| The Potential Role of Artificial Intelligence in Radiotherapy Treatment Planning |
| Paper ID : 1113-ISCH |
| Authors |
|
Hazem Wadaa Allah *1, Ahmed Mousa2, Marwa Yassin3, Magdy Khalil4 1Biomedical Physics, Department of Physics, Faculty of Science, Helwan University, Cairo, Egypt. 2Nasser Institute, Radiotherapy Department, Cairo, Egypt 3Department of Phyiscs, Faculty of Science, Helwan University 4Department of Physics, Faculty of Science, Helwan University |
| Abstract |
| Background Radiotherapy planning determines how radiation dose is distributed within the body, typically assessed through dose statistics and Dose Volume Histograms (DVHs). the process is time-consuming and requires multiple iterations to achieve an optimized plan that delivers the prescribed dose to the tumor while sparing organs at risk (OARs) according to the “As Low As Reasonably Achievable” (ALARA) principle. Recent advances in deep learning have demonstrated the potential to predict optimized dose distributions based on prior clinical plans. Purpose This study evaluates the ability of a deep learning model to generate predicted dose distributions and validate them against clinical reference plans using dosimetric and DVH metrics. Methods A retrospective dataset of radiotherapy plans was used for model training and testing. Predicted dose distributions were compared with corresponding clinical plans using standard parameters (D95, Dmean, Dmax) and DVH curves for targets and OARs. Agreement was assessed through statistical analyses. Results The model reproduced dose distributions closely aligned with clinical plans. Predicted PTV D95 differed by only 1.2%, while OAR mean dose deviations were within ±5%. The dice similarity coefficient between predicted and clinical dose distributions was 0.88 ± 0.03. The gamma index pass rate (3%/3 mm) exceeded 95% across all cases. Prediction time averaged <80 seconds per plan compared to 30–60 minutes for manual optimization. Conclusion Artificial Intelligence can accelerate radiotherapy planning by providing rapid, knowledge-based dose predictions. While promising for clinical decision support and quality assurance, further research is needed to achieve fully automated end-to-end planning. |
| Keywords |
| Artificial Intelligence, Deep Learning, Radiotherapy Planning, Dose Prediction, Treatment Planning Optimization |
| Status: Abstract Accepted (Oral Presentation) |