Artificial intelligence applications in brachytherapy: A literature review

No prior work had resolved the full scope of computational advancements in internal radiation therapy workflows. Traditional methods often rely on manual labor that consumes significant clinical time and resources. That uncertainty drove researchers to investigate how automated systems might assist medical physicists. It was already known that machine learning models could enhance image interpretation in various diagnostic settings. However, the specific integration of these tools into brachytherapy remained fragmented across diverse publications. This gap motivated a comprehensive assessment of current technological progress in the field. Prior research has shown that algorithmic assistance can improve consistency in complex medical procedures. The current landscape lacks a unified summary of how these diverse computational approaches impact clinical practice.

The aim of this review is to provide a comprehensive overview of how computational intelligence can simplify and optimize various stages of the internal radiation therapy workflow. The authors seek to synthesize existing evidence regarding the integration of automated systems into clinical practice. This work addresses the need to understand how diverse algorithmic approaches impact the efficiency of complex medical procedures. By categorizing current research, the study clarifies which specific tasks are most amenable to automation. The motivation stems from the potential for these technologies to reduce the manual burden on medical physicists and clinicians. The review also identifies gaps in current reporting practices that hinder the comparison of different models. Ultimately, the authors intend to highlight the current state of the field to guide future development. This effort provides a foundation for understanding the transition toward more automated and precise radiation delivery.

The review approach involved a systematic search of three major medical databases conducted in June 2022. Investigators targeted publications proposing computational applications for internal radiation procedures. The team screened records to identify studies that met specific inclusion and exclusion criteria. A total of eighty relevant documents were selected for detailed analysis. The authors organized these papers into six functional categories representing different stages of clinical practice. This classification scheme allowed for a structured evaluation of diverse technical methodologies. The researchers assessed a wide range of models, from classical statistical frameworks to modern deep learning architectures. This methodology provided a comprehensive overview of how various algorithms support or replace manual tasks.

Key findings from the literature indicate that eighty studies successfully implemented computational tools to optimize internal radiation workflows. Segmentation emerged as the most frequent application, accounting for twenty-four of the analyzed papers. Reconstruction of applicators, catheters, and needles represented the second largest category with sixteen identified studies. Preplanning and dose prediction were addressed in thirteen and eleven papers, respectively. Quality assurance tasks were the focus of ten publications, while registration and image processing appeared in six. Many of these algorithms demonstrated performance levels equivalent to human experts. The results suggest that these technological interventions offer significant improvements in the speed of clinical operations.

The authors suggest that automated algorithms possess the capacity to enhance, streamline, and expedite numerous stages of the internal radiation process. Evidence indicates that several models achieve performance levels comparable to human experts in specific tasks. These systems offer notable gains in operational efficiency for busy clinical departments. The review highlights a requirement for investigators to follow established reporting standards in future publications. Authors emphasize the necessity of utilizing larger datasets to ensure robust model validation. Reporting findings through metrics that clinicians can easily interpret remains a priority for the field. The synthesis indicates that computational tools are poised to transform standard workflows by reducing manual burdens. Future efforts should focus on standardized evaluation to facilitate broader clinical adoption of these technologies.