Abstract
The aim of this work is to describe the technical characteristics of an AI-powered radiotherapy workflow that enables full-process automation (All-in-One), evaluate its performance implemented for on-couch initial treatment of rectal cancer, and provide insight into the behavior of full-workflow automation in the specialty of radiotherapy. The All-in-One workflow was developed based on a CT-integrated linear accelerator. It incorporates routine radiotherapy procedures from simulation, autosegmentation, autoplanning, image guidance, beam delivery, and in vivo quality assurance (QA) into one scheme, with critical decision points involved, while the patient is on the treatment couch during the whole process. For the enrolled ten patients with rectal cancer, minor modifications of the autosegmented target volumes were required, and the Dice similarity coefficient and 95% Hausdorff distance before and after modifications were 0.892±0.061 and 18.2±13.0 mm, respectively. The autosegmented normal tissues and automatic plans were clinically acceptable without any modifications or reoptimization. The pretreatment IGRT corrections were within 2 mm in all directions, and the EPID-based in vivo QA showed a γ passing rate better than 97% (3%/3 mm/10% threshold). The duration of the whole process was 23.2±3.5 minutes, depending mostly on the time required for manual modification and plan evaluation. The All-in-One workflow enables full automation of the entire radiotherapy process by seamlessly integrating multiple routine procedures. The one-stop solution shortens the time scale it takes to ready the first treatment from days to minutes, significantly improving the patient experience and the efficiency of the workflow, and shows potential to facilitate the clinical application of online adaptive replanning.
Introduction
In conventional radiotherapy (RT), the journey from the initial planning CT scan to the first treatment delivery is typically a multi-day process. This delay introduces two primary challenges: anatomical drift, where internal changes in the patient's anatomy (such as organ filling or tumor shrinkage) render the original plan less accurate, and workflow inefficiency caused by fragmented software systems and manual data transfers.
To address these issues, researchers developed the "All-in-One" solution, an AI-powered workflow designed to compress the entire radiotherapy process into a single session while the patient remains on the treatment couch. This approach aims to improve treatment accuracy by matching the plan to the patient's anatomy at the exact moment of delivery.
The All-in-One Workflow
The study utilized the uRT-linac 506c, a hybrid platform combining a diagnostic-quality 16-slice helical CT scanner with a C-arm linear accelerator. This hardware integration allows for "on-couch" simulation and treatment without relocating the patient.
The software backbone of the workflow includes several automated components:
- AI Autosegmentation: A multiresolution VB-Net convolutional neural network (CNN) identifies the clinical target volume (CTV) and organs-at-risk (OARs).
- Autoplanning: Voxel-based optimization algorithms generate treatment plans based on preset clinical goals.
- Integrated QA: Real-time verification is performed using an Electronic Portal Imaging Device (EPID) and Monte Carlo dose calculations.
As shown in the workflow diagram, the process includes critical decision points where oncologists and physicists review the AI-generated contours and plans. If the automated output fails evaluation, manual corrections are performed before proceeding.
Clinical Implementation and Results
The workflow was tested on a prospective cohort of 10 rectal cancer patients. The AI segmentation models were previously trained on a database of 195 patients to ensure robust performance.
Key Performance Metrics:
- Efficiency: The entire process, from simulation to completion of delivery, averaged 23.2 ± 3.5 minutes. This represents a massive reduction from the standard multi-day clinical timeline.
- Accuracy: The Dice similarity coefficient for target volumes was 0.892, indicating high agreement between AI and final corrected contours. Autosegmented normal tissues and treatment plans were found to be clinically acceptable without any manual re-optimization in all cases.
- Precision and Safety: Pre-treatment image guidance (IGRT) corrections were within 2 mm in all directions. In vivo QA showed a gamma passing rate of over 97%, confirming high fidelity between the planned and delivered doses.
Discussion and Future Outlook
A key insight from the study is that the primary bottleneck in the workflow is human intervention rather than machine computation. The total duration was largely driven by the time required for manual modifications and plan evaluation. However, by keeping the patient on the couch, the system inherently functions as a form of online adaptive radiotherapy (ART), mitigating systematic errors like "couch sag" that occur when moving between different machines.
While promising, the "All-in-One" solution currently requires high staffing intensity, with both a radiation oncologist and a medical physicist present for the duration of the session. Future work will focus on expanding the workflow to more complex anatomical sites, such as the head and neck, and exploring further AI integration to reduce the need for constant human supervision. This one-stop solution successfully demonstrates that the time scale for readying radiotherapy can be reduced from days to minutes, significantly improving both workflow efficiency and the patient experience.
Summary updated