dorsal/arxiv
View SchemaFeasibility of a General-Purpose Deep Learning Dose Engine: A Multi-Site Validation Study
| Authors | Yao Zhao, Ka Ho Tam, Raphael Douglas, Kyuhak Oh, Xin Wang, Ergys Subashi, Jinzhong Yang, Laurence Court, Dong Joo Rhee |
|---|---|
| Categories | |
| ArXiv ID | 2601.05348vv1 |
| URL | https://arxiv.org/abs/2601.05348 |
| License | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ |
Abstract
Conventional radiotherapy dose calculation algorithms are often computationally slow and non-differentiable, creating bottlenecks for online adaptive radiotherapy (ART) and limiting end-to-end automatic planning. Deep learning provides consistent inference performance and a differentiable framework essential for rapid optimization. In this study, we developed a generalized, site-independent deep learning dose engine using a beamlet-based input strategy. This establishes a computationally consistent and differentiable module that enables end-to-end training for autoplanning while maintaining accuracy across diverse geometries. A dataset of 3,600 plans from 120 patients across six anatomical sites was used to train two 3D convolutional neural networks, a standard U-Net and a Cascade U-Net, to predict 3D dose distributions from CT images and divergent MLC/jaw projections. Performance was validated via 3D gamma analysis on an independent cohort of 60 VMAT plans. The optimal model (U-Net with MAE loss) achieved a mean gamma passing rate of $98.9 \pm 1.6\%$ (3%/2mm, 10% threshold). Performance remained robust across all sites (passing rates $>98\%$), demonstrating that the beamlet-based strategy generalizes effectively to complex geometries without site-specific training. These results indicate that a single, site-independent model can calculate radiotherapy dose distributions with clinical accuracy. This differentiable engine is highly suitable for integration into end-to-end automatic planning, online ART, and secondary dose verification workflows.
{
"annotation_id": "cea7207f-1e46-4dbd-a1e2-6fea192ec527",
"date_created": "2026-02-17T05:53:03.730000Z",
"date_modified": "2026-02-17T05:53:03.730000Z",
"file_hash": "69b18e8606cd3cc428654ea5795524b9a0321068caf558fbdc796cb41200febd",
"private": false,
"record": {
"abstract": "Conventional radiotherapy dose calculation algorithms are often computationally slow and non-differentiable, creating bottlenecks for online adaptive radiotherapy (ART) and limiting end-to-end automatic planning. Deep learning provides consistent inference performance and a differentiable framework essential for rapid optimization. In this study, we developed a generalized, site-independent deep learning dose engine using a beamlet-based input strategy. This establishes a computationally consistent and differentiable module that enables end-to-end training for autoplanning while maintaining accuracy across diverse geometries. A dataset of 3,600 plans from 120 patients across six anatomical sites was used to train two 3D convolutional neural networks, a standard U-Net and a Cascade U-Net, to predict 3D dose distributions from CT images and divergent MLC/jaw projections. Performance was validated via 3D gamma analysis on an independent cohort of 60 VMAT plans. The optimal model (U-Net with MAE loss) achieved a mean gamma passing rate of $98.9 \\pm 1.6\\%$ (3%/2mm, 10% threshold). Performance remained robust across all sites (passing rates $\u003e98\\%$), demonstrating that the beamlet-based strategy generalizes effectively to complex geometries without site-specific training. These results indicate that a single, site-independent model can calculate radiotherapy dose distributions with clinical accuracy. This differentiable engine is highly suitable for integration into end-to-end automatic planning, online ART, and secondary dose verification workflows.",
"arxiv_id": "2601.05348",
"authors": [
"Yao Zhao",
"Ka Ho Tam",
"Raphael Douglas",
"Kyuhak Oh",
"Xin Wang",
"Ergys Subashi",
"Jinzhong Yang",
"Laurence Court",
"Dong Joo Rhee"
],
"categories": [
"physics.med-ph"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"title": "Feasibility of a General-Purpose Deep Learning Dose Engine: A Multi-Site Validation Study",
"url": "https://arxiv.org/abs/2601.05348",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "1875178f-3987-4d4b-90e0-4066817a0f86",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}