Total marrow and lymph nodes irradiation in conditioning regimen for allogeneic hematopoietic stem cell transplantation Artificial Intelligence algorithms to automate the Total Marrow and Lymph-node Irradiation by VMAT optimization using WB-CT/MRI and synthetic WB-CT - The AutoMI project

Purpose/Objective The AuToMI project focuses on improving the TMLI clinical workflow and automating radiotherapy procedures through the application of Artificial Intelligence (AI). The primary goal is to streamline and enhance the efficiency of the radiotherapy treatment process by automating two critical steps: contouring of the target volumes and treatment plan optimization. In parallel, the project aims to integrate whole-body magnetic resonance imaging (WB-MRI) to improve the accuracy of lymph-node segmentation, thereby reducing uncertainties in treatment planning. A secondary but essential objective is to develop tools that enable the generation of synthetic WB-CT images from WB-MRI data, and optimizing clinical imaging for more precise radiotherapy treatments. Material/Methods This is a single-arm, mono-institutional study involving adult patients undergoing total marrow and lymph-node irradiation (TMLI) as part of the conditioning regimen for Bone Marrow Transplantation (BMT). The workflow integrates WB-CT and WB-MRI for target definition, focusing on automating contouring and planning steps using AI algorithms trained on retrospective and prospective patient data. WB-MRI is specifically employed to improve the segmentation of lymph nodes, addressing uncertainties typically encountered in WB-CT without alone. Additionally, the project explores the creation of synthetic WB-CT from WB-MRI, aiming to substitute conventional imaging steps. The primary endpoint was the efficiency and accuracy of the AI-driven automation in contouring and planning. The secondary endpoints were evaluating lymph-node segmentation and the implementation of synthetic WB-CT generation. Results Preliminary results demonstrate that AI significantly optimizes the clinical workflow, reducing the time required for both contouring and treatment plan generation without compromising accuracy and safety. Automated contouring and planning procedures delivered clinically acceptable outcomes, with comparable dosimetric parameters to manually optimized plans. The use of WB-MRI for lymph-node segmentation resulted in a significant reduction of treatment volumes and improved accuracy, lowering the uncertainty typically associated with WB-CT-based planning. Additionally, the synthetic WB-CT generated from WB-MRI showed a potential future alternative to the traditional CT-based approach in radiotherapy planning. Conclusion The AuToMI project successfully demonstrates that AI can substantially optimize the clinical workflow in radiotherapy by automating contouring and planning processes. This automation enhances efficiency and reduces manual workload, allowing for faster and more accurate treatment delivery. The use of WB-MRI improves lymph-node segmentation, decreasing uncertainty and potentially enhancing treatment precision. Moreover, the development of synthetic WB-CT from WB-MRI represents a promising innovation that could further streamline clinical workflows and improve imaging accuracy. These advancements position the AuToMI approach as a valuable tool for improving the quality and accessibility of radiotherapy in oncohematology.