Prospective comparison of volumetric post-contrast T1-Sampling Perfection with Application optimized Contrasts by using different flip angle Evolutions and Magnetization-Prepared Rapid Acquisition with Gradient Echo in patients with metastatic melanoma

Abstract

Introduction

Volumetric turbo spin echo (3D-TSE) T1-weighted imaging techniques such as T1-SPACE (Sampling Perfection with Application optimized Contrasts by using different flip angle Evolutions) improve detection of intracranial metastases (IM) compared to volumetric magnetisation-prepared gradient recalled echo techniques such as MPRAGE (Magnetization-Prepared Rapid Acquisition with Gradient Echo). However, incomplete vascular suppression can produce false positives when using 3D-TSE. Research into 3D-TSE has generally targeted patients with known or suspected IM, but the clinical implications of false positives are greater in patients with lower likelihood of IM. This study examined additional findings identified by T1-SPACE in patients with metastatic melanoma, targeting patients with a lower incidence of IM.

Methods

Patients with metastatic melanoma and an upcoming brain MRI booking were identified prospectively. Consent for adding post-contrast T1-SPACE to the MRI protocol (which included MPRAGE) was obtained. Imaging was initially assessed without T1-SPACE. Subsequently, T1-SPACE images were examined and additional findings identified were recorded, including their correlation with MPRAGE.

Results

One hundred examinations were performed, 24 having evidence of active IM. T1-SPACE allowed identification of additional lesions in five patients, including two with small solitary IM not identified when first assessing MPRAGE. In 18 examinations, T1-SPACE identified additional equivocal findings, confidently attributed to artefact (most commonly normal vessels) following correlation with MPRAGE.

Conclusion

T1-SPACE improves detection of small lesions in patients without known IM, changing patient management. False positives are common but can be clarified with MPRAGE. Combining T1-SPACE and MPRAGE allows both sensitivity and specificity to be optimised.

Introduction

Volumetric turbo spin echo (3D-TSE) T1-weighted imaging (T1WI) techniques are increasingly replacing volumetric magnetisation-prepared gradient recalled echo (3D-GRE) sequences such as MPRAGE (Magnetization-Prepared Rapid Acquisition with Gradient Echo). Examples of 3D-TSE sequences include T1-SPACE (Sampling Perfection with Application optimised Contrasts by using different flip angle Evolutions), CUBE, and VISTA (Volumetric ISotropic Turbo spin echo Acquisition). Key advantages of 3D-TSE over 3D-GRE include inherently greater suppression of normal vessels, which can otherwise obscure small metastases adjacent to vessels, ¹ and greater lesional enhancement after contrast administration.^1–4 Furthermore, the white matter is not as hyperintense on 3D-TSE compared with 3D-GRE, which may facilitate the appreciation of enhancement for lesions involving the white matter. ¹ These features combine to improve the detection of small IM,^3,5 although larger IM are generally identifiable with either technique. ⁶ As a result of these advantages, 3D-TSE has been included in recent consensus recommendations for a standardised brain tumour imaging protocol for clinical trials in brain metastases (BTIP-BM), as part of the ‘ideal’ MRI protocol. ¹

Despite the compelling advantages of 3D-TSE, there are some disadvantages. Most importantly, there is the potential for false positive (FP) findings, most commonly due to incompletely suppressed vessel segments mimicking lesions such as a metastasis or a meningioma. ⁶ Often, such a finding can be seen to be continuous with a suppressed vessel, and thus correctly interpreted, but occasionally this is less clear. This uncertainly can, however, be resolved by comparing with 3D-GRE, as the more homogeneous vascular enhancement provided by this technique facilitates attribution to a normal vessel. ⁶ 3D-TSE also has a higher FP rate due to artefacts, ⁶ and such artefacts may vary between the two sequences. ¹ For this reason, there is value in performing both 3D-TSE and 3D-GRE – 3D-TSE provides greater sensitivity, while 3D-GRE improves specificity by resolving possible FP.

