[Paper Review] Applications of Artificial Intelligence in Particle Radiotherapy

Chao Wu, Dan Nguyen+4 more|arXiv (Cornell University)|Feb 5, 2021

Advanced Radiotherapy Techniques118 references4 citations

TL;DR

This paper provides a comprehensive review of artificial intelligence (AI) applications in particle radiotherapy, focusing on treatment planning, adaptive therapy, range and dose verification, and other clinical workflows. It demonstrates how AI leverages the physical advantages of particle therapy to enhance accuracy and efficiency, though challenges remain for widespread clinical adoption.

ABSTRACT

Radiotherapy, due to its technology-intensive nature and reliance on digital data and human-machine interactions, is particularly suited to benefit from artificial intelligence (AI) to improve the accuracy and efficiency of its clinical workflow. Recently, various artificial intelligence (AI) methods have been successfully developed to exploit the benefit of the inherent physical properties of particle therapy. Many reviews about AI applications in radiotherapy have already been published, but none were specifically dedicated to particle therapy. In this article, we present a comprehensive review of the recent published works on AI applications in particle therapy, which can be classified into particle therapy treatment planning, adaptive particle therapy, range and dose verification and other applications in particle therapy. Although promising results reported in these works demonstrate how AI-based methods can help exploit the intrinsic physic advantages of particle therapy, challenges remained to be address before AI applications in particle therapy enjoy widespread implementation in clinical practice.

Motivation & Objective

• To address the gap in existing literature by focusing specifically on AI applications in particle radiotherapy, rather than general radiotherapy.
• To identify and categorize AI methods that exploit the intrinsic physical properties of particle therapy for improved clinical outcomes.
• To evaluate the current state of AI in particle therapy, highlighting promising results and persistent challenges to clinical implementation.
• To provide a structured overview of AI applications across key stages of the particle therapy workflow, including treatment planning and real-time adaptation.
• To guide future research by identifying underexplored areas and technical barriers limiting AI adoption in particle therapy.

Proposed method

• Systematic review of peer-reviewed literature on AI applications in particle therapy, focusing on clinical workflow integration.
• Classification of AI applications into four main categories: treatment planning, adaptive therapy, range and dose verification, and other emerging uses.
• Analysis of AI techniques such as deep learning, convolutional neural networks (CNNs), and reinforcement learning applied to medical imaging and treatment planning data.
• Evaluation of AI models trained on patient-specific imaging (e.g., CT, PET) and treatment plans to predict optimal beam arrangements and dose distributions.
• Assessment of real-time AI systems for online adaptive replanning and range uncertainty correction using imaging feedback.
• Comparison of AI-based methods with conventional approaches in terms of speed, accuracy, and robustness using reported clinical and phantom study results.

Experimental results

Research questions

• RQ1 How can AI improve the accuracy and efficiency of particle therapy treatment planning?
• RQ2 What role can AI play in enabling real-time adaptive particle therapy during fractionated treatments?
• RQ3 How do AI-based methods enhance range and dose verification in particle therapy?
• RQ4 What are the key technical and clinical challenges hindering the widespread adoption of AI in particle therapy?
• RQ5 In what ways do AI techniques exploit the unique physical advantages of particle beams (e.g., Bragg peak) for better clinical outcomes?

Key findings

• AI-based treatment planning methods significantly reduce planning time while maintaining or improving plan quality compared to conventional planning.
• Adaptive particle therapy using AI enables real-time replanning based on daily imaging, improving target coverage and organ-at-risk sparing.
• AI models for range and dose verification demonstrate improved accuracy in predicting proton range uncertainties, reducing the need for extensive imaging fractionation.
• Deep learning-based approaches show promise in predicting patient-specific dose distributions with high spatial fidelity, even in complex anatomical regions.
• Despite strong performance in research settings, clinical implementation of AI in particle therapy is limited by data scarcity, model interpretability, and regulatory hurdles.
• The integration of AI with existing particle therapy systems remains challenging due to the need for robust, generalizable models across diverse patient populations and anatomical sites.