Discussion
Accurate diagnosis of large vessel narrowing has clinical implications for stroke risk stratification and the initiation of and escalation of clinical interventions such as transfusions and surgical procedures (e.g. encephaloduroarteriosynangiosis) in pediatric patients with SCA. While there are many imaging techniques to examine the intracranial vasculature, MRA imaging without gadolinium is preferred as first line as it is noninvasive, avoids ionizing radiation and iodinated or gadolinium contrast, compared to Computed Tomography Angiography (CTA), catheter angiography, and contrast-enhanced MRA(18). Non-contrast MRA provides visual assessment of flow-related enhancement within the intracranial arteries as compared to velocity assessment with TCD, however it is uniquely subject to flow-related MRI artifact, which can lead to misidentification of stenosis when it is not truly present, or exaggeration of existing stenosis.
One disadvantage of noncontrast MRA is the potential loss of flow-related signal in regions of arterial tortuosity and/or branching, which are commonly present in patients with SCA, and the possibility for this artifact to simulate arterial stenosis. The possibility of artifactual stenosis is especially problematic in SCA, as the regions typically affected by artifact on noncontrast MRA overlap with the regions typically affected by sickle-cell related vasculopathy. Use of short-TE MRA techniques (ideally less than 5msec), can limit the artifactual loss of flow-related signal and thereby more accurately assess the true patency of intracerebral arterial vessels.
Of the 29 patients initially identified with sickle cell CNS vasculopathy, our study identified 12 patients (41%) in whom the level of cerebral stenosis was re-classified to a lower degree after implementation of the standardized MRA protocol. Clinically, an improvement in the level of arterial stenosis is not expected in patients with SCA, even for those on disease modifying therapy such as chronic transfusion therapy. This suggests that previously identified areas of stenosis were a result of artifact rather than true vasculopathy.
In six patients (21%) this reclassification led to a de-escalation of the stroke prevention plan and discontinuation of chronic transfusion therapy. In these six patients, the initially identified stenosis was noted to be most often in A1 branch of the anterior cerebral artery or supraclinoid segments of the internal carotid artery. Additionally, one patient was newly identified to have stenosis and one patient was determined to have an increased classification of stenosis, suggesting a progression of stenosis due to SCA.
As a retrospective chart review, this study has limitations in assessing whether the TE of the scanning protocol is responsible for leading to the decreased appearance of artifact and thus clinical change in treatment plan. After May 2016, MRA in three patients had TE >5msec, suggesting that the standardized scanning protocol was not uniformly implemented at our institution in May of 2016. For the six patients in which the degree of CNS stenosis was re-classified and treatment was de-escalated, two patients had TE > 5msec on their post May 2016 imaging. While TE was the major parameter standardized in this protocol, MRA time of flight angiography phase dispersions may also be dependent on other technical parameters. For example, a greater slice thickness or TR will make the contrast between stationary tissue and blood less obvious. To evaluate this, TR, matrix size, and slice thickness were reviewed for several patients and were not noted to be significantly different.
Our analysis included 29 patients; however, our initial chart sample identified 136 patients with only one MRA; reanalysis of all MRA exams done prior to May 2016 or with a high TE may identify other patients in which the stenosis can be re-classified.
Aside from the presence of MRI artifacts, interpretation of MRA imaging may be partially dependent on the interpreting neuroradiologist. To minimize this variability, a blinded expert neuroradiologist reviewed the pre-May 2016 images for the six patients that had a change in treatment plan. This reader agreed with previous MRA reports regarding the appearance of stenosis in five of the six patients, however for one patient this reader did not identify any stenosis on the pre-May 2016 imaging. This raises the question regarding variability of expert neuroradiologist interpretation of stenosis versus artifact in scans with higher TE and implications on clinical management.
To elucidate these results further, next steps could include a focused comparative analysis of TCD and MRA findings on a larger sample size, which may require a multicenter project to identify patients with CNS arterial stenosis identified using a higher TE and prospectively re-image them with a lower TE. Further analysis could also include other factors that might decrease the degree of artifact (slice thickness, matrix, contrast mechanisms, etc.), comparison to other vessel imaging techniques, long term follow up of patients identified in this study, and addressing the impact of MRI parameters in patients with other vascular conditions.