To investigate if a K^trans-based biopsy target selection is superior to the standard selection method in glioma patients.

In the brain, standard biopsy target selection is based on either contrast enhancement in T1W images (CET1W) or on hyperintensities in FLAIR sequences. With these techniques up to 10% of samples are non-diagnostic [1,2] – a high number considering that repetitive surgery is the consequence.

Approaches to select targets with quantitative imaging methods without the standard method as control group have been made [3 - 8]. The proof of an additional benefit would be, however, a strong criterion to implement (semi-)quantitative MRI in the clinical routine.

We therefore investigated whether a transfer constant K^trans “hot spot” based target selection improves the rate of diagnostically accurate glioma samples compared to the current selection standard.

The “hot spot” technique, derived from T1-DCE MRI, was chosen as K^trans is used to approximate both blood-brain barrier disruption (BBBD) and tissue perfusion [9]. These are two factors associated with tumor malignancy. Indeed, WHO grading of glioma by T1-DCE MRI parameters seems to be diagnostically reliable [10,11].

27 pre-operative patients with suspected glioma were enrolled after informed consent. They received 3D CET1W and FLAIR sequences as well as a transverse T1-DCE MRI at 3 T (Achieva TX (Philips Healthcare, Best/NL); 8-channel head coil; table 1).

Contrast was automatically injected at a dose of 0.1mmol/kg BW [Gadobutrol (Bayer Healthcare, Leverkusen/D); 24ml saline flush; flow rate 3ml/s].

The extended Tofts model was applied to derive kinetic parameter K^trans (fig. 1, Intellispace Portal 5.0, Philips Healthcare) [9]. The vascular input function ROI was placed in the superior sagittal sinus. 5/22 cases (WHO III) were excluded secondarily as K^trans was globally zero and no “hot spots” could be identified.

Independently 1 to 6 biopsy targets were marked by (1) a neuroradiologist on K^trans color maps and (2) by a neurosurgeon marked CET1W maps (in non-enhancing tumors on FLAIR maps). Data of both selection methods were exported to the neuronavigation unit (Brainlab, Feldkirchen/D). The selection method of targets was not recognizable on the navigation map. A neurosurgeon blinded to the targeting method retrieved the samples (table 2).

A blinded neuropathologist rated the HE-stained samples as “diagnostic” (matching the reference) or “non-diagnostic” (e.g. underestimating the WHO grade). The reference diagnosis was derived from the later fully resected glioma.

Concordance of samples with the reference diagnosis was statistically analyzed by alternating logistic regression accounting for unbalanced group sizes (SAS9.4, SAS Institute Inc., Cary/USA.)

Diagnostic samples were more likely retrieved from K^trans-based targets (vs. standard method): (1) glioblastoma: 20/39 vs. 11/20, p=0.085; and (2) WHO III/II: n=12/13 vs. n=6/11; p=0.061). 3/16 glioblastoma cases were exclusively correctly diagnosed based on K^trans “hot spot” samples.

Focal BBBD, as defined by K^trans elevation, was found in 4/9 WHO III cases without BBBD according to CET1W maps (no enhancement). Inter-individual variability of K^trans measured in tissue of the same tumor histology was high (1.9* to 511.7*10^-3/min; mean 140.3±135.1*10^-3/min). A differentiation by WHO was possible though (p=0.0003).

On an intra-individual level K^trans was higher in diagnostic samples than in non-diagnostic ones (109.3±113.6*10^-3/min vs. 70.5±88.7*10^-3/min). In a case with mixed WHO III and WHO IV foci, the WHO IV tissue could be discriminated by K^trans but not by CET1W MRI.

There was a strong trend in favor of a K^trans “hot spot” based target selection improving the success rate to identify diagnostic samples compared to CET1W or FLAIR-based selection. K^trans not only seemed more sensitive towards subtle BBBD, but also delivered a focal semi-quantitative discrimination of its degree. As biopsies must retrieve the most malignant parts of the tissue and malignancy itself correlates with BBBD in glioma, the K^trans method may facilitate the identification of suitable tissue targets.

Recurrent glioma often shows a partial de-differentiation, making the K^trans method particularly suitable for these cases. However, 5/27 patients (all WHO III and II) needed to be excluded from the study as K^trans was zero. This limits the application of the K^trans target selection method to glioma of a higher WHO grade (III-IV).

Main limitation of this study is its small sample size. Further, no repetitive T1-DCE MRI measurements could be made to confirm that K^trans “hot spots” were intra-individually reproducible. A subsequent histological analysis of glioma microvasculature and cellular density is planned to elucidate a possible correlation between kinetic parameters and histological tumor features.

The semi-quantitative information of K^trans maps seems to facilitate the identification of representative tissue targets in glioma compared to the purely anatomical standard selection methods of CET1W or FLAIR maps.