A Decision-Making Method for Photon/Proton Selection for Nasopharyngeal Cancer Based on Dose Prediction and NTCP

Area of Science:

• Oncology
• Medical Physics
• Radiotherapy

Background:

• Nasopharyngeal cancer treatment decisions between photon and proton therapy require complex plan comparisons.
• Current methods are time-consuming and demand specialized expertise.
• An automated tool can streamline patient selection for proton therapy.

Purpose of the Study:

• To develop and validate a fully automated decision tool for selecting nasopharyngeal cancer patients for proton therapy.
• To predict dose distributions for both proton and photon therapy using CT images.
• To assess the likelihood of xerostomia and dysphagia based on predicted organ doses and the NIPP protocol.

Main Methods:

• Developed deep learning models to predict photon therapy (PT) and proton therapy (XT) dose distributions from patient CT scans.
• Trained models on data from 48 nasopharyngeal cancer patients with manually generated plans.
• Utilized the Netherlands' National Indication Protocol Proton therapy (NIPP) to evaluate predicted normal tissue complication probabilities (NTCP) for xerostomia and dysphagia.

Main Results:

• Predicted dose distributions showed high accuracy compared to manual plans (Mean Absolute Error < 5%).
• The tool achieved a 93.8% correct patient selection rate with an Area Under the Curve (AUC) of 0.86.
• NTCP predictions for xerostomia and dysphagia demonstrated good performance, with no significant difference for proton therapy.

Conclusions:

• The developed deep learning-based decision tool accurately identifies nasopharyngeal cancer patients who can benefit from proton therapy.
• The tool significantly reduces planning time and enhances diagnostic efficiency for clinicians.
• This automated system is valuable for centers lacking proton therapy expertise, promoting wider adoption.