While computer simulations can accurately model light dose delivery in interstitial photodynamic therapy (iPDT), predicting individual clinical outcomes remains difficult. In addition to biological uncertainties, inaccuracies in light delivery must be considered. Using simulations on virtual brain tumour (glioblastoma) models, we analyze two sources of uncertainty: light source power variations of \(\pm 5\%\) , \(\pm 10\%\) , and \(\pm 20\%\) , and positional deviations during source insertion, modelled as angular errors producing up to \(3\textrm{mm}\) displacement. Simulated outcomes show minimal impact from power uncertainty, even at worst-case \(\pm 20\%\) : the percent difference between maximum and minimum \(v_{100}\) does not exceed \(9\%\) , with tumour coverage only dropping from the targeted \(98\%\) to \(96.9\%\) . Using a new power-uncertainty–aware option in the PDT-SPACE planning tool improves the worst-case minimum coverage from \(96.9\%\) to \(97.3\%\) , eliminating the risk of under-treating. Position uncertainty was simulated by discretizing the space and randomizing source placements, showing a larger negative effect. Power re-optimization on measured post-insertion positions restores tumour coverage to \(98\%\) , while PDT-SPACE source-position optimization reduces average healthy tissue damage by \(36\%\) . Combining both yields the most robust performance and minimizes sensitivity to positional deviations, thereby limiting light-delivery errors in iPDT.
