Understanding gender pattern differences in MET-PET Glioma patients with radiomics analysis

Ziel/Aim PET uptake pattern gender differences are under debate. However, gender specific PET uptake patterns especially in light of Glioma histological subgroups has not been investigated in detail. Our goal was to utilize radiomics analysis in order to understand gender differences in Glioma histological subgroups.

Methodik/Methods 56 of 105 MET-PET positive cases were involved in this study. Each case underwent semi-automated tumour delineation (Hybrid 3D) and tumour-to-background ratio (TBR) normalization by a background VOI in the contra-lateral region. Extraction of 50 radiomic features from each TBR VOI was performed [1]. Male-female cases with similar characteristics (histology, WHO stage, age, IDH mutation status) were matched and categorized to Astro (n=7-7), Oligo (n=8-8), OligoAstro (n=7-7) and GBM (n=6-6) subgroups according to the WHO 2007 classification system. Detection of prominent radiomic features differentiating gender was done by ensemble learning [2] in all the four pairwise subgroups.

Ergebnisse/Results The first three prominent features to differntiate genders in the 4 subgroups were textural heterogeneity based (NGTDM, GLZSM, GLCM and Histogram categories), with the OligoAstro subgroup being an exception, where the most prominent feature was shape-based (Spherical dice coefficient), followed by GLCM sum entropy. OligoAstro and GBM subgroups presented the most skewed feature weight distributions implying that these subgroups are the most prone to gender uptake pattern differences.

Schlussfolgerungen/Conclusions We presented a radiomic analysis approach to understand imaging pattern differences within genders in glioma patients. Our results may support why OligoAstro presented a challenge in the interpretation according to the WHO 2007 system. Even though the WHO 2016 system eliminates the term OligoAstro, our results imply that future classification systems shall also consider a gender specificity.

• Literatur/References

• 1 Papp L. , et al: Optimized feature extraction for radiomics analysis of 18F-FDG-PET imaging. .JNM 2018 , 10.2967/jnumed.118.217612.
• 2 Papp L. et al: Glioma Survival Prediction with Combined Analysis of In Vivo ¹¹C-MET PET Features, Ex Vivo Features, and Patient Features by Supervised Machine Learning. . JNM , 2018; ( (59) ): 892-899 .