PREDICTING SITES OF LOCAL TUMOUR PROGRESSION-WHAT SHOULD BE OUR IMAGING BIOMARKER?

Abstract Background Glioblastoma is the most aggressive primary brain tumour diagnosed in adults. Despite intensive treatment of maximal safe resection and chemoradiotherapy, the prognosis remains grim due to invasive tumour cells. Current treatment and standard imaging methods are highly limited in terms of managing these invasive cells as they are often located outside the area of surgical resection and are generally resistant to chemoradiation. DTI appears to be a promising tool for imaging tumour cell invasion and predicting the site of recurrence especially when decomposed into its anisotropic (q) and isotropic (p) components. The aim of this study is to investigate the sensitivity of imaging biomarkers as predictors of recurrence. Material and Methods All pre-op and recurrence sequences were co-registered to the pre-op post-contrast T1-weighted images as reference. Co-registration of images was performed using FSL and ANTs. The ROIs for 49 patients with a primary diagnosis of GBM were segmented using 3DSlicer. Each voxel was assigned to one of four status: true negative, false negative, true positive and false positive. Sensitivity and specificity between the pre-op ROIs and the progression region were calculated using FSL. The significance of the differences in sensitivity and specificity between the ROIs was computed in MATLAB. Results The sensitivity for the contrast enhancing region was 48.77 ± 26.13 (Mean ± SD) and 62.40 ± 23.07 (Mean ± SD). The abnormal q alone has a significantly greater sensitivity than the contrast enhancing region (t = -2.7327, df = 96, p-value = 0.0075). The sensitivity for the ROI of combined contrast enhancement and abnormal q was 65.86 ± 23.29 (Mean ± SD). There is an even more significant increase in sensitivity when the contrast enhancing ROI is combined with abnormal q region (t = -3.4133, df = 96, p-value = 9.4123e-04) compared to when it is alone. There was no statistical difference in the specificities of the different ROIs. Conclusion Current surgical and radiation volumes focus solely on pre-op contrast enhancement. However, these results suggest that combining the abnormal q with the standard contrast enhancing region is a more sensitive predictor of tumour recurrence than contrast enhancement alone, while still retaining high specificity. The higher sensitivity is an indicator of correct identification of tumour recurrence while the high specificity correctly identifies normal brain, or non-recurrent regions. These results are currently being prospectively assessed in a multi-centre study (PRaM-GBM).