Pre-therapeutic 68Ga-PSMA PET AI Based Dose Prediction for 177Lu-PSMA Targeted Radionuclide Therapy

Despite the early success of TRT, concerns have been raised about the risks of inadequate trade-off between therapeutic dose and side effects. Currently, the protocols for administering the radiopharmaceuticals are assessed on a population basis, and the activity to administer was determined for a specific patient group based on preceding studies . However, the European Council Directive (2013/59 Euratom) mandates that TRT treatments should be planned according to the optimal radiation dose tailored for individual patients, as has long been the case for external beam radiotherapy (EBRT) or brachytherapy. An essential requirement of TRT treatment planning is to estimate the absorbed dose in advance of therapy.

Prior knowledge of the biodistribution of the therapeutic agent via the pre-therapy imaging assists to optimize the trade-off between tumor destruction and irradiation of healthy tissues. Concepts, such as physiologically based pharmacokinetic (PBPK) modeling, have been proposed to estimate the spatiotemporal pharmacokinetics of imaging agents and then extrapolate to the treatment agents.

An alternative is to generate a 3D time integrated activity (TIA) map per patient based on the PBPK and the pre-therapy imaging. The TIA gives the information about number of decays that take place in each voxel during the total duration of the therapy. PBPK is an organ-based model, then the calculation of the 3D TIA raises the issue of organ segmentations on the pre-therapy nuclear imaging, which must be robust, automatic, and accurate. The absorbed dose to the patient can be estimated before the treatment using the 3D TIA and the patient anatomy (CT image) using Monte Carlo (MC) simulation. . This project will address two main challenges: (a) the robust and accurate metabolic segmentation in nuclear medicine for the 3D TIA calculation, and (b) the fast dose prediction based on MC and deep-learning approach.