Deep learning applied to dose prediction in external radiation therapy: A narrative review

Area of Science:

• Medical Physics
• Radiotherapy
• Artificial Intelligence

Background:

• The integration of artificial intelligence (AI), machine learning (ML), and deep learning (DL) in medicine has significantly increased over recent decades.
• These technologies have demonstrated success in medical image segmentation, motion management, and predicting treatment outcomes.
• Early investigations into ML and DL for dose calculation and quality assurance in radiotherapy date back to the year 2000.

Purpose of the Study:

• To provide a comprehensive overview of deep learning architectures and models applied in external radiotherapy.
• To discuss the performance and future potential of deep learning predictive models in optimizing radiotherapy workflows.
• To highlight how AI, ML, and DL enhance the accuracy and efficiency of dosimetry and quality assurance processes.

Main Methods:

• Review of existing literature on deep learning architectures and models used in external radiotherapy.
• Analysis of advancements in dose calculation models driven by information research and architectural innovations since 2014.
• Examination of knowledge-based and patient-specific methods for improved dose prediction accuracy.

Main Results:

• Deep learning models have shown significant promise in improving the speed and accuracy of dose calculations in radiotherapy.
• AI-driven approaches enhance quality assurance processes, reducing time-consuming aspects of clinical patient management.
• Innovations in model architectures and knowledge-based methods have led to more accurate dose predictions.

Conclusions:

• Deep learning is a transformative technology in external radiotherapy, offering substantial improvements in treatment planning and delivery.
• Continued research and development in deep learning architectures will further refine predictive models for enhanced patient care.
• The application of AI, ML, and DL is crucial for advancing the efficiency and accuracy of radiotherapy workflows.