Automated and Clinically Optimal Radiotherapy Cancer Treatment Planning


Planning cancer treatments with radiation to sterilize cancer cells is a global problem. Every year, worldwide, there are about 18 million new cases of cancer, more than a third of which are treated with radiation.

Radiotherapy is the use of carefully chosen beams of high-energy radiation to sterilize cancer and is often given in conjunction with other types of treatment such as chemotherapy and/or surgery. The radiation is delivered directly into the affected area to kill cancerous cells without harming the surrounding healthy organs and tissue. The treatment is complex and very patient specific; it must be uniquely tailored for each patient. 

Memorial Sloan Kettering Cancer Center (MSKCC) has developed and applied advanced optimization tools (e.g., hierarchical constrained optimization, convex approximation, Lagrangian multipliers), along with careful modeling of the radiation delivery process, to provide better planning, faster, and cheaper radiotherapy treatment. The need for complex, high-dimensional constrained optimization in cancer care is clear, because each cancer patient is unique in terms of the shape and location of the tumor and radiation sensitive surrounding healthy tissues. 

Radiation is delivered to the patient’s body from various orientations. Each "intensity modulated" radiation treatment requires customization of radiation intensity maps of delivered beams to kill cancerous cells without causing excessive harm to nearby normal organs and tissues. Finding these patient-specific settings, correctly prioritizing different dose goals to disease, and achieving the desired avoidance to multiple normal tissues, is a complex and labor-intensive task that must be achieved in a limited time. 

Most current planning optimization methods either rely on the radiotherapy planners’ experience and skill to guide these tradeoffs or seek to create a plan based on characteristics of past acceptable plans.

The Approach

MSKCC formulated this problem as a hierarchical constrained optimization problem to better capture the underlying clinical philosophy of controlling acceptable dose tradeoffs within the plan: some dose goals are more important than other goals. For example, delivering adequate tumoricidal dose to mouth cancer may be considered more important than avoiding a reduction in salivary function. 

For each patient, the delivery machine parameters are optimized by solving two large-scale sequential constrained optimization problems. The first-level optimization problem guarantees adequate radiation dose to the tumor; the second-level optimization problem minimizes the radiation to critical healthy organs. Excessive radiation dose to healthy organs exceeding tissue tolerances is strictly prohibited in each optimization problem using constraints. The resultant optimization problems are large and nonconvex with hundreds of thousands of variables and constraints. 

To allow quick patient access and solve the optimization problems in a clinically reasonable time frame, the team leveraged multiple advanced optimization tools (e.g., convex approximation, Lagrangian multipliers, and sequential convex programming) to reduce typical computational time to 1-2 hours while also maximizing patient plan quality.

Clinical Implementation

Memorial Sloan Kettering Cancer Center has clinically implemented this complex optimization technique, internally referred to as ECHO (Expedited Constrained Hierarchical Optimization). The technique was validated before clinical implementation, with extensive retrospective comparison studies comparing automated ECHO plans with the manually generated plans produced by experienced planners. Extensive work was also done to improve the dosimetric modeling, ensuring that the optimization problem solution is close to the dose distribution as delivered by the treatment machine.

The team demonstrated that ECHO plans have a consistent high quality, and in general provide better tumor irradiation, and reduced radiation dose to the key healthy organs, and are thus expected to improve outcomes. In addition, ECHO results in improved clinical workflow and shorter times between simulation and treatment (from 5 days to 4 days).

More than 3,000 patients have benefited thus far from ECHO radiotherapy treatments. Benefits to MSKCC and their patients include: 1) a streamlined and more efficient workflow (e.g., increased capacity); 2) reduced planning effort; 3) improved plan quality; 4) greater planning consistency; and 5) expedited treatments for numerous patients in severe pain and in urgent need of treatment. The ECHO system has been commissioned clinically for several disease sites (prostate, lung, paraspinal, oligometastatic, and head and neck), and will be expanded to the great majority of all radiotherapy treatment planning at MSKCC in the next two years.

The ECHO system will ultimately be explored as a way to impact patient care more broadly, most likely with a commercial partner, including in resource-constrained countries where access to highly skilled radiotherapy planners is very limited and cost-efficient resource utilization is a must to be able to meet the cancer treatment need.