Identification of a candidate biomarker from perfusion MRI to anticipate glioblastoma progression after chemoradiation.

Abstract

To identify relevant relative cerebral blood volume biomarkers from T2* dynamic-susceptibility contrast magnetic resonance imaging to anticipate glioblastoma progression after chemoradiation. Twenty-five patients from a prospective study with glioblastoma, primarily treated by chemoradiation, were included. According to the last follow-up MRI confirmed status, patients were divided into: relapse group (n = 13) and control group (n = 12). The time of last MR acquisition was t_end; MR acquisitions performed at t_end-2M, t_end-4M and t_end-6M (respectively 2, 4 and 6 months before t_end) were analyzed to extract relevant variations among eleven perfusion biomarkers (B). These variations were assessed through R(B), as the absolute value of the ratio between ∆B from t_end-4M to t_end-2M and ∆B from t_end-6M to t_end-4M. The optimal cut-off for R(B) was determined using receiver-operating-characteristic curve analysis. The fraction of hypoperfused tumor volume (F_hP_g) was a relevant biomarker. A ratio R(F_hP_g) ≥ 0.61 would have been able to anticipate relapse at the next follow-up with a sensitivity/specificity/accuracy of 92.3 %/63.6 %/79.2 %. High R(F_hPg) (≥0.61) was associated with more relapse at t_end compared to low R(F_hPg) (75 % vs 12.5 %, p = 0.008). Iterative analysis of F_hP_g from consecutive examinations could provide surrogate markers to predict progression at the next follow-up. • Related rCBV biomarkers from DSC were assessed to anticipate GBM progression. • Biomarkers were assessed through their patterns of variation during the follow-up. • The fraction of hypoperfused tumour volume (F_hP _g ) seemed to be a relevant biomarker. • An innovative ratio R(F_hP _g ) could be an early surrogate marker of relapse. • A significant time gain could be achieved in the management of GBM patients.