Understandably, research into 3D-TSE has generally targeted patients with known or suspected IM, and the BTIP-BM has been specifically developed for clinical trials in patients with known IM. ¹ In routine clinical practice, however, many patients undergoing a brain metastasis protocol MRI do not have known active IM; rather, MRI is being used to screen for asymptomatic IM. The clinical implications of an FP vary depending on the clinical setting. For example, incorrectly identifying an additional IM in a patient with multiple other IM may have limited impact on their management. In contrast, if a solitary IM were to be incorrectly reported in a patient without a prior diagnosis of IM, this could have substantial negative implications, for example, cessation of an effective therapy, unnecessary use of stereotactic radiosurgery (SRS) or exclusion from a clinical trial. Thus, while the BTIP-BM is likely to be a very effective protocol beyond the specific setting for which it was designed, there are additional considerations in routine clinical practice.

Intracranial metastases from melanoma have some differences compared to IM from other primaries. Firstly, they may exhibit intrinsic T1 hyperintensity due to melanin content. ⁷ Secondly, most IM develop at the grey-white matter junction, related to tapering of cortical vessels. ⁸ In contrast, recent research has shown that most small intracranial melanoma metastases (IMM) develop at the interface between the cortex and meninges, suggesting the pia mater as a preferential portal of entry. ⁹ This closer proximity to the sulcal vessels, in contrast to IM from other histologies, suggests that IMM may be at greater risk of misdiagnosis. For example, 3D-GRE may have a greater risk of missing IMM due to obscuration by adjacent normal vessels. Equally, 3D-TSE may produce a higher FP rate, given that incompletely suppressed vessels will be located closer to the expected location of IMM. Another important consideration is that brain MRI can be included as part of routine screening for patients with either known extracranial metastases from melanoma or a high-risk primary, ¹⁰ while with some other histologies, MRI may only be performed if specifically prompted by the patient’s symptoms. Thus, while melanoma frequently metastasises to the brain, ¹¹ the majority of MRI examinations in such a context are expected to be negative.

The purpose of this study was to assess the utilisation of both 3D-GRE and 3D-TSE in patients with metastatic melanoma across a broader and more typical clinical context, targeting patients with either no known IM or low-volume IM, to assess the utility of replacing 3D-GRE with 3D-TSE in our routine clinical brain metastases protocol. This included contrasting the strengths and weaknesses of the two sequences, assessing the added value of utilising both sequences and developing practical insights.

Methods

Patient identification

Institutional Human Research Ethics Committee approval was obtained. Each week, patients were identified prospectively on the basis of an upcoming brain MRI booking and a history of metastatic melanoma. Only patients with either no known active IMM or low-volume IMM (no IMM >10 mm in long axis), based on the preceding MRI report, were included. This included patients with successfully treated IMM. Verbal consent for the addition of a post-contrast T1-SPACE sequence to the MRI protocol was obtained from each patient. The additional sequence was performed when logistics allowed (i.e. when the MRI list was not behind schedule). A cohort of 100 examinations was planned; power analysis was not feasible given the uncertainty regarding the proportion of positive examinations and the added value of T1-SPACE. Examinations were performed between February 2020 and October 2020.

Image acquisition

All MRI examinations were performed on a 3-Tesla MRI (MAGNETOM Skyra, Siemens, Erlangen, Germany) utilising a standardised protocol, including pre- and post-contrast 3D-GRE (MPRAGE with water excitation), post-contrast 3D-TSE (T1-SPACE, with fat saturation), axial T2-weighted imaging, axial fluid attenuated inversion recovery, diffusion-weighted imaging and susceptibility-weighted imaging. Post-contrast T1-SPACE was performed after post-contrast MPRAGE, while T2WI was performed between contrast administration and post-contrast MPRAGE. MPRAGE was performed in the axial plane (TR: 1900 ms; TE: Min; flip angle: 8; frequency: 256; phase: 256; NEX: 1; FOV: 256 mm; slice thickness: 1 mm; fat suppression: water excitation; parallel imaging: 2). Initially, T1-SPACE was also performed in the axial plane (for the first 30 examinations), but wrap artefact occurred at the superior aspect of the field-of-view in some patients. As a result, the acquisition plane was changed to sagittal (for the remaining 70 examinations), although images were primarily assessed in the axial plane with 1.0 mm slice thickness, to be equivalent to MPRAGE. The parameters for T1-SPACE were TR: 700; TE: 11; frequency: 256; phase: 256; NEX 1.4; FOV: 256; slice thickness: 1 mm; fat Suppression: fat sat strong; blood suppression: free; parallel imaging: 2.

