Reduced-order constrained optimization (ROCO): clinical application to head-and-neck IMRT

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Optimization

Background:

• Intensity-modulated radiation therapy (IMRT) planning for head-and-neck (H&N) cancer is complex due to numerous organs-at-risk (OARs) and varying prescription levels.
• Traditional IMRT planning involves extensive trial-and-error, leading to significant time investment by expert planners.
• Previous applications of the reduced order constrained optimization (ROCO) method have shown success in prostate and lung cancer treatment planning.

Purpose of the Study:

• To apply and evaluate the reduced order constrained optimization (ROCO) method for head-and-neck (H&N) intensity-modulated radiation therapy (IMRT) planning.
• To demonstrate the method's ability to quickly and automatically generate clinically competitive IMRT plans for H&N sites.
• To provide guidelines for ROCO application in larynx, oropharynx, and nasopharynx cases and assess time savings in a live experiment.

Main Methods:

• The ROCO method involves sampling the treatment plan intensity space via unconstrained optimization.
• Dimensionality reduction using principal component analysis (PCA) identifies dominant modes in the intensity space.
• A constrained optimization problem is solved over this reduced basis to generate IMRT plans meeting OAR and target coverage constraints.

Main Results:

• Ten H&N IMRT plans generated by ROCO were quantitatively compared to existing clinical plans and found to be clinically acceptable.
• ROCO H&N plans require approximately 1.6 hours of offline computation and 5-15 minutes of semi-automatic planning time.
• A live experiment showed ROCO reduced planning time to 135 minutes, compared to an estimated 4 hours using traditional methods, with an expert deeming the plan clinically acceptable and potentially superior.

Conclusions:

• The ROCO method is a promising approach for head-and-neck IMRT planning, addressing the site's inherent complexities.
• ROCO facilitates the generation of clinically acceptable IMRT plans more rapidly and with substantially reduced human interaction.
• The method offers significant time savings and efficiency gains for expert radiation oncology planners.