Domain Knowledge Driven 3D Dose Prediction Using Moment-Based Loss Function
This work addresses the problem of accurate and efficient dose prediction for clinicians in radiation therapy planning, though it is incremental as it builds on existing deep learning frameworks with a new loss function.
The paper tackled the challenge of incorporating non-convex and non-differentiable dose volume histogram (DVH) metrics into deep learning models for 3D dose prediction in lung radiation therapy, by proposing a novel moment-based loss function that improved DVH-score by 11% compared to baseline methods while reducing computational cost by 48%.
Dose volume histogram (DVH) metrics are widely accepted evaluation criteria in the clinic. However, incorporating these metrics into deep learning dose prediction models is challenging due to their non-convexity and non-differentiability. We propose a novel moment-based loss function for predicting 3D dose distribution for the challenging conventional lung intensity modulated radiation therapy (IMRT) plans. The moment-based loss function is convex and differentiable and can easily incorporate DVH metrics in any deep learning framework without computational overhead. The moments can also be customized to reflect the clinical priorities in 3D dose prediction. For instance, using high-order moments allows better prediction in high-dose areas for serial structures. We used a large dataset of 360 conventional lung patients with 2Gy $\times$ 30 fractions to train the deep learning (DL) model using clinically treated plans. We trained a UNet-like CNN architecture using computed tomography (CT), planning target volume (PTV) and organ-at-risk contours (OAR) as input to infer corresponding voxel-wise 3D dose distribution. We evaluated three different loss functions: (1) Mean Absolute Error (MAE) Loss, (2) MAE + DVH Loss, and (3) the proposed MAE + Moments Loss. The quality of the predictions was compared using different DVH metrics as well as dose-score and DVH-score, recently introduced by the AAPM knowledge-based planning grand challenge. Model with (MAE + Moment) loss function outperformed the model with MAE loss by significantly improving the DVH-score (11%, p$<$0.01) while having similar computational cost. It also outperformed the model trained with (MAE+DVH) by significantly improving the computational cost (48%) and the DVH-score (8%, p$<$0.01). The code, models, docker container, and Google Colab project are available on our DoseRTX GitHub (https://github.com/nadeemlab/DoseRTX).