Imaging review

All MRI examinations were assessed by a single study neuroradiologist (with 8 years of subspeciality experience in neuro-oncology), blinded to the clinical history. Firstly, all sequences other than post-contrast T1-SPACE were assessed. Secondly, T1-SPACE images were assessed, noting additional findings identified by T1-SPACE. MPRAGE images were subsequently reviewed, to determine whether the additional findings demonstrated by T1-SPACE were considered consistent with additional metastases, convincingly FP (most commonly normal vessels) or remained equivocal. Imaging was assessed either as the initial step in the routine diagnostic read, or later (when reported by other radiologists at the institution). When reported by other radiologists, the imaging report was subsequently reviewed after interpretation for quality assurance purposes.

Results

One hundred examinations including post-contrast 3D-TSE were performed, in 93 patients (seven patients participated in the study on two separate occasions, although the study radiologist was unaware of their prior participation at the time of assessment). Reflecting a typical patient population with metastatic melanoma, fifty-four patients (58%) were male and the median patient age was 63 years (range 27–87 years). In 61 of the 100 examinations, there was no prior history of IMM. Eighty-four examinations were performed as follow-up after previous MRI(s); in 16 patients, the study MRI was the first MRI performed for metastatic melanoma to our knowledge. These background data are summarised in Table 1. For 24 examinations, the study read occurred at the commencement the routine clinical report; in the remaining 76, the study read occurred after the routine clinical read (by other radiologists).

Table 1.

Background patient data.

Number of study MRIs per patient
1	93 (93%)
2	7 (7%)
Patient sex
Male	54 (58%)
Female	39 (42%)
Previous history of IMM
Yes	39 (39%)
No	61 (61%)
Previous MRIs
Yes	84 (84%)
No	16 (16%)

Seventy-six examinations demonstrated no evidence of active IMM when combining both sequences (including stable residua of treated disease). Of the 24 examinations with IMM, T1-SPACE allowed the identification of additional lesions in five patients. In two of these patients, small solitary IMM (measuring 5 mm and 4 mm) were not identified when first assessing MPRAGE but were visible on T1-SPACE (Figure 1). Of note, both had also been missed during the routine clinical read (by non-study radiologists). In one of these patients, the imaging review as part of the study prompted earlier follow-up imaging; this confirmed an IMM, which was subsequently treated with SRS. In the other patient, immunotherapy (with ipilimumab and nivolumab) had already been commenced, and this produced a complete response in both the intra- and extracranial metastatic disease. Three further patients had some IMM visible on MPRAGE, but T1-SPACE identified additional IMM. In all five patients, the additional lesions identified on T1-SPACE were able to be subsequently identified on review of MPRAGE images.

Figure 1.

Axial post-contrast MPRAGE (left) and T1-SPACE (right) images in the two patients (top row and bottom row) in whom solitary intracranial metastases (arrowed) were not identified prospectively using MPRAGE, but were seen on T1-SPACE images. In both cases, the metastases are more conspicuous with T1-SPACE.

In 18 examinations, T1-SPACE identified additional equivocal findings which could not definitely be attributed to normal vessels when using T1-SPACE alone but were confidently attributed to artefact on subsequent review of the MPRAGE images. Susceptibility-weighted imaging also added confidence in some cases (Figure 2). Seventeen were considered to represent normal vessels, nine located in the cerebral cortex, six in the subarachnoid space and two at the periphery of the cerebellar hemispheres. The remaining finding was located in the subcortical white matter and was attributed to motion artefact. Thus, based on correlation with MPRAGE, these findings were considered normal, and no additional imaging was required beyond routine follow-up. For 10 of the 18, subsequent routine follow-up imaging supported that the findings were FP; in the remaining eight, later follow-up imaging was not yet available at the time of study completion. Figure 3 summarises the additional findings identified on T1-SPACE.

Figure 2.

