FullMonte: a framework for high-performance Monte Carlo simulation of light through turbid media with complex geometry

Emerging clinical applications including bioluminescence imaging require fast and accurate modelling of light propagation through turbid media with complex geometries. Monte Carlo simulations are widely recognized as the standard for high-quality modelling of light propagation in turbid media, albeit with high computational requirements. We present FullMonte: a flexible, extensible software framework for Monte Carlo modelling of light transport from extended sources through general 3D turbid media including anisotropic scattering and refractive index changes. The problem geometry is expressed using a tetrahedral mesh, giving accurate surface normals and avoiding artifacts introduced by voxel approaches. The software uses multithreading, Intel SSE vector instructions, and optimized data structures. It incorporates novel hardware-friendly performance optimizations that are also useful for software implementations. Results and performance are compared against existing implementations. We present a discussion of current state-of-the-art algorithms and accelerated implementations of the modelling problem. A new parameter permitting accuracy-performance tradeoffs is also shown which has significant implications including performance gains of over 25% for real applications. The advantages and limitations of both CPU and GPU implementations are discussed, with observations important to future advances. We also point the way towards custom hardware implementations with potentially large gains in performance and energy efficiency.

[1]  L Wang,et al.  MCML--Monte Carlo modeling of light transport in multi-layered tissues. , 1995, Computer methods and programs in biomedicine.

[2]  David A Boas,et al.  Monte Carlo simulation of photon migration in 3D turbid media accelerated by graphics processing units. , 2009, Optics express.

[3]  Daniel A. Rüfenacht,et al.  Light transport in tissue by 3D Monte Carlo: Influence of boundary voxelization , 2008, Comput. Methods Programs Biomed..

[4]  D. Boas,et al.  Three dimensional Monte Carlo code for photon migration through complex heterogeneous media including the adult human head. , 2002, Optics express.

[5]  Haiou Shen,et al.  A tetrahedron-based inhomogeneous Monte Carlo optical simulator , 2010, Physics in medicine and biology.

[6]  Hamid Dehghani,et al.  Near infrared optical tomography using NIRFAST: Algorithm for numerical model and image reconstruction. , 2009, Communications in numerical methods in engineering.

[7]  Qianqian Fang,et al.  Mesh-based Monte Carlo method using fast ray-tracing in Plücker coordinates , 2010, Biomedical optics express.

[8]  Lothar Lilge,et al.  Hardware acceleration of a Monte Carlo simulation for photodynamic therapy [corrected] treatment planning. , 2009, Journal of biomedical optics.

[9]  David R. Kaeli,et al.  Accelerating mesh-based Monte Carlo method on modern CPU architectures , 2012, Biomedical optics express.

[10]  Tomas Svensson,et al.  Parallel computing with graphics processing units for high-speed Monte Carlo simulation of photon migration. , 2008, Journal of biomedical optics.

[11]  Makoto Matsumoto,et al.  SIMD-Oriented Fast Mersenne Twister: a 128-bit Pseudorandom Number Generator , 2008 .

[12]  R. Leahy,et al.  Digimouse: a 3D whole body mouse atlas from CT and cryosection data , 2007, Physics in medicine and biology.