Prospective comparison of brain tumor detection and characterization using post-contrast Dixon uT1RESS and MPRAGE at 3 Tesla

Abstract

Purpose:

This study aimed to evaluate the image quality and lesion characterization with Dixon unbalanced T1 relaxation-enahnced steady-state (uT1RESS) in comparison to magnetization-prepared rapid acquisition gradient echo (MPRAGE) pulse sequence for detecting brain tumors, focusing on its potential to improve diagnostic accuracy and enhance brain tumor assessment.

Methods:

This single-center prospective study enrolled 20 patients (12 males, aged 35–83), with primary brain tumors and brain metastases. Both MPRAGE and Dixon uT1RESS were acquired at 3 T. Objective image quality was assessed by contrast-to-noise ratio (CNR), tumor-to-brain contrast, and tumor volume. Subjective image quality was assessed by two independent readers focusing on lesion visibility, lesion margins, motion, and static artifacts. A side-by-side comparison assessed diagnostic performance regarding lesion detection, evaluation of internal structure and vascular or dural invasion.

Results:

Dixon uT1RESS had a reduced acquisition time (2 min 51 s vs. 4 min 52 s for MPRAGE) and showed significantly higher CNR and tumor-to-brain contrast compared to MPRAGE (p < 0.001). Subjectively, both sequences showed similar overall image quality. Dixon uT1RESS achieved more conspicuous lesions with better-defined lesion margins (p < 0.001), while MPRAGE performed better in evaluating internal lesion structure (p < 0.05). Dixon uT1RESS was rated better for lesion detection, with three lesions additionally identified on this sequence. Wilcoxon signed-rank test was used to assess differences.

Conclusions:

Dixon uT1RESS significantly improved tumor conspicuity and detection, particularly for small metastatic lesions. This new technique offers promising potential for enhancing clinical brain tumor imaging, though additional sequences may be necessary for comprehensive lesion characterization.

1. Introduction

Magnetic resonance imaging (MRI) is recognized as the most accurate modality for assessing brain tumors and is adept at evaluating lesion morphology and enhancement characteristics [1]. In particular, contrast-enhanced MRI provides essential information for monitoring disease progression and treatment efficacy, as well as for determining therapeutic strategies, such as surgical resection or radiotherapy, as these decisions rely heavily on the number, size, and location of intracranial lesions [2–7].

Traditional contrast-enhanced T1-weighted imaging (T1WI) techniques, such as magnetization-prepared rapid acquisition gradient echo (MPRAGE), are commonly employed [7]. MPRAGE offers high spatial resolution and excellent tissue contrast between gray and white matter. However, the bright signal from enhancing blood vessels can obscure small lesions, particularly those near the cortical surface [8–11].

In recent years, advanced sequences have been developed to address these limitations. One such technique, the unbalanced T1 relaxation-enhanced steady-state (uT1RESS) pulse sequence, has demonstrated notable improvements in the visualization of brain tumors [12]. uT1RESS is designed to selectively suppress the signal intensity (SI) of non-enhancing background tissues while preserving the signal from gadolinium (Gd)-enhancing lesions, thereby enhancing diagnostic clarity. Initial studies have highlighted several advantages of this new technique in brain tumor imaging, including increased tumor-to-brain contrast as a consequence of the improved background tissue suppression, and consistent dark blood effect [13–15]. However, a detailed comparative analysis between uT1RESS and MPRAGE, the most commonly used standard of care technique, is lacking. Therefore, the aim of our study was to assess image quality and lesion characterization using Dixon uT1RESS compared to MPRAGE, to identify additional findings revealed by this new technique, and explore the potential benefits of incorporating it into standard imaging protocols for brain tumor assessment.

