Can Spectroscopic Magnetic Resonance Imaging be Used to Delineate Recurrent Glioblastoma?

Abstract

<h3>Purpose/Objective(s)</h3> Local glioblastoma cell volumes strongly correlate with whole brain spectroscopic MRI (sMRI) generated relative choline to N-Acetyl-Aspartate ratio (rChoNAA) maps at initial presentation. Significantly more disease is identified when rChoNAA maps, reflective of pathologically elevated metabolism and tumor replacement of normal brain, are combined with T1 post-contrast (T1PC) MRI. Despite primary radiotherapy (RT), glioblastoma almost always recurs. However, post-RT effects confound conventional MRI in recurrent glioblastoma (rGBM) where radiation necrosis appears like rGBM on T1PC, while benign edema and RT-induced leukoencephalopathy enlarge T2/FLAIR volumes. The uncertainty about tumor volume and fear of repeat radiotherapy (re-RT) toxicity in rGBM lead to re-RT volumes that typically only use T1PC and likely underestimate gross disease. We hypothesize that rChoNAA maps may uncover a larger volume of high-risk rGBM for enhanced re-RT targeting and dose escalation. <h3>Materials/Methods</h3> Retrospective analysis of an IRB-approved prospective cohort was performed on 14 rGBM patients that underwent sMRI as part of planning for re-RT. sMRI at 3T was acquired over the whole brain at 5.6 × 5.6 × 5 mm resolution in 15 minutes. Target volumes were delineated using rChoNAA ratio over 2 (rChoNAA>2) on sMRI similar to a prior prospective trial in upfront glioblastoma. rChoNAA maps and rChoNAA>2 contours were made available to the treating radiation oncologist through export to standard contouring software, and in all treated cases rChoNAA>2 volumes were included as part of the gross tumor volume. Volumes of rChoNAA>2 were compared with conventional multiparametric sequences on a routine clinical 3T MRI neuroimaging protocol including T1PC, T2/FLAIR, diffusion-restricted regions on apparent diffusion coefficient (ADC) maps, and perfusion relative cerebral blood volume (rCBV). <h3>Results</h3> rChoNAA>2 volumes (mean 29 cc, range 7-79 cc) were significantly different from all other imaging modalities using paired t-tests (p<0.01). T1PC volumes had a mean of 11 cc (range 1-31 cc). Compared to T1PC on a per patient basis, rChoNAA>2 volumes were 262% larger (range 23% smaller – 873% larger). FLAIR hyperintense volumes (mean 112 cc, range 19-233 cc) were much larger than all other modalities, reflecting likely benign and post-RT changes. Volumes of increased perfusion relative to surrounding brain were often small (mean 6.7 cc, range 0.2-19.1 cc) and volumes of restricted diffusion were often not identified and generally tiny (mean 0.8 cc, range 0-3.7 cc). <h3>Conclusion</h3> Multiparametric MRI techniques such as T1PC, FLAIR, ADC, and rCBV are confounded by post-RT effects in rGBM and are of questionable utility in re-RT planning. Similar to upfront treatment of GBM, sMRI suggests high-risk targets in rGBM undetected by conventional MRI. Pattern of failure analyses and radiologic-pathologic correlation studies are underway to confirm the utility of sMRI in rGBM.