<ul data-eligibleForWebStory="true"><li>Anatomical changes during head-and-neck cancer proton therapy can necessitate treatment replanning, but current methods are labor-intensive.</li><li>A patient-specific deep reinforcement learning framework for automated replanning using a reward-shaping mechanism was proposed.</li><li>The framework trained personalized agents for each patient to adjust optimization priorities for maximizing plan quality.</li><li>The approach leverages anatomical similarities to adapt plans effectively throughout treatment.</li><li>Two deep reinforcement learning algorithms, Deep Q-Network and Proximal Policy Optimization, were implemented.</li><li>Both algorithms improved initial plan scores significantly, surpassing manual replans from a human planner.</li><li>Evaluation on five head-and-neck cancer patients showed improved tumor coverage and organ-at-risk sparing.</li><li>The work highlights the potential of deep reinforcement learning in adaptive proton therapy for addressing dosimetric complexities.</li></ul>

Patient-Specific Deep Reinforcement Learning for Automatic Replanning in Head-and-Neck Cancer Proton Therapy

Discover more