In the first year of the project, we have developed innovative algorithms that will be part of the proprietary software to enable MR-based and MR-guided radiotherapy planning on the fly. These algorithms serve to:
1. Perform anatomical simulation using MRI images that lead to rapidly updating the treatment plan on-the-fly.
2. Automatically generate pseudo CTs from T1 MR Images of the brain, abdomen and pelvis using an ensemble strategy of data-driven artificial intelligence techniques. To this end, an end-to-end deep learning framework of 3D generative adversarial networks (GANs) that takes as input the original MRI and generates a pseudo-CT has been developed.
3. Perform automatic contouring on Gadolinium enhanced T1 Sequence MRIs. Contouring organs and tumors for each radiotherapy session from scratch is not practically feasible and a major bottleneck in moving into a more dynamic planning system. With our new solution, plans can be adapted to the changing size/shape of the tumor during the course of the treatment period.
4. Achieve real-time dose planning through deformable registration between the planning CT and MR image of each session. The resulting deformation field allows rapid assessment of the plan for each treatment session using dose volume histograms while an efficient local search technique enables plan adjustment to the current position of the patient. MR-linear accelerators enable the change of imaging modality to MRI during treatment sessions, reducing radiotoxicity and resulting in a more accurate patient positioning. This solution uses our proprietary algorithms for registration of multimodal medical images and pseudo-CT generation.
5. Finalize the prototype of our unique AI based radiotherapy software for MR-based radiotherapy. The software is cloud based, capable of running both on public or private cloud, granting access to the full power of the cloud for medical practitioners and decreasing the burden on IT professionals of hospitals. SmartPlan offers automatic contouring and AI-enhanced treatment plans by leveraging the latest advances in AI and GPU technology. This first version of the software will be used in the clinical trials performed by our three clinical partners.
In addition, from a less technical point of view, we have 1) ensured smooth communication and collaboration with our partners through regular meetings and well-defined next steps and deadline, 2) hired a project manager and research manager who is facilitating this liaison and following up closely with our teams to ensure timely progress of the milestones, 3) undertaken many different dissemination strategies, including the participation at different key conferences, preparation of abstracts of our partners, new development of our website, press-releases, 4) worked on the quality and regulatory aspect of the project through a newly hired QMS manager, 5) reacted fast to unforeseen risks, especially to the unforeseen COVID-situation, 6) secured and started the clinical trial to assess the performance of the developed solution, 7) expanded the work started through ART-AI centers (e.g. Acibadem University Hospital) which will not only be early adopters of our product but also facilitate market acceptance via their extensive industry and academic networks.
