (0.89)	10/14 (0.71)	0.8031 (0.6702–0.9359)
C-MRA with VR	33/37 (0.89)	4/14 (0.29)	37/51 (0.73)	33/43 (0.77)	4/8 (0.5)	0.5888 (0.4560–0.7217)
HR-MRA with VR	35/37 (0.95)	11/14 (0.79)	46/51 (0.9)	35/38 (0.92)	11/13 (0.85)	0.8658 (0.7483–0.9833)

Using the grade of MRA							0.27	0.24	0.14
C-MRA	25/37 (0.68)	14/14 (1)	39/51 (0.76)	25/25 (1)	14/26 (0.54)	0.8687 (0.7866–0.9509)
HR-MRA	25/37 (0.68)	14/14 (1)	39/51 (0.76)	25/25 (1)	14/26 (0.54)	0.9228 (0.8555–0.9901)
C-MRA with VR	29/37 (0.78)	13/14 (0.93)	42/51 (0.82)	29/30 (0.97)	13/21 (0.62)	0.8986 (0.8038–0.9935)
HR-MRA with VR	29/37 (0.78)	14/14 (1)	43/51 (0.84)	29/29 (1)	14/22 (0.64)	0.9614 (0.9199–1.0000)

Values are presented as number (%) or AUC (95% CI of AUC) unless otherwise stated.

MRA, magnetic resonance angiography; PPV, positive predictive value; NPV, negative predictive value; AUC, area under the receiver-operating characteristic curve; CI, confidence interval; C-MRA, conventional magnetic resonance angiography; HR-MRA, high-resolution magnetic resonance angiography; VR, volume rendering.

^a) The P-values for the comparison of the pairs of AUCs were calculated by Delong’s method.

Table 3

Evaluation of the added predictive ability of various MRA techniques for the detection of cerebral aneurysm by using categorization or grading of MRA findings (aneurysm-based analysis of 51 possible aneurysms)

Variable	NRI			IDI

	Index (%)	Z	P-value	Index (%)	Z	P-value
Using categorized MRA finding
C-MRA vs. HR-MRA	21.2	4.62	<0.01	36.7	4.42	<0.01
C-MRA with VR vs. HR-MRA with VR	46.3	6.32	<0.01	51.4	5.89	<0.01
HR-MRA without VR vs. HR-MRA with VR	46.3	6.32	<0.01	19.4	3.18	<0.01

Using the grade of MRA finding
C-MRA vs. HR-MRA	−13.9	−0.50	0.62	−2.9	0.31	0.01
C-MRA with VR vs. HR-MRA with VR	30.1	5.28	<0.01	16.2	1.57	0.12
HR-MRA without VR vs. HR-MRA with VR	73.0	15.39	<0.01	29.5	4.30	<0.01

MRA, magnetic resonance angiography; NRI, Net Reclassification Index; IDI, Integrated Discrimination Improvement; C-MRA, conventional magnetic resonance angiography; HR-MRA, high-resolution magnetic resonance angiography; VR, volume rendering.

Table 4

Diagnostic accuracy of various MRA techniques for the detection of cerebral aneurysms by size (aneurysm-based analysis of 51 possible aneurysms)

Variable	True-positive	True-negative	False-positive	False-negative	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Accuracy (%)
Total lesions (n=51)
C-MRA	34	1	11	5	82.1	8.3	74.4	12.5	64.7
C-MRA (VR)	34	3	9	5	87.2	25.0	79.1	37.5	72.5
HR-MRA	33	8	4	6	84.6	66.7	89.2	57.1	80.4
HR-MRA (VR)	35	9	3	4	89.7	75.0	92.1	69.2	86.3

Aneurysms ≥3 mm (n=22)
C-MRA	21			1	95.5
C-MRA (VR)	21			1	95.5
HR-MRA	21			1	95.5
HR-MRA (VR)	21			1	95.5

Aneurysms <3 mm (n=15)
C-MRA	10			5	66.7
C-MRA (VR)	12			3	80.0
HR-MRA	12			3	80.0
HR-MRA (VR)	14			1	93.3

MRA, magnetic resonance angiography; PPV, positive predictive value; NPV, negative predictive value; C-MRA, conventional magnetic resonance angiography; HR-MRA, high-resolution magnetic resonance angiography; VR, volume rendering.

REFERENCES

1. Meyers PM, Schumacher HC, Higashida RT, Derdeyn CP, Nesbit GM, Sacks D, et al. Reporting standards for endovascular repair of saccular intracranial cerebral aneurysms. Stroke 2009;40: e366-79.

2. Okahara M, Kiyosue H, Yamashita M, Nagatomi H, Hata H, Saginoya T, et al. Diagnostic accuracy of magnetic resonance angiography for cerebral aneurysms in correlation with 3D-digital subtraction angiographic images: a study of 133 aneurysms. Stroke 2002;33: 1803-8.

3. Li H, Yan L, Li MH, Li YD, Tan HQ, Gu BX, et al. Evaluation of intracranial aneurysms with high-resolution MR angiography using single-artery highlighting technique: correlation with digital subtraction angiography. Radiol Med 2013;118: 1379-87.

4. Masaryk TJ, Modic MT, Ross JS, Ruggieri PM, Laub GA, Lenz GW, et al. Intracranial circulation: preliminary clinical results with three-dimensional (volume) MR angiography. Radiology 1989;171: 793-9.

5. Keller PJ, Drayer BP, Fram EK, Williams KD, Dumoulin CL, Souza SP. MR angiography with two-dimensional acquisition and three-dimensional display: work in progress. Radiology 1989;173: 527-32.

