Histogram analysis reveals a better delineation of tumor volume from background in 18F-FET PET compared to CBV maps in a hybrid PET–MR studie in gliomas

Abstract

Anatomical imaging with magnetic resonance imaging (MRI) is currently the method of first choice for diagnostic investigation of glial tumors. However, different MR sequences may over- or underestimate tumor size and thus it may not be possible to delineate tumor from adjacent brain. In order to compensate this confinement additonal MR sequences like perfusion weighted MRI (PWI) with regional cerebral blood volume (rCBV) or positron emission tomography (PET) with aminoacids are used to gain further information. Recent studies suggest that both of theses image modalities provide similar diagnostic information. For comparison tumor to brain ratios (TBR) with mean and maximum values are frequently used but results from different studies can often not be checked against each other. Furthermore, especially the maximum TBR in rCBV is at risk to be falsified by artifacts (e.g. blood vessels). These confinements are reduced by the use of histograms since all information of the VOIs are equally displayed. In this study we measured and compared the intersection of tumor and reference tissue histograms in 18F-FET PET and rCBV maps in glioma patients. MethodsTwenty-seven glioma patients with contrast enhancing lesion on T1-weighted MR images were investigated using static 18F-FET PET and rCBV in MRI using a PET–MR hybrid scanner. In all patients diagnosis was confirmed histologically (7 grade II gliomas, 6 grade III gliomas and 14 grade IV gliomas). We generated a set of tumor and reference tissue Volumes-of-Interest (VOIs) based on T1 weighted images in MRI with the tumor VOI defined by contrast enhancement and transferred these VOIs to the corresponding 18F-FET PET scans and rCBV maps. From these VOIs we generated tumor and reference tissue histograms with a unity of one for each curve integral and measured the proportion of the area under the tumor curve that falls into the reference curve for 18F-FET PET and rCBV maps for each patient. ResultsThe mean proportion of the area under the tumor curve that falls into the reference curve was smaller in FET PET than in rCBV maps (13.1±19.5% vs. 45.0±21.3%; p<0.001). ConclusionsMeasuring the intersection of tumor and reference tissue histograms in 18F-FET PET and rCBV maps based on identical VOIs defined in anatomical images with tumor marked by contrast enhancement in T1-weighted images can easily be performed and indicates that tumor volume can better be discriminated from background in 18F-FET PET compared to rCBV in MRI. With this approach artefacts (e.g. blood vessels) impair the results to a lesser degree compared to the mean and especially maximal TBRs which are often used when comparing these two image modalities.