Background and Aims: The ability of advanced radiotherapy techniques to produce an optimal plan varies according to the planner's experience. Additionally, developing the treatment plans may be a complex, time-consuming, and iterative process characterized by trial-and-error practices.
Artificial Intelligence (AI) can assist in radiotherapy and cancer treatment by personalizing treatment plans and optimizing radiation doses for individual patients, improving treatment outcomes, and minimizing side effects. In this paper, we aim to provide an overview of artificial intelligence models in dose prediction and automated radiotherapy treatment planning. Methods: We conducted a systematic search and collected scholarly articles published from ۲۰۱۸ through ۲۰۲۴, utilizing keywords “artificial intelligence,” “radiotherapy,” “automated planning,” and “dose prediction.” Articles were included if they addressed any aspect of artificial intelligence in relation to dose prediction and radiotherapy treatment planning in different tumor sites. Results: Most of the studies on dose prediction and treatment planning, focused on breast cancer, prostate cancer, head and neck cancer, and cervical cancer. The previous model architectures include those based on Knowledge Based Planning (KBP), Convolutional Neural Networks (CNN), UNet, Generative Adversarial Networks (GANs), ResNet, and DenseNet. Developed methods employ the raw data (anatomical structures and/or CT images) as input and use the architectures to uncover the latent features concealed within the original raw data. Furthermore, an accurate dataset would require a predetermined number of OARs and/or PTVs as input data and are commonly customized to a particular treatment. On the other hand, the lack of sufficient and large datasets in the field of medical sciences has led to the limitation of the use and generalizability of previous models, which are also limited to the tumor site and the center under treatment. Conclusion: Most previous studies have been done retrospectively and include the limitations of this type of study, including differences among treatment planners, treatment machines, radiation oncologists, and dataset size. Despite the current limitations and challenges, artificial intelligence models for radiotherapy dose
