A physics-informed, plug-and-play dose engine for gradient-based radiotherapy treatment planning

TL;DR

The paper addresses the bottleneck of TPS-dependent radiotherapy planning by introducing PyDoseRT, a gradient-enabled, physics-informed dose engine implemented in PyTorch that maps delivery parameters to 3D dose distributions while preserving differentiability. The modular seven-layer architecture enables direct optimization of hardware-constrained parameters (e.g., leaf/jaw positions, gantry angles, and MU) with gradient flow, eliminating reliance on commercial TPS during research. Comprehensive validation across two prostate cohorts and water phantoms demonstrates high gamma pass rates (e.g., 3%/3 mm around 99% and 2%/2 mm around 97% in the Umeå cohort) and deliverable optimization outcomes, with gradient-based planning converging in about 10 minutes per patient on commodity GPUs. The open-source framework supports rapid exploration of adaptive planning, novel objective functions, and TPS-independent optimization, offering a practical platform for reproducible radiotherapy research and real-time planning concepts.

Abstract

Radiotherapy treatment planning remains a time-intensive iterative process requiring expert intervention in commercial treatment planning system (TPS). While machine learning approaches have demonstrated promise, most remain depedent on TPS-based dose calculation or surrogate dose models, preventing direct optimization of deliverable treatment plan parameters. We propose PyDoseRT (PDRT), a physics-informed, GPU-accelerated dose engine implemented in PyTorch that computes dose distributions directly from treatment delivery parameters (i.e., MLC leaf positions, jaw positions, gantry angles, and monitor units). The engine preserves gradient information throughout the dose computation pipeline, enabling gradient-based optimization of hardware-constrained treatment plans without the reliance on a commercial TPS. PDRT was evaluated on 19 and 162 clinical VMAT prostate cancer plans from two hospitals (with different treatment machines). When recalculating clinical plans, PDRT achieved high 3D gamma pass rates (mean 96.8% for 2%/2 mm and 98.9% for 3%/3 mm, depending on cohort). All optimized plans converged to clinically acceptable solutions and passed deliverability verification when imported into a commercial TPS. This physics-informed framework eliminates TPS dependency for radiotherapy optimization research by enabling gradient-based planning while ensuring that delivery parameters remain in the machine-feasible range. The gradient-enabled dose engine allows exploration of novel optimization strategies and objective functions while maintaining clinical validity. The proposed approach provides a research platform for investigating real-time adaptive radiotherapy concepts, automated planning workflows, and TPS-independent optimization strategies, and democratizing radiotherapy research, by exposing gradient-enabled, hardware-aware, open-source dose computation.

Figures (4)

• Figure 1: Schematic overview of . Treatment parameters and patient CT are processed through physics-based modules---hardware constrained projection, fluence generation with physics corrections, 3D projection with beam divergence, radiological depth via ray-tracing, depth-dependent kernel convolution, and coordinate transformation---producing final 3D dose distributions while preserving gradients throughout.
• Figure 2: Representative water phantom dose profiles for the Varian TrueBeam linac. Central-axis depth dose curves (left) comparing (solid line) with measurements (dashed). (right) Lateral profiles (right) at 10 cm depth showing central region agreement and penumbra modeling.
• Figure 3: Two examples of dose recalculated with (left) and the dose (middle) with two perpendicular line profiles through the isocenter (right) for a patient from the Umeå cohort (top) and from the Vienna cohort (bottom).
• Figure 4: Two example results of gradient-based optimization from the Umeå patient cohort.