Quantification of intra-tumor 18F-FDG PET tracer uptake heterogeneity through textural features analysis: Which minimum functional tumor volume to consider?

International audienceIntra-tumor uptake heterogeneity in PET has been associated with patient outcome in several cancer models. Amongst the proposed methods for its quantification, textural features (TF) analysis provides several local and regional parameters. One of the issues associated with the use of TF analysis is its applicability may depend on a minimum metabolically active tumor volume (MATV), that has been claimed to be 45 cm3, based on entropy (1). The objective of this work was to verify this assertion for the entire range of TF.Several patients' cohorts in different cancer models (head&neck, lung, esophageal, and cervix) were pooled to create a database of 308 tumors with FDG PET images. A subset of 6 local and 5 regional TF were investigated according to their reproducibility and robustness assessed in previous studies. The correlation with MATV was investigated with spearman rank coefficients (rho) and a threshold of rho=0.80 was considered to differentiate features with potential complementary information with respect to MATV based on previous studies.Range of MATV was 3-415 cm3 (mean=35, median=21). Correlation between MATV and TF (table 1) ranged from 0.31 to 0.98. No significant differences were found in different cancer models. Even ≤10 cm3, 4 local and 3 regional features provided complementary information with respect to volume (rho≤0.80). Correlations decreased with increasing MATV and most features fell below the threshold (even entropy) except intensity variability and size-zone variability.Our results suggest that heterogeneity quantification through TF may potentially provide complementary information even for volumes below 45 cm3, although it is reduced with smaller volumes