Precise Fluence Map Prediction for Intensity-Modulated Radiation Therapy by Leveraging nmODE

Intensity-modulated radiation therapy (IMRT) is a vital non-surgical approach in cancer treatment, with fluence map prediction playing a pivotal role in treatment planning. We introduce a novel model, "nmODE-UNet", which enhances fluence map prediction accuracy by incorporating dynamic modeling through the "nmODE block" within the UNet architecture. Our study focuses on rectal cancer cases and employs a dataset of 90 patients. Comparative evaluations against advanced models reveal the superiority of nmODE-UNet, evident through metrics like Mean Absolute Error (MAE) and Peak Signal-to-Noise Ratio (PSNR). The experimental results reveal a significant performance advantage of our proposed nmODE-UNet.