Investigation of using a power function as a cost function in inverse planning optimization

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Biology

Background:

• Inverse planning optimization in radiation therapy relies on cost functions to guide treatment plan creation.
• Classical quadratic cost functions are commonly used but may have limitations in achieving optimal dose distributions.
• Exploring alternative cost functions, such as power functions, can potentially improve treatment planning outcomes.

Purpose of the Study:

• To investigate the efficacy of a power function, specifically an exponential power function, as a cost function in inverse planning optimization.
• To compare the performance of power functions with varying exponents against the traditional quadratic cost function (exponent of 2).

Main Methods:

• Developed an independent optimization module interfaced with a research treatment planning system.
• Implemented cost functions as exponential power functions of dose deviation for various structures.
• Tested three clinical cases using different exponent settings for tumor targets and sensitive structures.
• Evaluated treatment plans using dose-volume histograms (DVHs).

Main Results:

• Employing exponents higher than 2 in the cost function for the target significantly improved dose homogeneity compared to an exponent of 2.
• Exponents greater than 2 for serial sensitive structures effectively reduced the maximum dose.
• The most substantial improvements in DVHs were observed when varying exponents from 2 to 4, with diminishing returns between 4 and 8, indicating saturation.

Conclusions:

• A power function with exponents greater than 2 serves as an effective cost function in inverse planning optimization.
• This approach can achieve superior dose homogeneity within the target volume.
• It also effectively minimizes the maximum dose delivered to critical structures, offering potential clinical benefits.