6. Dumoulin CL, Hart HR Jr. Magnetic resonance angiography. Radiology 1986;161: 717-20.

7. Huston J 3rd, Nichols DA, Luetmer PH, Goodwin JT, Meyer FB, Wiebers DO, et al. Blinded prospective evaluation of sensitivity of MR angiography to known intracranial aneurysms: importance of aneurysm size. AJNR Am J Neuroradiol 1994;15: 1607-14.

8. Ross JS, Masaryk TJ, Modic MT, Ruggieri PM, Haacke EM, Selman WR. Intracranial aneurysms: evaluation by MR angiography. AJNR Am J Neuroradiol 1990;11: 449-55.

9. Sun ZK, Li YD, Li MH, Chen SW, Tan HQ. Detection of infundibula using three-dimensional time-of-flight magnetic resonance angiography with volume rendering at 3.0 Tesla compared to digital subtraction angiography. J Clin Neurosci 2011;18: 504-8.

10. Sahs AL. Cooperative study of intracranial aneurysms and subarachnoid hemorrhage: report on a randomized treatment study: I. Introduction. Stroke 1974;5: 550-1.

11. McCormick WF, Acosta-Rua GJ. The size of intracranial saccular aneurysms: an autopsy study. J Neurosurg 1970;33: 422-7.

12. Atkinson JL, Sundt TM Jr, Houser OW, Whisnant JP. Angiographic frequency of anterior circulation intracranial aneurysms. J Neurosurg 1989;70: 551-5.

13. Nakagawa T, Hashi K. The incidence and treatment of asymptomatic, unruptured cerebral aneurysms. J Neurosurg 1994;80: 217-23.

14. Ronkainen A, Miettinen H, Karkola K, Papinaho S, Vanninen R, Puranen M, et al. Risk of harboring an unruptured intracranial aneurysm. Stroke 1998;29: 359-62.

15. Wiebers DO, Whisnant JP, Huston J 3rd, Meissner I, Brown RD Jr, Piepgras DG, et al. Unruptured intracranial aneurysms: natural history, clinical outcome, and risks of surgical and endovascular treatment. Lancet 2003;362: 103-10.

16. Leffers AM, Wagner A. Neurologic complications of cerebral angiography: a retrospective study of complication rate and patient risk factors. Acta Radiol 2000;41: 204-10.

17. Naggara O, Louillet F, Touze E, Roy D, Leclerc X, Mas JL, et al. Added value of high-resolution MR imaging in the diagnosis of vertebral artery dissection. AJNR Am J Neuroradiol 2010;31: 1707-12.

18. Chung GH, Kwak HS, Hwang SB, Jin GY. High resolution MR imaging in patients with symptomatic middle cerebral artery stenosis. Eur J Radiol 2012;81: 4069-74.

19. Kadota T, Hosomi N, Kuroda C, Nakagawa H. Unruptured intracranial aneurysms: evaluation with high-resolution MR angiography with magnetization transfer contrast (MTC) and tilted optimized nonsaturating excitation (TONE). Nihon Igaku Hoshasen Gakkai Zasshi 1997;57: 853-9.

20. Mallouhi A, Schocke M, Judmaier W, Wolf C, Dessl A, Czermak BV, et al. 3D MR angiography of renal arteries: comparison of volume rendering and maximum intensity projection algorithms. Radiology 2002;223: 509-16.

21. Fellner FA, Fellner C, Wutke R, Lang W, Laub G, Schmidt M, et al. Fluoroscopically triggered contrast-enhanced 3D MR DSA and 3D time-of-flight turbo MRA of the carotid arteries: first clinical experiences in correlation with ultrasound, x-ray angiography, and endarterectomy findings. Magn Reson Imaging 2000;18: 575-85.

22. Maeder PP, Meuli RA, de Tribolet N. Three-dimensional volume rendering for magnetic resonance angiography in the screening and preoperative workup of intracranial aneurysms. J Neurosurg 1996;85: 1050-5.

23. Tsuchiya K, Katase S, Yoshino A, Hachiya J, Yodo K. Preliminary evaluation of volume-rendered three-dimensional display of time-of-flight MR angiography in the diagnosis of intracranial aneurysms. Neuroradiology 2001;43: 633-6.

24. Gibbs GF, Huston J 3rd, Bernstein MA, Riederer SJ, Brown RD Jr. Improved image quality of intracranial aneurysms: 3.0-T versus 1.5-T time-of-flight MR angiography. AJNR Am J Neuroradiol 2004;25: 84-7.

25. Monninghoff C, Maderwald S, Theysohn JM, Kraff O, Ladd SC, Ladd ME, et al. Evaluation of intracranial aneurysms with 7 T versus 1.5 T time-of-flight MR angiography: initial experience. Rofo 2009;181: 16-23.

26. Kapsalaki EZ, Rountas CD, Fountas KN. The role of 3 Tesla MRA in the detection of intracranial aneurysms. Int J Vasc Med 2012;2012: 792834.

27. Mallouhi A, Felber S, Chemelli A, Dessl A, Auer A, Schocke M, et al. Detection and characterization of intracranial aneurysms with MR angiography: comparison of volume-rendering and maximum-intensity-projection algorithms. AJR Am J Roentgenol 2003;180: 55-64.

28. Mallouhi A, Chemelli A, Judmaier W, Giacomuzzi S, Jaschke WR, Waldenberger P. Investigation of cerebrovascular disease with MR angiography: comparison of volume rendering and maximum intensity projection algorithms: initial assessment. Neuroradiology 2002;44: 961-7.