At present, traditional dose verification in radiotherapy plan mainly uses homogeneous phantoms instead of patient for measurement verification. This method cannot evaluate the dose distribution in the human body. The development of artificial intelligence (AI) and Monte Carlo (MC) has brought opportunities in the dose verification.

For the two types of equipment of conventional electronic linear accelerator and nuclear magnetic guided accelerator, first of all, the source model of the electronic linear accelerator is established, and the commission of the measured machine measurement data is completed. Before the implementation of radiotherapy, the GPU-based MC calculation is used to generate the three-dimensional dose distribution for dose verification. Then in each fraction of radiotherapy, the MRI or CBCT images of the patient were obtained to generate the synthetic CT based on the AI, and the contours on the planned CT is propagated to the synthetic CT. The GPU-based MC calculation is used again to calculate the three-dimensional dose distribution on the pseudo CT according to the accelerator logfiles. Finally, the dose verification and tracking for patient are realized in the whole radiotherapy process.

Compared with Monaco Treatment Planning System (TPS) for 34 clinical cases of MR-guided radiotherapy, the γ-pass rates of ArcherQA and Monaco were greater than 90% (including 31 cases more than 95%) with the criteria of 3 mm/3% . Compared with Truebeam TPS for 40 clinical cases, compared with Eclipse, theγ-pass rates of ArcherQA and Truebeam were all greater than 95%. The average time of MC dose calculation for conventional radiotherapy plan and MR-guided radiotherapy plan were 34s and 80s, respectively. Compared with the traditional B-spline elastic registration, the accuracy and speed of the AI-based elastic registration have been greatly improved.

AI and MC can greatly improve the efficiency and accuracy of dose verification in radiotherapy plan, which are expected to be widely used in clinical radiotherapy in the future.