2. Methods

2.1. Patient cohort

This single center, prospective study was approved by the Institutional Review Board of the Medical University of South Carolina (Pro00128013), and written informed consent was obtained from all participants. This study was conducted in compliance with the Health Insurance Portability and Accountability Act to ensure the protection of patient privacy and security. Between August 2023 and May 2024, eligible patients undergoing contrast-enhanced brain MRI for known tumor were consecutively enrolled. Inclusion criteria were: (1) at least one contrast-enhancing primary or secondary intra-axial or extra-axial brain tumor; (2) ability to tolerate the MRI procedure; and (3) no contraindications to Gd-based contrast agents. Exclusion criteria included: (1) inability to complete the informed consent form; (2) safety-related contraindication to the MRI examination as determined by standard institutional guidelines and policies; (3) estimated glomerular filtration rate < 30 ml/min/1.73 m² within the past 60 days; and (4) pregnancy or lactating.

2.2. MRI acquisition protocol

MRI imaging was performed on a 3 Tesla system (MAGNETOM Skyra, Siemens Healthineers, Erlangen, Germany) equipped with a 20-channel head and neck coil (Skyra, Siemens). Standard clinical and research sequences were acquired, including localizers, pre-contrast axial MPRAGE, axial T2-weighted imaging, fluid-attenuated inversion recovery, post-contrast MPRAGE, and post-contrast Dixon uT1RESS acquisitions. Post-contrast imaging was performed after the intravenous administration of 0.1 mmol/kg gadobutrol (Gadavist, Bayer Healthcare, Malvern, PA, USA). 3D MPRAGE (T1-MPRAGE) acquisition was performed in the axial plane with TR, 1900 ms; TE, 2.54 ms; TI, 900 ms; flip angle, 8-9°; 1.0 mm isotropic resolution; 220 Hz/pixel bandwidth, and a scan time of 4 min 52 s. Finally, Dixon uT1RESS was acquired in the sagittal plane with TR, 539 ms; TE, 1.37 ms; 1.0 mm isotropic resolution, sampling bandwidth 975 Hz/pixel, and a scan time of 2 min 51 s. Detailed information on the image acquisition parameters is provided in Table 1. For the purpose of this study, only MPRAGE and Dixon uT1RESS images were evaluated. Multiplanar reconstructions were employed to generate axial, sagittal and coronal views for further analysis.

Table 1.

Scan parameters for MPRAGE and Dixon uT1RESS.

	MPRAGE	Dixon uT1RESS
Scan time	4 min 52 s	2 min 51 s
Signal averages	1	3
Voxel size (mm3)	1.0	1.0
Orientation	Axial	Sagittal
Field of view (mm)	256	256
Slices per slab	176	176
Shots per slice	1	1
Excitation	Spatially selective	Non-selective
k-Space trajectory	3D Cartesian	3D Cartesian
TR (ms)	1900	539
TE (ms)	2.54	1.37
TI (ms)	900	317
Echo spacing (ms)	6.6	4.6
Excitation flip angle (degrees)	8° – 9°	43°
Contrast modifying flip angle	n/a	90°
RF spoiling	Yes	No
Gradient spoiling factor	1.0	0.2
Bandwidth (Hz/pixel)	220	975

MPRAGE = Magnetization-prepared rapid acquisition gradient-echo; TE = Time to echo, TI = Time to inversion, TR = Time to repetition, uT1RESS = unbalanced T1 relaxation-enhanced steady-state.

2.3. Quantitative analysis

Quantitative image quality analysis was performed by a radiology trainee (A.T., 2 years of experience) using a picture archiving and communication system workstation (Sectra Medical, Sectra AB, Linkoping, Sweden). Region-of-interest measurements of SI were obtained in the enhancing brain lesions, normal white matter (WM), and air adjacent to the neurocranium. Tumor regions-of-interest encompassed the entire enhancing lesion. Given that signal-to-noise ratio per voxel exceeded the Rose threshold of 4 [16], a metric analogous to Weber contrast served as the primary measure of lesion visibility [17], and was calculated as follows:

$$\text{Weber contrast}=\frac{{\text{SI}}_{\text{tumor}}-{\text{SI}}_{\text{WM}}}{{\text{SI}}_{\text{WM}}}$$

where SI_tumor and SI_WM are the SIs measured in the enhancing lesion and the normal WM, respectively. All mention of “tumor-to-brain contrast” refer to the Weber contrast. Additionally, contrast-to-noise ratio (CNR) was employed as a secondary metric, calculated as follows [18]:

$$\text{CNR}=\frac{{\text{SI}}_{\text{tumor}}-{\text{SI}}_{\text{WM}}}{{\text{SD}}_{\text{air}}}\times 0.6555$$

where SI_tumor and SI_WM are the SIs measured in the enhancing lesion and the normal WM, respectively, while SDair is the standard deviation (SD) of the air adjacent to the neurocranium.

