Georgios Kalantzis

Investigations of a GPU-based levy-firefly algorithm for constrained optimization of radiation therapy treatment planning

Abstract Intensity modulated radiation therapy (IMRT) affords the potential to decrease radiation therapy associated toxicity by creating highly conformal dose distribution to tumor. Inverse optimization of IMRT treatment plans is often a time intensive task due to the large scale solution space, and the indubitably complexity of the task. Furthermore, the incorporation of conflicting dose constraints in the treatment plan, usually introduces an additional degree of intricacy. Metaheuristic algorithms have been proposed in the past for global optimization in IMRT treatment planning. However one disadvantage of the aforementioned methods is their extensive computational cost. One way to ameliorate their performance deficiency is to parallelize the application. In the current study we propose a GPU-based levy-firefly algorithm (LFA) for constrained optimization of IMRT treatment planning. The evaluation of our method was realized for two treatment cases: a prostate and a head and neck (H&N) cancer IMRT plans. The studies indicated an ascendable increase of the speedup factor as a function of the number of pencil beams with a maximum of ~11, whereas the performance of the algorithm was decreasing as a function of the population of the swarm particles. In addition, from our simulation results we concluded that 200 fireflies were sufficient for the algorithm to converge in less than 80 iterations. Finally, we demonstrated the effect of penalizing factors on constraining the maximum dose at the organs at risk (OAR) by impeding the dose coverage of the tumor target. The impetus behind our study was to elucidate the performance and generic attributes of the proposed algorithm, as well as the potential of its applicability for IMRT optimization problems.

Dosimetric and radiobiological comparison of CyberKnife M6TM InCise multileaf collimator over IRISTM variable collimator in prostate stereotactic body radiation therapy

The impetus behind our study was to establish a quantitative comparison between the IRIS collimator and the InCise multileaf collimator (MLC) (Accuray Inc. Synnyvale, CA) for prostate stereotactic body radiation therapy (SBRT). Treatment plans for ten prostate cancer patients were performed on MultiPlan™ 5.1.2 treatment planning system utilizing MLC and IRIS for 36.25 Gy in five fractions. To reduce the magnitude of variations between cases, the planning tumor volume (PTV) was defined and outlined for treating prostate gland only, assuming no seminal vesicle or ex-capsule involvement. Evaluation indices of each plan include PTV coverage, conformity index (CI), Paddicks new CI, homogeneity index, and gradient index. Organ at risk (OAR) dose sparing was analyzed by the bladder wall Dmaxand V37Gy, rectum Dmaxand V36Gy. The radiobiological response was evaluated by tumor control probability and normal tissue complication probability based on equivalent uniform dose. The dose delivery efficiency was evaluated on the basis of planned monitor units (MUs) and the reported treatment time per fraction. Statistical significance was tested using the Wilcoxon signed rank test. The studies indicated that CyberKnife M6™ IRIS and InCise™ MLC produce equivalent SBRT prostate treatment plans in terms of dosimetry, radiobiology, and OAR sparing, except that the MLC plans offer improvement of the dose fall-off gradient by 29% over IRIS. The main advantage of replacing the IRIS collimator with MLC is the improved efficiency, determined from the reduction of MUs by 42%, and a 36% faster delivery time.

A self-tuned bat algorithm for optimization in radiation therapy treatment planning

The performance of any optimization algorithm largely depends on the setting of its algorithm-dependent parameters. Swarm intelligence algorithms are popular methods in optimization since they have been proved very efficient. One drawback of those methods though, is that the appropriate setting of the algorithm-dependent parameters has a significant impact on the algorithms performance. The “parameter tuning” of an algorithm in such a way to be able to find the optimal solution by using the minimum number of iterations, quite often is a difficult and time consuming task depending on the optimization problem. Essentially this is a hyper-optimization problem, that is, the optimization of the optimization algorithm. In this paper, a novel self-tuned metaheuristic algorithm is presented for optimization in radiation therapy treatment planning. The proposed Self-Tuned Bat Algorithm (STBA) finds itself the optimal set of algorithm-dependent parameters and therefore minimizes the number of iterations required for the optimization to reach sub-optimal solution. The applicability of the proposed algorithm is demonstrated in the optimization of a prostate case using intensity modulation radiation therapy (IMRT).

A GPU Accelerated Finite Differences Method of the Bioheat Transfer Equation for Ultrasound Thermal Ablation

Over the years, high intensity focused ultrasound (FUS) therapy has become a promising therapeutic alternative for non-invasive tumor treatment. The basic idea of FUS therapy is the elevation of the tissue temperature by the application of focused ultrasound beams to focal spot in the tumor. Biothermal modeling is utilized to predict dynamic temperature distributions generated and altered by the therapeutic heating modality, tissue energy storage and dissipation, and blood flow. Implementation of biothermal modeling in the planning, monitoring, control and evaluation of MR guided Focused Ultrasound (MRgFUS) therapies can help to minimize treatment time, maximize efficacy, and ensure the safety of healthy normal tissues, while increasing clinical confidence in MRgFUS treatments. Fast calculations of thermal doses can support in planning, conduction, and monitoring of such treatments. In the current study a GPU-based method in Matlab is proposed, for fast calculations of the temperature and cumulative equivalent minutes at 43° (CEM 43°) based on the bioheat equation. The performance of our proposed method was assessed with three GPUs (GTX 750, GTX 770 and Tesla C2050) for five grid sizes. The maximum speedup was achieved with the Tesla C2050 (~29) while GTX 750 demonstrated the lower performance (~15).

