Prediction of outcome in sarcoma patients based on spatial heterogeneity of F18-FDG uptake using a multiscale variance technique

Ziel/Aim:

Spatial heterogeneity of tracer uptake in the tumor is a promising predictor of patient outcome in various tumor entities. Different definitions have been proposed for spatial heterogeneity of tracer uptake. However, as heterogeneity is a very complex characteristic, the most appropriate mathematical approach for quantitative description of heterogeneity has still to be determined. Here we propose the use of a multiscale variance technique (MSVT) that provides a vector of heterogeneity values at different spatial scales (distances). The technique is widely used for the characterization of geographical data (maps). The prognostic value of spatial heterogeneity based on MSVT was evaluated in sarcoma patients with F18-FDG-PET for initial staging.

Methodik/Methods:

In total 51 sarcoma patients (29 m, median age 15 y, range 2-61 y) were included retrospectively. Histological subtypes were 19 Ewing sarcomas (EWS), 16 osteosarcomas (OS) and 16 sarcomas with other subtypes (MS). The primary tumor was segmented fully automatically using the ROVER 3D segmentation tool which is based on thresholding at the volume-reproducible intensity threshold after subtraction of the local background. For the computation of heterogeneity, SUV values were first binarized at a given threshold. Then MSVT was used to compute heterogeneity at the following spatial scales: 8mm, 16mm, 32mm. For each scale, heterogeneity was plotted as function of the binarization threshold. The area under the curve (AUC) from SUVmin to SUVmax was computed. The AUC ratio of heterogeneity (HR) at 8mm to 16mm was used for the univariate analysis presented here. The best threshold for discrimination between survivors and non-survivors was determined by ROC analysis. Overall survival rates between the resulting groups were compared by Kaplan-Meier curves and log-rank tests.

Ergebnisse/Results:

36 patients (15 EWS, 14 OS, 7 MS) survived till the end of the follow-up ranging from 9.1 to 97.8 months (median 46.5 months). Overall survival of non-survivors ranged from 2.7 to 49.6 months (median 16.3 months). The highest HR was observed in the group of miscellaneous sarcomas (median 44.6, IQR 38.5-53.6) followed by osteosarcomas (median 43.6, IQR 35.3-46.0) and Ewing sarcomas (median 34.4, IQR, 31.2-39.6) (Kruskal-Wallis p<0.05). HR was significantly higher in non-survivors (median, 39.6; IQR, 14.6) than in survivors (median 36.3, IQR 31.5-44.8, Mann-Whitney p<0.05) with a threshold of 38.3 in ROC analysis (p<0.05). Kaplan-Meier-analysis revealed a higher overall 5-year-survival rate of 85.6% in patients with HR≤38.3 compared to 37.7% in patients with HR>38.3 (log-rank-test p<0.01). Significantly different distribution of survival rates was also observed in subgroups of EWS (p<0.05) and OS (p<0.05), but not in MS (p=0.23).

Schlussfolgerungen/Conclusions:

MSVT based spatial heterogeneity is promising for outcome prediction based on F18-FDG-PET in sarcoma. Studies with enlarged patient numbers using multivariate statistical models with adjustment of confounders are necessary for further evaluation of the potential of MSVT.