Volumetric 3D tumor measurements were conducted using an open-source software (3D Slicer, version 5.6.2., Massachusetts Institute of Technology and Birgham and Women’s Hospital, Boston, MA, USA) [19]. For the volumetric assessment, one lesion per patient was selected based on specific criteria: a minimum lesion diameter of 10 mm, absence of cystic cavities, as homogeneous enhancement as possible, and well-defined lesion margins [20]. The lesion that best met these criteria was chosen for the analysis. After image co-registration, the MPRAGE and Dixon uT1RESS datasets were reviewed independently. 2D segmentation of the lesions was performed on perpendicular slices and automatically 3D-interpolated by the software. Margins of the lesions were refined manually to help the region-growing algorithm. To evaluate the reliability of the measurements, segmentations were repeated for all lesions during a subsequent session. The average of the two measurements was analyzed to determine differences between the sequences.

2.4. Qualitative analysis

MPRAGE and Dixon uT1RESS images were presented in a randomized order to two board-certified radiologists with 5 (M.T.H.) and 6 (D.K.) years of experience in interpreting brain MRI studies. Readers were blinded to both clinical and demographic information and were un-aware of the quantitative results. Initially, both readers assessed the individual MPRAGE and Dixon uT1RESS images. The longest lesion diameter [20] and the product of the two longest perpendicular diameters [21] of one selected lesion per patient were taken from a single axial section of the images. Subsequently, readers evaluated the following aspects of qualitative image quality based on all detected lesions, using both source images and multiplanar reformations: (1) overall image quality; (2) tumor conspicuity; (3) assessment of lesion margins; (4) motion artifacts; and (5) static artifacts. These assessments were scored on a 4-point Likert scale (1 – unacceptable; 4 – excellent). Finally, a side-by-side comparison was conducted to directly evaluate (1) diagnostic performance, (2) internal lesion structure, and (3) the presence of dural or vascular invasion. Diagnostic performance of lesion detection was scored on a 5-point scale: − 2, one or more enhancing lesions only shown by MPRAGE; − 1, one or more lesions better shown by MPRAGE; 0, lesions equally well shown by MPRAGE and Dixon uT1RESS; +1, one or more enhancing lesions better shown by Dixon uT1RESS; +2, one or more lesion only shown by Dixon uT1RESS. Internal lesion structure and the presence of dural or vascular invasion was scored on a 3-point scale: − 1, better shown by MPRAGE; 0, equally well shown by MPRAGE and Dixon uT1RESS; +1, better shown by Dixon uT1RESS.

2.5. Statistical analysis

Patient characteristics were summarized descriptively. The normality of data distribution was evaluated using the Shapiro-Wilk test. For qualitative and non-normally distributed quantitative image quality scores, the two-tailed Wilcoxon signed-rank test was used to assess differences. Scores from the qualitative analysis were aggregated across readers. A one-tailed Wilcoxon signed-rank test was employed to determine if MPRAGE or Dixon uT1RESS enhanced diagnostic performance, internal structure assessment, and the detection of vascular or dural involvement. Within-sequence reproducibility of 3D volume measurements was assessed by calculating the intraclass correlation coefficients with a 2-way mixed consistency, average-measured approach (0.0 – 0.5; poor agreement, 0.51 – 0.75; moderate agreement, 0.76 – 0.9; good agreement, above 0.9; excellent agreement). Inter-reader agreement of qualitative scores between the two readers was quantified with weighted κ statistics (< 0.20: slight agreement; 0.21 – 0.40: fair agreement; 0.41 – 0.60: moderate agreement; 0.61 – 0.80: substantial agreement, > 0.81: almost perfect agreement). The Benjamini-Hochberg method was applied to correct for multiple comparisons, and a p-value less than 0.05 was considered statistically significant. Unless stated otherwise, all data are presented as mean ± SD. All statistical analyses were performed using IBM, SPSS, version 28.0.1.

