[F-18]FMCH intratumor heterogeneity analysis by advanced texture analysis in patients with prostate cancer recurrence

Background: The role of [18F]FMCH in patients with biochem- ical recurrence for prostate cancer (PCa) is well established. An alternative to visual image analysis, which represent the current standard for PET/CT assessment, is the evaluation using textur- al features with the scope of a more objective and quantitative evaluation. In this work we evaluated the feasibility of radiomics analysis of [18F]FMCH PET/CT in patients with recurrent PCa after primary radical therapy. Methods: Within a prospective trial, we evaluated a series of 82 patients (mean age 72±7 years, medi- an age 71 years, range 56-83) with recurrent PCa (mean PSA at the time of [18F]FMCH PET/CT 10,39 ng/ml). Histology data and current treatment were recorded in all patients. Whole-body PET/CT (GE Discovery ST) was acquired about 45 mins after [18F] FMCH (4 MBq/kg of body weight) administration. Based on [18F] FMCH PET/CT results patients were classi ed in on Oligometa- static (< 3 lesions) or Multimetastatic (> 3 lesions) disease. A to- tal of 339 lesions were found and classi ed according to TNM in skeleton (n=180), distant Ln (n=78), regional Ln (n=55), prostat- ic lodge (n=12), other (n=34). Radiomics texture features were extracted using the LIFEx package from semiautomatically seg- mented PET and CT images. Statistical analysis was performed with JMP Statistical Discoverytm. The Spearman correlation coef- cient two-sided Wilcoxon rank-sum test or Kruskal-Wallis test were used and corrected for multiple testing using the Bonfer- roni-Holm method (signi cance level α = 0.05). Results: Radio- mic analysis allowed to discriminate between patients with oli- gometastatic and multimetastatic recurrence as well as among among PCa subtypes (Gleason score) thus characterizing a sig- nature of a more aggressive PCa. In addition, texture parameters showed ability in distinguishing among site of recurrence being di erent in bone lesions recurrence, distant Ln recurrence and local Ln recurrence. Hormone-therapy impact on textural fea- tures was assessed to identify texture features characterizing lesions independently from ongoing therapy. Conclusions: Our preliminary results after these rst series of patients showed that texture analysis of [18F]FMCH PET/CT is able to characterize the heterogeneity of lesions in patients with recurrent PCa, more ef- ciently than the semi-quantitative indexes. Radiomic analysis proved its ability in investigating lesions heterogenity, providing imaging biomarker to be add to the current available to better characterize disease PCa burden and biological aggressiveness, thus guiding patients’ treatment decision making. This work has been founded by Bando AIRC 2017 IG 20819