Knowledge-based radiation treatment planning: A data-driven method survey.

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Biology

Background:

• Knowledge-based planning (KBP) in radiation therapy relies on historical treatment data.
• Accurate dose prediction is crucial for optimizing treatment plans and patient outcomes.
• Data-driven methods have emerged as key tools for advancing KBP.

Purpose of the Study:

• To comprehensively survey data-driven dose prediction methods in KBP over the past decade.
• To classify and analyze traditional KBP and deep learning (DL) approaches.
• To discuss the performance and future trends of these methods.

Main Methods:

• Systematic literature review of data-driven dose prediction methods for KBP.
• Classification of methods into traditional KBP and deep learning (DL) categories.
• Analysis based on framework, cancer site, and utilization of prior knowledge.

Main Results:

• Identified two main categories: traditional KBP methods using features and DL methods using neural networks.
• Reviewed advancements, key features, and methodologies within each category.
• Separated cited works by framework and cancer site for detailed analysis.

Conclusions:

• Both traditional and DL methods have shown promise in data-driven dose prediction for KBP.
• Deep learning methods offer potential for improved accuracy and efficiency.
• Future trends point towards hybrid approaches and further integration of DL in KBP.