3. Results

3.1. Patient population

Twenty adult patients (8 females; median age, 65.5 years; IQR, 52 – 74.25 years) were enrolled in this study. The identified pathologies included primary brain tumors (n = 8) and brain metastases (n = 12). Table 2 provides patient characteristics and detailed information about the identified lesions.

Table 2.

Patient demographics and tumor types.

Characteristics	All patients (N = 20)
Sex
Male	12 (60 %)
Female	8 (40 %)
Age, years (median [IQR])	65.5 [52 – 74.25]
Tumor type
Meningioma, n (%)	3 (15)
Glioblastoma, n (%)	2 (10)
Astrocytoma, n (%)	2 (10)
Hemangiopericytoma, n (%)	1 (5)
Metastatic disease, n (%)	12 (60)
Lung, n (%)	3 (15)
Melanoma, n (%)	2 (10)
Renal, n (%)	2 (10)
Colon, n (%)	1 (5)
Other/Unknown origin, n (%)	4 (20)

IQR = interquartile range; N = number.

3.2. Quantitative comparisons

Both CNR and tumor-to-brain contrast were significantly higher with Dixon uT1RESS compared to MPRAGE. Specifically, Dixon uT1RESS achieved nearly a three-fold increase in tumor-to-brain contrast, with mean values of 0.5 ± 0.5 for MPRAGE and 1.4 ± 0.7 for Dixon uT1RESS (p < 0.001).

Regarding tumor volume measurements, both 2D and 3D assessments using Dixon uT1RESS indicated slightly larger volumes compared to MPRAGE. The mean lesion area was 2.7 ± 3.4 cm² and 2.4 ± 3.3 cm² for Dixon uT1RESS and MPRAGE, respectively, although this difference was not statistically significant (p = 0.21). In contrast, the 3D volume measurements indicated a significant difference, with Dixon uT1RESS reporting a mean volume of 4.5 ± 8.6 cm³ compared to 4.2 ± 8.4 cm³ for MPRAGE (p < 0.001).

Within-sequence reproducibility of 3D volume measurements was found to be almost perfect (0.98, 95 % CI 0.97–0.99).

3.3. Qualitative comparisons

Both imaging sequences exhibited comparable image quality (3.5 ± 0.7 for Dixon uT1RESS vs 3.4 ± 0.7 for MPRAGE, p = 0.09) and static artifacts (3.2 ± 0.7 for Dixon uT1RESS vs 3.5 ± 0.7 for MPRAGE, p = 0.06). Dixon uT1RESS was rated better for motion artifacts compared with MPRAGE (3.7 ± 0.5 vs 3.4 ± 0.6, respectively, p = 0.005). Lesions consistently appeared more conspicuous with Dixon uT1RESS (3.9 ± 0.4 vs 3.4 ± 0.7, p < 0.001), with better defined lesion margins (3.9 ± 0.4 vs 2.9 ± 0.9, p < 0.001) (see Table 3 and Figs. 1–3).

Table 3.

Quantitative and qualitative comparisons.

	MPRAGE	Dixon uT1RESS	p
Quantitative analysis
CNR	43.0 ± 48.2	54.7 ± 37.7	0.02
Tumor-to-brain contrast	0.5 ± 0.5	1.4 ± 0.7	< 0.001
2D Lesion size (cm2)	2.4 ± 3.3	2.7 ± 3.4	0.21
3D Lesion volume (cm3)	4.2 ± 8.4	4.5 ± 8.6	< 0.001
Qualitative Analysis
Image quality	3.4 ± 0.7	3.5 ± 0.7	0.09
Motion artifacts	3.4 ± 0.6	3.7 ± 0.5	0.005
Static artifacts	3.5 ± 0.7	3.2 ± 0.7	0.06
Lesion conspicuity	3.4 ± 0.7	3.9 ± 0.4	< 0.001
Lesion margins	2.9 ± 0.9	3.9 ± 0.4	< 0.001

MPRAGE = Magnetization-prepared rapid acquisition gradient-echo; uT1RESS = unbalanced T1 relaxation-enhanced steady-state.

