PO31: FRACTAL ANALYSIS OF PREOPERATIVE IMAGING AS A TOOL FOR PREDICTING PROGNOSIS IN GLIOBLASTOMA MULTIFORMAE – PRELIMINARY RESULTS

INTRODUCTION: The fractal analysis objectively quantifies the geometric complexity of structures in unitless measure index. We hypothesise it may reflect the degree of malignancy in glioblastoma multiformae (GBM). METHOD: A group of 33 patients with postoperative diagnosis of GBM were retrospectively enrolled into the study. Preoperative brain MR (T1-weighted post-gadolinium) were processed and analyzed to determine the average and maximum fractal dimension (F̄D and FDmax) of the inner and outer surface of the contrast enhancing part of the tumour. Box-count method and ImageJ 1.49 software were used. Corellations between fractal dimensions and: 1) demographics; 2) immunohistochemical characteristics; 3) progression free (PFS) and overal survival (OS); 4) tumour volume were determined by mean of nonparametric tests. RESULTS: There were 8 women and 25 men in the study group, mean age was 67 ± 10 years. Fractal analysis was possible to be perfomed in all cases with good inter-observer reproducibility (Kappa >0.80). The median FD̄ was 1.2550 (IQR 0.0897) and FDmax 1.3848 (IQR 0.1057). There was a statisticaly significant Spearman Rank correlation between fractal dimensions and Ki-67 index (-0.44 and -0.52 for FD̄ and FDmax respectively). Both FD̄ and FDmax were significantly (Fi2; p < 0.05) correlated with: age, tumor volume, PFS, OS, and Phosphohistone-H3 index but not with the expression of neurofilaments, p53, GFAP, EGFR and IDH-1. CONCLUSION: Fractal analysis of the preoperative MR images may pottentially act as an adjunct in determing the degree of malignancy and prognosis in GBM. Further studies on a wider cohort of patients are warranted.