Synopsis

Keywords: Tumors, Tumor, Metastasis

We compared standard T1-weighted thin-slice SE and T1-weighted SPACE sequences in the detection of metastatic brain lesions at 1.5 and 3 Tesla in 56 patients. Three raters evaluated the presence of lesions in 2 sessions minimally 6 weeks apart. Our results show that T1-weighted SPACE is not inferior to standard thin-slice SE sequences in the detection of brain metastases. All three experienced raters reached excellent consistency between SE and SPACE and agreement with ground truth.

INTRODUCTION

The accurate detection of metastatic brain lesions before radiotherapy is critical. Although spin echo (SE) are superior to gradient echo sequences in detecting small metastatic brain lesions^1,2, thin-slice whole-brain coverage is time-consuming. The sampling perfection with application optimized contrasts using different flip angle evolution (SPACE) sequence shares many advantages with SE, but with shorter acquisition times and greater resolution, while achieving whole-brain coverage and allowing multiplanar reconstruction. We therefore aimed to compare the accuracy of three readers in detecting metastatic brain lesions using thin-slice T1-weighted 2D SE and T1-weighted SPACE sequences with whole-brain coverage, at both at 1.5 and 3 Tesla field strengths.

METHODS

Fifty-six patients with suspected or known brain metastases were included and underwent a standard brain imaging protocol, including standard proton density-weighted, T2-weighted and fluid-attenuated inversion recovery sequences, as well as standard diffusion-weighted imaging. All subjects were additionally imaged with a thin-slice T1-weighted 2D SE sequence (1.5 T: TR=483–630 ms, TE=17 ms, in-plane resolution=0.94x0.94 mm, slice thickness=3 mm, slice gap=3.3 mm; 3 T: TR=640–871 ms, TE=8.2–10 ms, in-plane resolution=0.43x0.43 mm, slice thickness=2–3 mm, slice gap=2.4–3.9 mm) and a T1-weighted SPACE sequence (1.5 T: TR=500 ms, TE=26 ms, resolution=1 mm isotropic; 3 T: TR=700 ms, TE=11–12 ms, in-plane resolution=0.75–0.91 mm, slice thickness=0.9–1 mm). Seven of the SPACE sequences at 3 Tesla had 0.9 mm isotropic resolution, while the remainder had slightly higher in-plane resolution with a slice thickness of 1 mm. The median acquisition time for the SE sequences was 10.4 minutes (range 8.3–16.8) at 1.5 Tesla and 9.9 minutes (range 7.7–15.7) at 3 Tesla, while the acquisition time of the SPACE sequences was 4.6 minutes at 1.5 Tesla and median 4.3 minutes (range 3.9–5) at 3 Tesla. The order of the T1-weighted sequences after contrast injection varied, such that in 26 patients the SE sequence preceded the SPACE sequence, while in 30 patients the SPACE sequence preceded the SE sequence. Gadolinium-based contrast agents were administered at standard doses; 54 of 56 patients received Gadobutrol (1 molar). Rating was performed in 2 sessions minimally 6 weeks apart. The images were de-identified and the order was randomized for each session. The raters were instructed to count the number of likely metastatic lesions in the brain. The raters had access to only T1-weighted post-contrast images, either SE or SPACE, in each session, and each subject had only 1 sequence (SE or SPACE) in each session. Therefore, each rater evaluated 56 image volumes, covering 56 subjects, per session. The true number of metastatic lesions was determined after rating was finished, using all sequences available for that session, as well as radiological reports and follow-up imaging. Inter- and intra-rater metrics were determined by intraclass correlations (ICC), specifically, consistency for intra-rater evaluation (SE vs. SPACE) and agreement for inter-rater assessment (same sequence). A paired t-test was used to evaluate the impact of sequence order with respect to contrast administration (first or second).

RESULTS

A total of 135 metastatic lesions were identified in 56 subjects, and considered as the true number of lesions (mean per subject 2.41, SD 6.4, range 0–46). Relatively fewer lesions were identified as suspected metastasis on the sequence temporally closer to contrast administration (irrespective of sequence type), however the difference did not reach statistical significance at the pooled level (p=0.08), nor at the level of individual raters (R1, p=0.16; R2, p=0.19; R3, p=0.5). Intra-rater consistency (SE vs. SPACE) was excellent in all 3 raters (ICC: R1, 0.984; R2, 0.971; R3, 0.946). Inter-rater agreement was also excellent in all 3 raters, with ICC values of 0.984 and 0.969 for SE and SPACE sequences, respectively. Finally, agreement between individual sequences and the true number of lesions was excellent in all raters (SE ICC: R1, 0.981; R2, 0.973; R3, 0.977; SPACE ICC: R1, 0.984; R2, 0.971; R3, 0.965).

DISCUSSION

The emergence of fast computer-assisted treatment planning of targeted radiosurgery techniques in the management of brain metastases requires precise, fast and reliable MRI workup. The reliable detection of metastatic brain lesions with MRI depends on a number of factors, particularly pulse sequence and contrast agent type, dose and application delay³. Although SE sequences are superior to gradient echo sequences in the detection of small brain metastases, they have relatively long acquisition times and are prone to artifacts. To our knowledge this is the first comparison between thin-slice T1-weighted SE and SPACE sequences in the detection of brain metastases at 1.5 Tesla and 3 Tesla. Our results show that T1-weighted SPACE is not inferior to standard thin-slice SE sequences in the detection of brain metastases. All three experienced raters reached excellent consistency between SE and SPACE and agreement with ground truth.

Acknowledgements

Supported by Ministry of Health CZ, IG204303.