Dose-Response analysis for a patient cohort treated with $^{90}Y$-TARE according to a personalized dosimetric workflow: Preliminary results.

The aim of the work is to investigate the role of both dosimetric and clinical parameters as classifiers or predictors of response to the $^{90}Y$ radioembolization treatment and to present possible response cut-off. Dosimetric data were extracted from personalized absorbed dose maps obtained with convolution of Voxel S-values kernel with the patient actual $^{90}Y$ activity distribution, derived from PET images. Beside common parameters as Mean Absorbed Dose to the Tumor (MADt) and to the Healthy Liver, other variables were also selected from Dose Volume Histograms, such as D30,t, D95,t and V120,t, along with clinical data. ${D95} \geq$ 103.50 ${Gy}$ and ${MADt} \geq$ 228.75 ${Gy}$ were found to be optimal cut-off values for complete response, while ${D30} \geq$ 180.30 ${Gy}$ and ${MADt} \geq$ 117.30 ${Gy}$ were selected as cut-off values for at least partial response and predicted a better survival. CT-based contouring resulted to be a more accurate delineation approach than the PET-based one: MADt,PET showed neither correlation nor predictive power with response or survival. Finally, the results also indirectly demonstrate the reliability of the dosimetric workflow developed in this Institute, based on a patient specific approach.