An approach for estimating mean excitation energy and stopping power ratio for proton therapy

A method that uses machine learning idea was developed for computing mean excitation energy (I) and proton stopping power ratio (SPR) using relative electron density (ρe ). The designed method was applied to theoretical ρe, as well as ρe computed on image domain. The method provided a good estimate for the parameters; specifically in the case of proton stopping power ratio which has total root mean square training error that is less or equal to 0.32% and total root mean square testing error less or equal to 0.92% when theoretical ρe was used, and total modeling root mean square error as low as 0.22% and total testing root mean square error as low as 0.86% for human tissues used when image domain estimated ρe was used. This will improve proton therapy for cancer treatment by improving the precision in killing cancer cells and sparing healthy ones.

To be updated shortly..