Apparent Diffusion Coefficient Metrics to Differentiate between Treatment-Related Abnormalities and Tumor Progression in Post-Treatment Glioblastoma Patients: A Retrospective Study

Abstract

Distinguishing treatment-related abnormalities (TRA) from tumor progression (TP) in glioblastoma patients is a diagnostic imaging challenge due to the identical morphology of conventional MR imaging sequences. Diffusion-weighted imaging (DWI) and its derived images of the apparent diffusion coefficient (ADC) have been suggested as diagnostic tools for this problem. The aim of this study is to determine the diagnostic accuracy of different cut-off values of the ADC to differentiate between TP and TRA. In total, 76 post-treatment glioblastoma patients with new contrast-enhancing lesions were selected. Lesions were segmented using a T1-weighted, contrast-enhanced scan. The mean ADC values of the segmentations were compared between TRA and TP groups. Diagnostic accuracy was compared by use of the area under the curve (AUC) and the derived sensitivity and specificity values from cutoff points. Although ADC values in TP (mean = 1.32 × 10^-3 mm²/s; SD = 0.31 × 10^-3 mm²/s) were significantly different compared to TRA (mean = 1.53 × 10^-3 mm²/s; SD = 0.28 × 10^-3 mm²/s) (p = 0.003), considerable overlap in their distributions exists. The AUC of ADC values to distinguish TP from TRA was 0.71, with a sensitivity and specificity of 65% and 70%, respectively, at an ADC value of 1.47 × 10^-3 mm²/s. These findings therefore indicate that ADC maps should not be used in discerning between TP and TRA at a certain timepoint without information on temporal evolution.

Keywords: apparent diffusion coefficient; diffusion imaging; glioblastoma; receiver operating characteristic curve; treatment-related abnormality; tumor progression.

Figure 1

Axial T1-CE sequences and ADC maps of confirmed TP and TRA patients. (A) Contrast-enhanced T1-weighted images with segmenation; (B) ADC map with segmentation overlay The mean ADC values of the lesions in this figure are TP = 1.14 × 10^–3 mm²/s, TRA = 1.68 × 10^–3 mm²/s.

Figure 2

ROC analysis based on the ADC values found and literature cut-off. An AUC of 0.71 (95% confidence interval: 0.58–0.83) was yielded with regard to distinguishing TP and TRA on ADC values only. The AUC for combined MGMT status and mean ADC led to an AUC of 0.78 (95% CI: 0.66–0.90). The literature-derived cut-off ADC value of 1.33 × 10^–3 mm²/s (13) yielded an AUROC of 0.635 (95% confidence interval: 0.505–0.764).