Prediction of chemotherapeutic response with volumetric parameters or texture features on FDG PET in recurrent gynecological malignancies

Objectives: The purpose of this study was to evaluate if volumetric parameters (VPs) or texture features of FDG PET were useful in predicting treatment response early after chemotherapy in recurrent gynecological malignancies. Methods: This study included 25 patients with recurrent gynecological malignancies (11 uterine, 11 ovarian, and 3 others). FDG PET/CT exam was performed before (pre) and after 1 cycle of chemotherapy (post1c). Metabolic tumor volume (MTV, SUV threshold 2.5) and total lesion glycolysis (TLG) were obtained as VPs in addition to SUVmax at the hottest lesion in each exam. Texture features were obtained from a 3D ROI covering the same hottest lesion in pre exam (preTF). TFs evaluated in this study included entropy, homogeneity, low-, high gray-level zone emphasis, short-, and long-run emphasis as recommended by Orlhac, et al (JNM 2014). Pre, post1c, and changes (expressed as Δ) of VPs as well as SUVmax and preTF were compared each other and to the treatment response after last cycle of chemotherapy, which was decided with all clinical information available including imaging data. Pre- VPs, SUVmax, and TF were also compared to the post1c treatment response. Treatment response was defined as more than 30% reduction of SUVmax as compared to pre values.

Results: PreMTV ranged from 5.8 to 146ml and preSUVmax from 3.4 to 22.4. Entropy, high gray-level zone emphasis, short-, and long-run emphasis correlated with pre- VPs or SUVmax. Post1c response was observed in 10 of 25 pts. The values of entropy in the post1c responder were significantly lower than those in the non-responder (p<0.05) (homogenous lesions showed good response). No difference was noted in other pre- TFs, VPs, or SUVmax between the post1c responder and non-responder. Entropy was considered the parameter to predict post1c response from ROC analysis (AUC: 0.7). After last cycle of chemotherapy, 11 of 25 pts represented response. The values of post1c VPs were significantly lower and values of ΔVPs and ΔSUVmax were significantly higher in responder than those in the non-responder (p<0.05). Among them, ROC analysis revealed post1c MTV the best parameter to predict response (AUC: 0.76). No significant findings were observed between pre- VPs, SUVmax, TF, or post1c SUVmax and treatment response. Conclusion: This study demonstrated that texture analysis on FDG PET allowed prediction of response after 1 cycle of chemotherapy in pts with recurrent gynecological malignancies. However, volumetric parameters were more useful than TFs in predicting final treatment response.