Advancing Radiotherapy Quality Assurance in Clinical Trials Through AI
· Radiotherapy Quality Assurance (RT QA) is a critical determinant of trial validity, yet manual contour review remains time-consuming and prone to inter-expert variability.
· AQUILAB by Coexya has integrated MVision AI Contour+ AI segmentation into the ONCO PLACE platform to standardize and accelerate RT QA across multicenter trials.
· The integration supports both prospective Individual Case Review and retrospective re-segmentation for ancillary research on completed or ongoing trials.
· Sponsors and CROs benefit from improved scientific rigor, reduced review burden, and stronger regulatory readiness - while preserving expert clinical oversight.
Radiotherapy (RT) plays an essential role in cancer treatment, used in over 50% of patients [1,2], with major advances in tumor targeting and treatment adaptation that have made it an increasingly sophisticated component of multi-modal cancer care. It also features as an active treatment component in a substantial proportion of oncology clinical trials - a role set to expand with the rapid growth of radiotherapy-immunotherapy combination evaluations [3,4].
In this context, high-quality Radiotherapy Quality Assurance (RT QA) is vital to controlling and minimizing inter-operator and inter-site variability in contouring and plan execution and, ultimately, the credibility and interpretability of trial outcomes [5]. However, RT QA of multicenter clinical trials remains challenging as it is time and resource consuming, notwithstanding the fact that there may be substantial inter-expert variability during case review.
| AQUILAB by Coexya and MVision AI have entered into partnership to integrate Contour+ segmentation models into RT QA clinical trials workflows. This integration enhances the efficiency, consistency and scientific rigor of QA processes, while preserving a necessary clinical judgment and expert review – key priorities for Sponsors and Contract Research Organization (CROs). |
Radiotherapy protocols in clinical trials require precise definition of target volumes and organs at risk. Variability in contouring and plan quality has been demonstrated to significantly impact both clinical endpoints and patient outcomes in multiple studies [6][7][8].
Traditional RT QA relies on extensive manual review of contours, which increases timeline variability and resource burden in large, multicenter trials. AI-augmented methods represent an opportunity to standardize review criteria, reduce workload, and accelerate decision timelines.
Recent French clinical research guidelines [9] highlight the inherent complexity of radiotherapy trials and the critical importance of rigorous, centralized quality control processes. The AQUILAB – MVision AI approach directly addresses these structural challenges, supporting scalable, high-quality RT QA for modern and efficient oncology trials.
Beyond prospective RT QA, the ability to retrospectively re-contour imaging datasets has become increasingly important in RT clinical research. Many trials evolve over time, with new scientific questions emerging that were not anticipated during initial protocol design, such as assessing dose received by newly relevant organs at risk, exploring predictive biomarkers, or conducting radiomics analyses.
As historical datasets often lack standardized or sufficiently detailed contouring, retrospective segmentation enables the creation of harmonized anatomical structures across all patients and centers. This process ensures comparability, reduces bias introduced by inter-observer variability, and provides analysis-ready cohorts for ancillary or exploratory studies.
AI-driven re-contouring allows these retrospective enhancements to be performed at scale, without the resource burden of manual delineation, thereby unlocking additional scientific value from completed or ongoing trials.
Company Profiles
AQUILAB by Coexya |
| AQUILAB by Coexya is a French company specialized in quality control solutions for radiotherapy and medical imaging with more than 25 years of experience. Its ONCO PLACE platform supports data collection and RT QA in more than 100 clinical trials worldwide. It allows for centralized management of DICOM RT data, facilitates the anonymization and organization of imaging and RT data, supports expert review processes, and standardizes radiotherapy planning and imaging data across multiple sites. |
AQUILAB's ONCO PLACE platform enables:
▸ coordinated workflows across investigators, sponsors, and expert reviewers;
▸ standardized data structuring for RT and imaging;
▸ expert review tools integrated with its web-viewer ARTWEB;
▸ audit trails, protocol-specific alerting, and reporting features.
AQUILAB’s white paper publication on RTQA further underscores the clinical impact of RT QA on study validity by citing studies that link QA compliance with improved outcomes and protocol reliability, and emphasizes the importance of prospective, centralized review.
MVision AI |
| MVision AI is a Finnish company based in Helsinki specializing in cloud-based, AI-driven solutions that accelerate and standardize radiotherapy workflows globally. The company's Workspace+ platform integrates multiple AI technologies to support clinical teams across treatment preparation activities. |
At the core of Workspace+ is Contour+, a guideline-based segmentation solution deployed across 24 countries that automates the delineation of organs at risk and lymph node regions following international consensus guidelines.
