Diffusion Model-based Lu-177 Dose Rate Mapping From Single-Time-Point SPECT/CT Images

Accurate voxel-level dosimetry is crucial for assessing the effectiveness of Lu-177 PSMA therapy for prostate cancer. Current dosimetry practices based on the MIRD schema rely heavily on approximations, limiting their personalization and accuracy. Monte Carlo simulations provide high accuracy but are computationally expensive. This paper proposes a diffusion probabilistic model to predict dose rate maps from single-time-point SPECT/CT imaging. Our DPM operates by simulating a diffusion process of dose rate images and learning their distribution, resulting in accurate dose rate predictions. We compare our approach with a U-Net architecture, demonstrating the efficacy of diffusion models in improving dose estimation accuracy while handling grainy Monte Carlo simulations as targets. The proposed method offers a promising approach for achieving personalized and efficient dosimetry in Lu-177 PSMA therapy.