Abstract
To deliver uniform photodynamic dose to the prostate gland, it is necessary to develop algorithms that optimize the location and strength (emitted power × illumination time) of each light source. Since tissue optical properties may change with time, rapid (almost real-time) optimization is desirable. We use the Cimmino algorithm because it is fast, linear, and always converges reliably. A phase I motexafin lutetium (MLu)-mediated photodynamic therapy (PDT) protocol is on-going at the University of Pennsylvania. The standard plan for the protocol uses equal source strength and equal spaced loading (1-cm). PDT for the prostate is performed with cylindrical diffusing fibers (CDF) of various lengths inserted to longitudinal coverage within the matrix of parallel catheters perpendicular to a base plate. We developed several search procedures to aid the user in choosing the positions, lengths, and intensities of the CDFs. The Cimmino algorithm is used in these procedures to optimize the strengths of the light catheters at each step of the iterative selection process. Maximum and minimum bounds on allowed doses to points in four volumes (prostate, urethra, rectum, and background) constrain the solutions for the strengths of the linear light sources. Uniform optical properties are assumed. To study how different opacities of the prostate would affect optimization, optical kernels of different light penetration were used. Another goal is to see whether the urethra and rectum can be spared, with minimal effect on PTV treatment delivery, by manipulating light illumination times of the sources. Importance weights are chosen beforehand for organ volumes, and normalized. Compared with the standard plan, our algorithm is shown to produce a plan that better spares the urethra and rectum and is very fast. Thus the combined selection of positions, lengths, and strengths of interstitial light sources improves outcome.