Workspace+ extends beyond segmentation to support modern radiotherapy workflows through synthetic CT generation for MR-only treatment planning, CBCT enhancement and deformable contour propagation for offline replanning, and AI-driven dose prediction that establishes personalized optimization objectives to streamline the planning process.
Through its integrated cloud-based platform, MVision AI reduces manual workload, increases anatomical consistency, and supports clinical decision-making across radiotherapy teams worldwide.
The collaboration integrates MVision AI's Contour+ segmentation models into the ONCO PLACE platform at two pivotal stages in RT clinical trials:
Individual Case Review (ICR)
Upon submission of a case by an investigator site, Contour+ automatically generates new DICOM contour sets based on international consensus guidelines. These AI-predicted contours serve as a reference standard for comparison against site-provided structures.
This comparison can identify missing structures, laterality errors, or significant morphological discrepancies, allowing expert reviewers to focus their attention on cases requiring clinical judgment rather than routine protocol compliance.
Benefits include consistent and rapid error detection, reduced review time, standardized feedback generation, and improved site training through automated quality flags.
This workflow reduces the manual review burden while preserving the expert oversight — a critical compliance objective for Sponsors and CROs.
Retrospective Re-Segmentation and Harmonization
Many trials accumulate large repositories of imaging and RT data that may lack complete structure sets for emerging research questions. The integrated Contour+ models enable:
▸ consistent harmonized segmentation across existing datasets;
▸ automated delineation of newly required structures for ancillary analyses;
▸ creation of uniform cohorts for secondary endpoints including radiomics or dosimetric outcome studies.
This retrospective process generates analysis-ready datasets efficiently without requiring exhaustive manual rework, unlocking additional scientific value from completed or ongoing trials.
Scientific Rigor and Data Quality
▸ AI-assisted segmentation increases concordance between reviewed contours and protocol-defined reference standards.
▸ Standardized reference contours reduce inter-site variability, improving the scientific robustness of trial data.
Efficiency and Resource Optimization
▸ Automated contouring reduces expert review time and supports faster case turnaround.
▸ Scalable workflows better accommodate large, multicenter global trials.
Regulatory & Trial Readiness
▸ Documented AI-assisted workflows and audit trails strengthen RT QA reporting.
▸ Enhanced standardization aligns with regulatory expectations for data integrity and reproducibility.
The AQUILAB by Coexya – MVision AI collaboration demonstrates how AI can be systematically integrated into RT QA workflows to support sponsors and CROs in executing high-quality clinical trials.
By enhancing consistency, reducing manual workload, and enabling harmonized dataset creation, this partnership exemplifies a scalable model of AI-augmented quality assurance aligned with scientific, operational, and regulatory priorities.
Interested in integrating AI into your RT QA workflows? Contact AQUILAB by Coexya: info@aquilab.com |
USE CASE 1 Automatic Contouring RT QA in Head & Neck Studies AQUILAB by Coexya / GORTEC - Prof. Juliette Thariat (François Baclesse Cancer Center – Caen) Several GORTEC clinical trials in head and neck cancer included prospective RT QA focused primarily on the verification of GTV delineation [10]. To expand and strengthen the assessment of treatment consistency, AQUILAB by Coexya and Prof. Juliette Thariat (François Baclesse Cancer Center – Caen) initiated a retrospective automated RT QA program to evaluate the quality of contouring across organs at risk (OARs). By applying AI-driven segmentation and automated structure-checking algorithms to historical datasets, the workflow identified multiple categories of deviations that had not been systematically reviewed during the original prospective QA, including laterality errors, incorrect structure naming, and contouring discrepancies on key OARs.
This automated retrospective assessment provides a comprehensive view of data quality, highlighting previously undetected protocol deviations and demonstrates the value of standardized, scalable QA processes for multicenter cooperative group trials. The work is being conducted as part of an ongoing PhD project in partnership between AQUILAB by Coexya and GORTEC, aiming to formalize automated RT QA methodologies and generate evidence for their integration into future prospective GORTEC studies. |
USE CASE 2 AI-Based Prostate MRI Segmentation for Radiomics in the HYPO-RT UNICANCER Project Prof. David Pasquier - Oscar Lambret Cancer Center - Lille / UNICANCER Network The HYPO-RT project, coordinated by Prof. David Pasquier (Oscar Lambret Cancer Center – Lille, France) within the UNICANCER network, aims to evaluate and compare different prostate hypofractionation regimens across 13 participating centers. As part of this multicenter evaluation, AQUILAB by Coexya supports the extraction of imaging and radiotherapy data.
