Memorial Sloan Kettering Partners Lumonus To Advance AI Radiation Planning
AI-driven radiation planning at MSK moves toward wider clinical use through improving consistency and handling complex patient-specific constraints.
Memorial Sloan Kettering Cancer Center is collaborating with Lumonus to bring a clinically validated optimization system into broader use, aiming to improve how radiation therapy plans are created and delivered.
The New York–based cancer center will license and co-develop its ECHO optimization engine within Lumonus’ platform, according to a press release from Lumonus. The effort focuses on automating one of the most complex parts of radiation oncology: generating treatment plans that balance competing clinical constraints for each patient.
ECHO was developed at Memorial Sloan Kettering Cancer Center over nearly a decade and uses constraint-based mathematical optimization to compute radiation treatment plans. The system is designed to determine the best possible configuration based on clinical goals, rather than relying on manual iteration.
“Our goal in developing ECHO has always been to improve the quality, consistency, and efficiency of radiation therapy treatment planning,” Masoud Zarepisheh, Associate Attending and lead developer of ECHO at MSK, said in the release.
MSK’s work builds on years of deploying AI inside radiation oncology workflows. The center has developed models that automatically identify tumors and surrounding organs, reducing the need for manual contouring and improving consistency across cases. These systems have been used across more than 40 tissue types and in over 6,000 treatments, indicating routine clinical use rather than pilot-stage testing.
AI systems at MSK also account for anatomical variability that affects treatment accuracy. Organs can shift due to breathing or digestion, requiring plans to adapt to changing conditions. Models trained on imaging data help track these changes and improve targeting precision during therapy.
Despite these advances, treatment planning remains difficult to automate fully. Each case involves multiple competing objectives, including tumor coverage and protection of healthy tissue, which must be balanced based on physician intent. Existing systems have shown clinical value but have not consistently produced optimal plans at scale.
“Over the last twenty years, automated planning systems have demonstrated real clinical value,” Tim Fox, Chief Product Officer at Lumonus, said in the release. “But scaled adoption has remained elusive.”
The collaboration is aimed at addressing that gap. By integrating ECHO’s optimization framework into a clinical workflow platform, the companies are working to move automated planning from partial adoption to routine use.
Lumonus said the combined system will pair mathematically grounded optimization with tools that capture physician judgment, allowing plans to be generated and evaluated in a way that aligns with clinical decision-making.