The advantages of a flat-panel X-ray source(FPXS)make it a promising candidate for imaging applications.Accurate imaging-system modeling and projection simulation are critical for analyzing imaging performance and res...The advantages of a flat-panel X-ray source(FPXS)make it a promising candidate for imaging applications.Accurate imaging-system modeling and projection simulation are critical for analyzing imaging performance and resolving overlapping projection issues in FPXS.The conventional analytical ray-tracing approach is limited by the number of patterns and is not applicable to FPXS-projection calculations.However,the computation time of Monte Carlo(MC)simulation is independent of the size of the patterned arrays in FPXS.This study proposes two high-efficiency MC projection simulators for FPXS:a graphics processing unit(GPU)-based phase-space sampling MC(gPSMC)simulator and GPU-based fluence sampling MC(gFSMC)simulator.The two simulators comprise three components:imaging-system modeling,photon initialization,and physical-interaction simulations in the phantom.Imaging-system modeling was performed by modeling the FPXS,imaging geometry,and detector.The gPSMC simulator samples the initial photons from the phase space,whereas the gFSMC simulator performs photon initialization from the calculated energy spectrum and fluence map.The entire process of photon interaction with the geometry and arrival at the detector was simulated in parallel using multiple GPU kernels,and projections based on the two simulators were calculated.The accuracies of the two simulators were evaluated by comparing them with the conventional analytical ray-tracing approach and acquired projections,and the efficiencies were evaluated by comparing the computation time.The results of simulated and realistic experiments illustrate the accuracy and efficiency of the proposed gPSMC and gFSMC simulators in the projection calculation of various phantoms.展开更多
We discuss the concept of typicality of quantum states at quantum-critical points, using projector Monte Carlo simu- lations of an S ---- 1/2 bilayer Heisenberg antiferromagnet as an illustration. With the projection ...We discuss the concept of typicality of quantum states at quantum-critical points, using projector Monte Carlo simu- lations of an S ---- 1/2 bilayer Heisenberg antiferromagnet as an illustration. With the projection （imaginary） time t scaled as t= aLz, L being the system length and z the dynamic critical exponent （which takes the value z = 1 in the bilayer model studied here）, a critical point can be identified which asymptotically flows to the correct location and universality class with increasing L, independently of the prefactor a and the initial state. Varying the proportionality factor a and the initial state only changes the cross-over behavior into the asymptotic large-L behavior. In some cases, choosing an optimal factor a may also lead to the vanishing of the leading finite-size corrections. The observation of typicality can be used to speed up simulations of quantum criticality, not only within the Monte Carlo approach but also with other numerical methods where imaginary-time evolution is employed, e.g., tensor network states, as it is not necessary to evolve fully to the ground state but only for sufficiently long times to reach the typicality regime.展开更多
文摘The advantages of a flat-panel X-ray source(FPXS)make it a promising candidate for imaging applications.Accurate imaging-system modeling and projection simulation are critical for analyzing imaging performance and resolving overlapping projection issues in FPXS.The conventional analytical ray-tracing approach is limited by the number of patterns and is not applicable to FPXS-projection calculations.However,the computation time of Monte Carlo(MC)simulation is independent of the size of the patterned arrays in FPXS.This study proposes two high-efficiency MC projection simulators for FPXS:a graphics processing unit(GPU)-based phase-space sampling MC(gPSMC)simulator and GPU-based fluence sampling MC(gFSMC)simulator.The two simulators comprise three components:imaging-system modeling,photon initialization,and physical-interaction simulations in the phantom.Imaging-system modeling was performed by modeling the FPXS,imaging geometry,and detector.The gPSMC simulator samples the initial photons from the phase space,whereas the gFSMC simulator performs photon initialization from the calculated energy spectrum and fluence map.The entire process of photon interaction with the geometry and arrival at the detector was simulated in parallel using multiple GPU kernels,and projections based on the two simulators were calculated.The accuracies of the two simulators were evaluated by comparing them with the conventional analytical ray-tracing approach and acquired projections,and the efficiencies were evaluated by comparing the computation time.The results of simulated and realistic experiments illustrate the accuracy and efficiency of the proposed gPSMC and gFSMC simulators in the projection calculation of various phantoms.
基金Project supported by the National Natural Science Foundation of China(Grant Nos.11734002 and 11775021)the National Science Foundation(Grant No.DMR-1710170)a Simons Investigator Award
文摘We discuss the concept of typicality of quantum states at quantum-critical points, using projector Monte Carlo simu- lations of an S ---- 1/2 bilayer Heisenberg antiferromagnet as an illustration. With the projection （imaginary） time t scaled as t= aLz, L being the system length and z the dynamic critical exponent （which takes the value z = 1 in the bilayer model studied here）, a critical point can be identified which asymptotically flows to the correct location and universality class with increasing L, independently of the prefactor a and the initial state. Varying the proportionality factor a and the initial state only changes the cross-over behavior into the asymptotic large-L behavior. In some cases, choosing an optimal factor a may also lead to the vanishing of the leading finite-size corrections. The observation of typicality can be used to speed up simulations of quantum criticality, not only within the Monte Carlo approach but also with other numerical methods where imaginary-time evolution is employed, e.g., tensor network states, as it is not necessary to evolve fully to the ground state but only for sufficiently long times to reach the typicality regime.
