LuCaS2: Efficient Monte Carlo simulations of serial PET scans for assessing detection and quantification methods used in patient monitoring

Objective: PET imaging is a promising approach for the early assessment of tumor response to therapy. However, there are currently no widely accepted rules to interpret the changes seen in successive PET scans. An objective evaluation of the performance of methods dedicated to detecting and characterizing the changes occurring between successive PET scans is needed. In this work, we propose a method to perform realistic Monte Carlo simulations of patient scans in the context of therapy monitoring. Material and Methods: Serial tumor-free sinograms of a patient with no tumor in the regions of interest (here the lungs) are first simulated with the GATE simulation toolkit, based on real PET/CT acquisitions. In addition, tumor-only sinograms are simulated, based on real tumor contours and tumor changes seen in patients. A specific strategy is used to properly relocate tumors from scan to scan. After corrections, the tumor-free and tumor-only sinograms are combined and reconstructed to produce serial scans including realistic tumor changes. We illustrate the relevance of such data for assessing the performance of the EORTC criteria to detect SUV and volume changes in tumors. Results: Using our simulation approach, realistic PET images can be easily produced in reasonable computational times. The results obtained from the simulated images illustrate the limits of the EORTC criteria to assess tumor changes. Conclusion: Monte Carlo simulations can be used to simulate realistic serial PET scans appropriate for assessing the performance of methods dedicated to the detection and quantification of tumor changes.