Fig. 1. Two patients with small brain metastases.

Sagittal plane contrast-enhanced MPRAGE (a, c) and Dixon uT1RESS (b, d) images show small lesions (arrows) with faint enhancement on the MPRAGE images, appearing more conspicuous on Dixon uT1RESS.

Fig. 3. Stacked bar charts display the result of the qualitative image quality analysis from the individual assessment of the MPRAGE and Dixon uT1RESS images.

The images were rated by two independent readers using a 4-point Likert scale (1 – unacceptable; 4 – excellent).

Dixon uT1RESS demonstrated a significantly better lesion detection rate than MPRAGE (p < 0.001). Reader 1 deemed Dixon uT1RESS superior in identifying lesions in 7 out of 20 cases, while reader 2 noted that it enhanced the visibility of one or more lesions in 9 cases. Additionally, three lesion were detected only on Dixon uT1RESS images. These lesions were later identified on MPRAGE and were confirmed as metastatic lesions by the readers. Most cases in which Dixon uT1RESS improved diagnostic confidence involved small brain metastases from various origins (Fig. 4). In no case did MPRAGE surpass Dixon uT1RESS for lesion detection.

Fig. 4. Diverging bar chart displays the result of the qualitative image quality analysis from the side-by-side assessment of the MPRAGE and Dixon uT1RESS images.

The images were rated by two independent readers using ordinal scales. Example scale used to assess diagnostic performance: (− 2) one or more enhancing lesions only shown by MPRAGE; (− 1) one or more lesions better shown by MPRAGE; (0) lesions equally well shown by MPRAGE and Dixon uT1RESS; (+1) one or more enhancing lesions better shown by Dixon uT1RESS; (+2) one or more lesion only shown by Dixon uT1RESS.

In contrast, the assessment of internal lesion structure was rated equal to or better with MPRAGE in the majority of cases (p < 0.05). MPRAGE was preferred for the evaluation of internal lesion structure in 7 cases by Reader 1 and in 5 cases by Reader 2, comprising 40 % primary brain tumors and 60 % brain metastases, respectively (Fig. 5).

Fig. 5. 65-year-old male with hemangiopericytoma.

MPRAGE (a, sagittal; c, coronal) and Dixon uT1RESS (b, sagittal; d, coronal) images show the complex, heterogeneously enhancing left periventricular mass, extending along the thalamus and corpus callosum, with a peripherally enhancing cystic component (arrows).

For detecting vascular or dural involvement, Dixon uT1RESS generally showed improvement, as assessed both by Reader 1 (8 out of 20 cases) and Reader 2 (4 out of 20 cases) (p < 0.001). Notably, MPRAGE was never rated higher than Dixon uT1RESS for the assessment of vascular or dural involvement.

Interreader agreement for the qualitative image scores was moderate (κ = 0.58).

4. Discussion

In this study, we compared the image quality and diagnostic performance of two 3D post-contrast T1-weighted sequences, MPRAGE and Dixon uT1RESS, for imaging primary brain tumors and brain metastases. The main results are as follows: i) Dixon uT1RESS yielded a significantly improved CNR and tumor-to-brain contrast; ii) both uT1RESS and MPRAGE maintained a similarly high overall image quality; and iii) the use of uT1RESS resulted in the detection of three lesions primarily not seen with MPRAGE.