A computational study on different penalty approaches for constrained optimization in radiation therapy treatment planning with a simulated annealing algorithm

In intensity modulated radiation therapy (IMRT) a treatment plan is a high dimensionality optimization problem with the goal to give the prescribed radiation dose to the Planning Target Volume (PTV) while sparing critical organs. A clinically acceptable plan is usually generated by a numerical optimization process in pursuit of attaining the above mentioned goal. Incorporation of dose volume constraints (DVCs) for the OARs introduce an additional degree of impediment to the optimization task. Heuristic algorithms have been ascertained in the past as a powerful tool for various problems in radiation therapy, such as beam angle optimization (BAO) and IMRT treatment planning. Simulated Annealing (SA) algorithm has the capability to find global minima for bound-constrained optimization problems. However, its performance depends on the penalty method which is utilized for the constraints. In the current study we investigate the performance of three penalty methods for IMRT treatment planning for five prostate cases with dose volume constraints. In addition, sensitivity analysis was performed in order to study the effect of the penalty variables on the treatment plan. Finally the merits and demerits of each method have been discussed.

SU-E-T-37: A GPU-Based Pencil Beam Algorithm for Dose Calculations in Proton Radiation Therapy

Purpose: Recent developments in radiation therapy have been focused on applications of charged particles, especially protons. Over the years several dose calculation methods have been proposed in proton therapy. A common characteristic of all these methods is their extensive computational burden. In the current study we present for the first time, to our best knowledge, a GPU-based PBA for proton dose calculations in Matlab. Methods: In the current study we employed an analytical expression for the protons depth dose distribution. The central-axis term is taken from the broad-beam central-axis depth dose in water modified by an inverse square correction while the distribution of the off-axis term was considered Gaussian. The serial code was implemented in MATLAB and was launched on a desktop with a quad core Intel Xeon X5550 at 2.67GHz with 8 GB of RAM. For the parallelization on the GPU, the parallel computing toolbox was employed and the code was launched on a GTX 770 with Kepler architecture. The performance comparison was established on the speedup factors. Results: The performance of the GPU code was evaluated for three different energies: low (50 MeV), medium (100 MeV) and high (150 MeV). Four square fields were selected for each energy, and the dose calculations were performed with both the serial and parallel codes for a homogeneous water phantom with size 300×300×300 mm3. The resolution of the PBs was set to 1.0 mm. The maximum speedup of ∼127 was achieved for the highest energy and the largest field size. Conclusion: A GPU-based PB algorithm for proton dose calculations in Matlab was presented. A maximum speedup of ∼127 was achieved. Future directions of the current work include extension of our method for dose calculation in heterogeneous phantoms.

A computational tool for patient specific dosimetry and radiobiological modeling of selective internal radiation therapy with 90Y microspheres

In recent years we have witnessed tremendous progress in selective internal radiation therapy. In clinical practice, quite often, radionuclide therapy is planned using simple models based on standard activity values or activity administered per unit body weight or surface area in spite of the admission that radiation-dose methods provide more accurate dosimetric results. To address that issue, the authors developed a Matlab-based computational software, named Patient Specific Yttrium-90 Dosimetry Toolkit (PSYDT). PSYDT was designed for patient specific voxel-based dosimetric calculations and radiobiological modeling of selective internal radiation therapy with (90)Y microspheres. The developed toolkit is composed of three dimensional dose calculations for both bremsstrahlung and beta emissions. Subsequently, radiobiological modeling is performed on a per-voxel basis and cumulative dose volume histograms (DVHs) are generated. In this report we describe the functionality and visualization features of PSYDT.

A parallelized GPU-based simulating annealing algorithm for intensity modulated radiation therapy optimization

Intensity modulated radiation therapy (IMRT) exhibits the ability to deliver the prescribed dose to the planning target volume (PTV), while minimizing the delivered dose to the organs at risk (OARs). Metaheuristic algorithms, among them the simulating annealing algorithm (SAA), have been proposed in the past for optimization of IMRT. Despite the advantage of the SAA to be a global optimizer, IMRT optimization is an extensive computational task due to the large scale of the optimization variables. Therefore stochastic algorithms, such as the SAA, require significant computational resources. In an effort to elucidate the performance improvement of the SAA in highly dimensional optimization tasks, such as the IMRT optimization, we introduce for the first time to our best knowledge a parallel graphic processing unit (GPU)-based SAA developed in MATLAB platform and compliant with the computational environment for radiotherapy research (CERR) for IMRT treatment planning. Our strategy was firstly to identify the major “bottlenecks” of our code and secondly to parallelize those on the GPU accordingly. Performance tests were conducted on four different GPU cards in comparison to a serial version of the algorithm executed on a CPU. Our studies have shown a gradual increase of the speedup factor as a function of the number of beamlets for all four GPUs. Particularly, a maximum speedup factor of ∼33 was achieved when the K40m card was utilized.

A GPU-Based Pencil Beam Algorithm for Dose Calculations in Proton Radiation Therapy

Recent developments in radiation therapy have been focused on applications of charged particles, especially—protons. Proton therapy can allow higher dose conformality compared to conventional radiation therapy. Dose calculations have an integral role in the successful application of proton therapy. Over the years several dose calculation methods have been proposed in proton therapy. A common characteristic of all these methods is their extensive computational burden. One way to ameliorate that issue is the parallelization of the algorithm. Graphics processing units (GPUs) have recently been employed to accelerate the proton dose calculation process. Pencil-beam dose calculation algorithms for proton therapy have been widely utilized in clinical routine for treatment planning purposes in most clinical settings, due to their simplicity of calculation scheme and acceptable accuracy. In the current study a GPU-based pencil beam algorithm for dose calculations with protons is proposed. The studies indicated a maximum speedup factor of \(\sim \)127 in a homogeneous phantom.