Systematic review of artificial intelligence in brachytherapy

Introduction
The objective of this study was to systematically review the scientific literature on the use of artificial intelligence (AI) in brachytherapy (BT), including deep learning and machine learning approaches. AI methods were quantitively and/or qualitatively compared with current clinical standards.

Material and methods
Included studies were accepted, peer-reviewed journal articles on AI in BT, published from January 1, 1980 till August 1, 2025 in PubMed, Google Scholar, Cochrane Library, and multi-institutional library databases. Articles were reviewed, and either included or excluded due to inclusion criteria or scope. Studies were searched for the application to BT, AI description, training and testing datasets, input and output of AI, treatment description, ground truth classification, accuracy compared with ground truth, and time for results. This review adhered to the Preferred Reporting Items for Systemic Reviews and Meta-Analyses (PRISMA) guidelines.

Results/Conclusions
A total of 232 studies were identified, which fulfilled inclusion criteria and scope after an initial yield of 10,820 results from database searches. Studies per application were 38, 74, 8, 40, 27, and 15 for applicator/needle reconstruction, segmentation, imaging applications, dose calculation, outcome prediction, and other planning applications, respectively. Studies per disease sites were 56, 66, 2, 3, 3, and 35 for prostate, gynecological, breast, choroidal, head and neck, and no specific site or multiple sites, respectively. The selected research demonstrated that AI may produce clinically acceptable planning data in significantly less time than required currently. The literature remains highly retrospective, but AI has the potential to reduce human effort and increase efficiency in repetitive tasks.