Vessels within the brain parenchyma may mimic parenchymal metastases on T1-SPACE (left, arrowed) if the remainder of the vessel cannot be identified due to flow suppression. Correlation with MPRAGE (middle) usually allows visualisation of communication with a vessel in the subarachnoid space. Such a communication may sometimes also be seen on Susceptibility-Weighted Imaging (right). The left cingulate gyrus metastasis is well visualised with both T1-SPACE and MPRAGE.

Figure 3.

Description of the additional findings identified using T1-SPACE.

Discussion

Our findings are consistent with the existing literature, confirming that 3D-TSE improves detection of intracranial metastases. The identification of solitary small IMM with the aid of T1-SPACE substantially altered the management of one of these two patients, by allowing treatment at a smaller size. In a further three patients, T1-SPACE identified more IMM than MPRAGE. Identifying additional IMM does not necessarily alter management, if the plan is for systemic therapy (as occurred in one of the patients with a small solitary IMM). Nevertheless, there is potential benefit in identifying additional metastases if SRS is planned as these additional lesions can be treated simultaneously.

More patients had FP using T1-SPACE than additional true positives, although we acknowledge that there was a low threshold for noting such findings, knowing that MPRAGE was available to clarify (and indeed this was routinely performed based on the study design). Most were considered unlikely to represent metastases and may have been discounted if only T1-SPACE was available; thus, the FP rate would have been lower if the T1-SPACE images were interpreted without access to MPRAGE. Nevertheless, there is value in performing a confirmatory sequence even when a finding is suspected to be an FP as occasional true positives may be identified. ¹² This is particularly relevant in the context of sulcal hyperintensities, given that incorrectly diagnosing or missing sulcal leptomeningeal metastatic disease may have greater implications for the patient than would incorrect interpretation of a parenchymal metastasis. While our findings support that using 3D-TSE alone may result in a higher FP rate, correlation with 3D-GRE allowed the uncertainties associated with 3D-TSE to be readily resolved. Alternatively, if additional findings are identified when using 3D-TSE alone, 3D-GRE can be performed at a later date depending on the degree of suspicion and potential clinical implications.

A higher FP rate can also be expected during the early stages of utilising 3D-TSE. Subjectively, experience gained through this study suggests that there is a learning curve in evaluating 3D-TSE images, especially for determining whether an apparent finding may instead represent a normal vessel, and it can be expected that the FP rate would decrease over time. Thus, when planning to replace 3D-GRE with 3D-TSE, it may be worthwhile performing both sequences for a period of time, for radiologists to familiarise themselves with the subtle differences and potential pitfalls. This learning curve is likely to be affected by the experience of the reader, and more caution may be warranted (i.e. with a longer period of overlap) when studies are interpreted by general radiologists and/or non-radiologists (e.g. radiation oncologists planning SRS). Of note, Nagao et al. found that performing 3D-TSE in addition to 3D-GRE did not increase interpretation time, ⁶ although of course scan time is longer.

One would anticipate that the likelihood of an FP on T1-SPACE is independent of the pre-test probability of IM. However, the likelihood of a finding on T1-SPACE being a true finding (i.e. IM) rather than an FP is strongly related to the pre-test probability of IM. Specifically, the lower the pre-test probability of IM, the higher the likelihood that a finding on T1-SPACE represents an FP rather than an IM. The likelihood of IM will vary greatly depending on the scenario – from being low in the screening context with prior negative imaging, to high when following up known IM. We suggest that, the lower the pre-test probability of IM, the greater the potential benefit of including both 3D-TSE and 3D-GRE – or, rather, the greater the potential detriment (by incorrectly diagnosing metastases) if only 3D-TSE is performed post-contrast.

The presence of an FP represents an interplay between the technical characteristics of the 3D-TSE sequence and the patient’s individual intracranial (in particular vascular) anatomy, which can be expected to be essentially stable over time. To our knowledge, it is not clear from the literature whether FP findings are relatively consistent over time or can vary substantially across examinations. If relatively consistent, including 3D-GRE in the protocol (in addition to 3D-TSE) may be helpful for the first examination, but its value would decrease thereafter. In addition, even if 3D-TSE is used without 3D-GRE, it may be possible to confidently characterise this as an FP by comparing with a prior examination utilising 3D-GRE.