As a complementary research objective, the team is conducting a radiomics study based on pre-treatment prostate MRI to identify potential predictive biomarkers. Given that the MRI contours from the original dataset are not available, there is a need to retrospectively contour all MRIs using MVision AI's dedicated prostate models. These harmonized, AI-generated contours enable robust radiomics feature extraction and support predictive modeling efforts, enhancing the scientific value of the HYPO-RT dataset while avoiding extensive manual contouring. |
USE CASE 3 Retrospective AI-Based Re-Segmentation for Cardiac Substructure Dosimetry in the RTEP7 Trial Prof. Pierre Vera & Prof. Sébastien Thureau (Henri Becquerel Cancer Center - Rouen) The RTEP7 study is a multicenter phase 2 study investigating adaptive radiotherapy guided by interim [¹⁸F]FDG-PET in patients with inoperable stage III non-small-cell lung cancer (NSCLC) [11]. The study benefited from prospective retrieval of all imaging data and radiotherapy treatment plans in ONCO PLACE platform. As the research team progresses toward a deeper understanding of treatment-related toxicity, a new objective has emerged: assessing dose distributions to cardiac substructures, which were not originally segmented during the study.
To enable this analysis, the full imaging dataset has been reprocessed using MVision AI's segmentation models. This retrospective AI-driven re-segmentation provides harmonized, high-resolution delineation of cardiac substructures across all patients, ensuring consistent and reproducible structure definitions. The resulting standardized contours will allow precise dosimetric extraction and correlation with adverse cardiac events. By leveraging AI to enrich an existing clinical dataset, this workflow enables the generation of new scientific insights without requiring any additional imaging or manual contouring resources, thereby accelerating ancillary research and increasing the long-term value of the clinical trial. |
References
[1] Barton MB, Jacob S, Shafiq J, et al. Estimating the demand for radiotherapy from the evidence: A review of changes from 2003 to 2012. Radiotherapy and Oncology 2014;112:140–4. https://doi.org/10.1016/j.radonc.2014.03.024
[2] Zhu H, Chua MLK, Chitapanarux I, et al. Global radiotherapy demands and corresponding radiotherapy-professional workforce requirements in 2022 and predicted to 2050: a population-based study. Lancet Glob Health 2024;12:e1945–53. https://doi.org/10.1016/S2214-109X(24)00355-3
[3] Vanneste BGL, van Limbergen EJ, Reynders K, et al. An overview of the published and running randomized phase 3 clinical results of radiotherapy in combination with immunotherapy. Transl Lung Cancer Res 2021;10:2048–58. https://doi.org/10.21037/tlcr-20-304
[4] Li M, Ding P. Characteristics and innovative points of clinical trials of radiotherapy combined with immune checkpoint inhibitors in NSCLC over the past decade. Front Med (Lausanne) 2025;12. https://doi.org/10.3389/fmed.2025.1598505
[5] Peters LJ, O’Sullivan B, Giralt J, et al. Critical Impact of Radiotherapy Protocol Compliance and Quality in the Treatment of Advanced Head and Neck Cancer: Results From TROG 02.02. Journal of Clinical Oncology 2010;28:2996–3001. https://doi.org/10.1200/JCO.2009.27.4498
[6] Zhong H, Men K, Wang J, et al. The impact of clinical trial quality assurance on outcome in head and neck radiotherapy treatment. Front Oncol 2019;9:1–7. https://doi.org/10.3389/fonc.2019.00792
[7] Mir R, Groom N, Mistry HB, et al. Association between radiotherapy protocol variations and outcome in the CONVERT trial. Clin Transl Radiat Oncol 2023;39. https://doi.org/10.1016/j.ctro.2022.100560
[8] Jomy J, Sharma R, Lu R, et al. Clinical impact of radiotherapy quality assurance results in contemporary cancer trials: a systematic review and meta-analysis. Radiotherapy and Oncology 2025;207. https://doi.org/10.1016/j.radonc.2025.110875
[9] Créhange G, Hennequin C, Allignet B, et al. Specific features of clinical research in radiotherapy in France. Cancer/Radiothérapie 2025;29:104768. https://doi.org/10.1016/j.canrad.2025.104768
[10] Özer Ö, Shafi H, O’Reilly D, et al. Need for standardization in the use of structures in the intensity-modulated radiation therapy planning of head and neck cancers, a GORTEC study. Radiotherapy and Oncology 2023;188:109895. https://doi.org/10.1016/j.radonc.2023.109895
[11] Vera P, Thureau S, Le Tinier F, et al. Adaptive radiotherapy (up to 74 Gy) or standard radiotherapy (66 Gy) for patients with stage III non-small-cell lung cancer, according to [18F]FDG-PET tumour residual uptake at 42 Gy (RTEP7–IFCT-1402): a multicentre, randomised, controlled phase 2 trial. Lancet Oncol 2024;25:1176–87. https://doi.org/10.1016/S1470-2045(24)00320-6