The enhanced detection of smaller metastases holds crucial clinical implications, potentially influencing prognosis and treatment strategies, such as the selection between whole-brain and targeted radiation therapy or the planning of stereotactic radiosurgery, as these additional lesions can be treated simultaneously [6]. Dixon uT1RESS was highly effective at suppressing intravascular signal. However, as with other dark blood imaging techniques, such as T1 Sampling Perfection with Application-optimized Contrast using different flip angle Evolution (SPACE)[22], flow suppression can be incomplete in some small veins with very slow or absent flow. A recent study by Lasocki et al. highlighted similar issues, reporting improved detection of small melanoma metastases with T1 SPACE, but also noting an increased false-positive burden attributed to vascular artifacts as correlated with MPRAGE [23]. As the authors state, while 3D turbo spin-echo sequences alone can result in increased false positive rates, correlating findings with 3D gradient-recalled echo may help mitigate these uncertainties. Moreover, familiarity with the imaging features of the sequence can assist readers in recognizing residual blood signals, potentially reducing false-positive rates [24]. Future research should investigate the diagnostic utility of the uT1RESS technique specifically in patient populations with small metastases and conduct a systematic assessment of false-positive rates alongside the additional lesions detected. Such investigations could clarify the technique’s impact on treatment strategies.

Another promising aspect of the Dixon uT1RESS technique is its potential to reduce contrast media usage. Preliminary results from studies using stack-of-stars echo-uT1RESS suggested that a half-dose of contrast could be effective compared to full-dose post-contrast MPRAGE [15]. While these findings require validation in larger cohorts, they present an exciting opportunity for integrating uT1RESS into clinical practice, particularly for patients at risk of nephrogenic systemic fibrosis or Gd deposition disease, especially in patient cohort that undergo repetitive imaging in a short time frame [25].

Regarding volumetric assessment, we observed that tumors appeared larger in both 2D and 3D measurements. We hypothesize that this discrepancy may arise from the improved delineation of lesion margins, facilitated by the enhanced contrast ratio of the technique and potentially the sequence acquisition order, given that Dixon uT1RESS was always acquired after MPRAGE. This aspect warrants further investigation, particularly in light of findings by Danieli et al. [26], which reported similar morphometric discrepancies across various MRI techniques independent of acquisition order, emphasizing the need to consider these factors in longitudinal lesion follow-up. The acquisition order might also figure in the superior performance of MPRAGE for characterizing the internal structure of enhancing lesions. Additionally, while uT1RESS excels at detecting small enhancing lesions, it lacks sufficient detail for accurate gray/white matter segmentation due to the bland appearance of normal brain tissue. Therefore, additional sequences may be necessary before or after uT1RESS acquisition to obtain comprehensive morphological information.

Our study has some limitations: first, the relatively small patient population limited the ability to draw conclusions about specific tumor types, mandating confirmatory research. Second, prior treatment of some lesions complicated the analysis of treatment-related changes. Third, performing Dixon uT1RESS after MPRAGE may have introduced bias due to the time-dependent nature of tumor enhancement, potentially aiding the detection of small lesions. Fourth, it is important to note that our current study was not designed to make definitive conclusions regarding false-positive rates due to the diverse nature of tumor etiologies and the relatively small cohort size. Finally, the absence of biopsy confirmation means not all enhancing lesions may represent true tumors, necessitating long-term clinical and imaging follow-up studies to determine their nature and clinical significance.

5. Conclusions

In conclusion, our study highlights the superior diagnostic capabilities of the Dixon uT1RESS technique over MPRAGE in imaging primary brain tumors and metastases, particularly in detecting smaller intracranial metastases. Further research is warranted to assess its application in diverse patient populations and to clarify its implications for treatment planning.

Fig. 2. Multifocal glioblastoma in a 75-year-old male.

The figure displays sagittal (a) and axial (b) plane MPRAGE; and sagittal (c) and axial (d) plane Dixon uT1RESS images. Dixon uT1RESS demonstrates enhanced lesion conspicuity, particularly highlighting the low Gd-enhancing lesion (indicated by arrows) through improved tumor-to-brain contrast compared to MPRAGE. Lesion margins were also better appreciated with Dixon uT1RESS.

Abbreviations:

CNR

Contrast-to-Noise Ratio

MPRAGE

Magnetization-Prepared Rapid Acquisition Gradient-Echo

SI

Signal Intensity

SPACE

Sampling Perfection with Application-optimized Contrast using different flip angle Evolution

TE

Time to Echo

TI

Time to Inversion

TR

Time to Repetition

uT1RESS

Unbalanced T1 Relaxation-Enhanced Steady-State