TrDosePred: A deep learning dose prediction algorithm based on transformers for head and neck cancer radiotherapy.

Area of Science:

• Medical Physics
• Artificial Intelligence in Medicine
• Radiation Oncology

Background:

• Intensity-Modulated Radiation Therapy (IMRT) is a standard cancer treatment.
• IMRT treatment planning is currently time-consuming and labor-intensive.

Purpose of the Study:

• To develop a novel deep learning algorithm, TrDosePred, for automated dose prediction in head and neck cancer radiotherapy.
• To reduce the time and effort required for IMRT treatment planning.

Main Methods:

• TrDosePred utilizes a U-shape network with convolutional patch embedding and self-attention transformers.
• The model generates dose distributions from contoured CT images, incorporating data augmentation and an ensemble approach.
• Training and evaluation were performed using the Open Knowledge-Based Planning Challenge (OpenKBP) dataset and metrics.

Main Results:

• The TrDosePred ensemble achieved a Dose score of 2.426 Gy and a DVH score of 1.592 Gy on the test dataset.
• Performance ranked 3rd and 9th on the CodaLab leaderboard, demonstrating competitive results.
• Relative Mean Absolute Error (MAE) against clinical plans was 2.25% for targets and 2.17% for organs at risk.

Conclusions:

• A transformer-based framework, TrDosePred, was successfully developed for radiation dose prediction.
• The algorithm demonstrates comparable or superior performance to state-of-the-art methods.
• Transformers show potential for significantly improving radiotherapy treatment planning efficiency.