As an aside, the consensus guidelines recommend fat saturation in order to improve visualisation of osseous metastases, and cite the inability of 3D-GRE techniques to perform fat saturation as an advantage of 3D-TSE. ¹ While we agree with the value of fat saturation, similar results can be achieved utilising MPRAGE with water excitation, ¹³ as we have done in this study, and also use in our routine clinical practice. A further limitation of 3D-TSE is its incompatibility with metallic SRS immobilisation frames. ¹ The importance of this limitation will depend on the utilisation of SRS at the given institution. While an earlier 3D-TSE study can be readily correlated with subsequent in-frame 3D-GRE imaging, performing 3D-TSE and 3D-GRE during the same examination may facilitate subsequent SRS planning. Finally, 3D-TSE is more susceptible to motion artefact than 3D-GRE, ¹ and we did have one FP using 3D-TSE related to this.

Our study performed T1-SPACE after MPRAGE, which may create potential bias as the greater post-contrast delay on T1-SPACE may result in more pronounced enhancement, thus aiding the detection of metastases. ¹⁴ This was done deliberately, however, as a predictor of how the routine MRI protocol could be performed after completion of the study. Our goal was not specifically to determine which of the two sequences had higher sensitivity as this has already been conclusively determined in prior literature; rather, the goal was to assess the logistics of performing T1-SPACE in a broader clinical setting – including a majority of patients without known active intracranial metastatic disease. If both sequences are to be performed in routine practice, it is logical to perform 3D-TSE after 3D-GRE, as the longer post-contrast delay may marginally improve lesion conspicuity on 3D-TSE and this sequence is the most important for lesion detection.

The single-reader nature of the assessment is a limitation, although we consider this to be less significant than in other settings. Firstly, in many cases, the study read was performed after (but initially blinded to) the clinical read, and correlation with the formal report provided some degree of comparison. Indeed, this revealed important discrepancies in two cases, as discussed above. While the reasons for the false negative interpretation in these two cases cannot be determined, we expect this was due to predominant utilisation of MPRAGE as this had been the sole post-contrast T1WI sequence performed prior to the commencement of the study. Secondly, a focus of the study was to develop practical clinical insights rather than solely focussing on detection, as the improved detection using 3D-TSE is well established in the literature. Another potential limitation was our specific focus on patients with metastatic melanoma as opposed to other histologies. The goal of this study was to focus on a patient population with a lower incidence of active IM – this is a relative gap in the existing literature, and our findings have been able to provide insights in this common clinical scenario. Our focus on metastatic melanoma relates to the routine use of MRI at our institution for intracranial screening and surveillance in melanoma patients, while for other histologies MRI is predominantly used in patients with neurological symptoms or abnormalities identified on CT. However, as discussed above, the assessment in melanoma may be more challenging than in other primaries; thus, one would expect that our findings can be comfortably extrapolated to IM more broadly. We note that some of the initial T1-SPACE examinations were suboptimal due to wrap artefact, but reassuringly this did not lead to any false negatives on T1-SPACE compared to MPRAGE. If anything, this may have obscured some additional findings on T1-SPACE, although this does not alter the overall conclusions of our study.

Conclusions

Our findings support the utilisation to 3D-TSE in the broader context of patients undergoing investigation for known or possible IM, not just patients with known IM. The benefits of 3D-TSE, in particularly the greater sensitivity for detecting small IM, outweigh its disadvantages. However, it is important for radiologists to be aware of the limitations of 3D-TSE, most notably the higher FP rate, and we suggest a low threshold for supplementing the examination with 3D-GRE where there is some doubt as to the validity of the findings. This is particularly relevant in patients with a lower pre-test probability of IM and/or when the clinical implications of an incorrect diagnosis are greater. There may also be value routinely performing both 3D-TSE in 3D-GRE in some situations, for example on the first imaging investigation, during the transition from 3D-GRE to 3D-TSE, and in patients with a higher likelihood of subsequently receiving SRS. Using both sequences in combination optimises both sensitivity and specificity.

ORCID iD

Arian Lasocki https://orcid.org/0000-0001-8176-3015