PDT Treatment Planning using the Terasic DE3 Stratix III FPGA
Nov. 10, 2009
Photodynamic Therapy (PDT) is a rapidly growing area of medical treatment that started in the late 1980s emerging as a practical medical technology used at several institutions throughout the world. PDT is currently used in a number of medical fields including oncology (cancer), dermatology (skin), and cosmetic surgery. PDFT is a medical treatment that involves the use of light-activated dyes (photosensitizers) that localize in target cells producing an activated oxygen module that can destroy nearby cells.
To improve the PDT success, particularly for interstitial applications, faster computational tools is required to enable efficient treatment planning. The PDT treatment planning depends on the light dosimetry which is distributed using the Monte Carlo (MC) method. The MC method is used extensively in the field of medical biophysics, particularly for modeling light propagation in tissues. Unfortunately, the high computation time for MC limits optimization problems such as PDT treatment planning which meant a faster means for performing MC simulations would accelerate MC-based light dose computation. The Terasic DE3 development board with four Stratix I FPGAs was chosen to build the MC simulation for PDT that is based on the Monte Carlo for Multi-Layered media (MCML) code. Using the MCML program as the custom pipelined hardware designed on a FPGA platform, an 80 times speedup was achieved compared to a 3-GHz Intel Xeon Processor. The FPGAs offer greater flexibility in the design as future modification can be readily implemented due to the use of modularized pipelined architecture.
The limitations of the current prototype design are removed as the newer FPGA platforms can be relied upon that offer more on-chip memory and other resources. In particular the Terasic DE3 board with the Stratix III FPGA, it was able to project 240 times speedup from the current design.
For more information on the Photodynamic Therapy (PDT) report, please refer to the following link: