P. Després

GPUMCD: A new GPU‐oriented Monte Carlo dose calculation platform

PURPOSE Monte Carlo methods are considered as the gold standard for dosimetric computations in radiotherapy. Their execution time is, however, still an obstacle to the routine use of Monte Carlo packages in a clinical setting. To address this problem, a completely new, and designed from the ground up for the GPU, Monte Carlo dose calculation package for voxelized geometries is proposed: GPUMCD. METHOD GPUMCD implements a coupled photon-electron Monte Carlo simulation for energies in the range of 0.01-20 MeV. An analog simulation of photon interactions is used and a class II condensed history method has been implemented for the simulation of electrons. A new GPU random number generator, some divergence reduction methods, as well as other optimization strategies are also described. GPUMCD was run on a NVIDIA GTX480, while single threaded implementations of EGSnrc and DPM were run on an Intel Core i7 860. RESULTS Dosimetric results obtained with GPUMCD were compared to EGSnrc. In all but one test case, 98% or more of all significant voxels passed the gamma criteria of 2%-2 mm. In terms of execution speed and efficiency, GPUMCD is more than 900 times faster than EGSnrc and more than 200 times faster than DPM, a Monte Carlo package aiming fast executions. Absolute execution times of less than 0.3 s are found for the simulation of 1M electrons and 4M photons in water for monoenergetic beams of 15 MeV, including GPU-CPU memory transfers. CONCLUSION GPUMCD, a new GPU-oriented Monte Carlo dose calculation platform, has been compared to EGSnrc and DPM in terms of dosimetric results and execution speed. Its accuracy and speed make it an interesting solution for full Monte Carlo dose calculation in radiation oncology.

Fast convolution-superposition dose calculation on graphics hardware.

The numerical calculation of dose is central to treatment planning in radiation therapy and is at the core of optimization strategies for modern delivery techniques. In a clinical environment, dose calculation algorithms are required to be accurate and fast. The accuracy is typically achieved through the integration of patient-specific data and extensive beam modeling, which generally results in slower algorithms. In order to alleviate execution speed problems, the authors have implemented a modern dose calculation algorithm on a massively parallel hardware architecture. More specifically, they have implemented a convolution-superposition photon beam dose calculation algorithm on a commodity graphics processing unit (GPU). They have investigated a simple porting scenario as well as slightly more complex GPU optimization strategies. They have achieved speed improvement factors ranging from 10 to 20 times with GPU implementations compared to central processing unit (CPU) implementations, with higher values corresponding to larger kernel and calculation grid sizes. In all cases, they preserved the numerical accuracy of the GPU calculations with respect to the CPU calculations. These results show that streaming architectures such as GPUs can significantly accelerate dose calculation algorithms and let envision benefits for numerically intensive processes such as optimizing strategies, in particular, for complex delivery techniques such as IMRT and are therapy.

18F-FDG-PET imaging in radiotherapy tumor volume delineation in treatment of head and neck cancer

PURPOSE To determine the impact of (18)F-fluorodeoxyglucose positron emission tomography (PET) in radiotherapy target delineation and patient management for head and neck squamous cell carcinoma (HNSCC) compared to computed tomography (CT) alone. MATERIALS AND METHODS Twenty-nine patients with HNSCC were included. CT and PET/CT obtained for treatment planning purposes were reviewed respectively by a neuroradiologist and a nuclear medicine specialist who were blinded to the findings from each other. The attending radiation oncologist together with the neuroradiologist initially defined all gross tumor volume of the primary (GTVp) and the suspicious lymph nodes (GTVn) on CT. Subsequently, the same radiation oncologist and the nuclear medicine specialist defined the GTVp and GTVn on (18)F-FDG-PET/CT. Upon disagreement between CT and (18)F-FDG-PET on the status of a particular lymph node, an ultrasound-guided fine needle aspiration was performed. Volumes based on CT and (18)F-FDG-PET were compared with a paired Students t-test. RESULTS For the primary disease, four patients had previous diagnostic tonsillectomy and therefore, FDG uptake occurred in 25 patients. For these patients, GTVp contoured on (18)F-FDG-PET (GTVp-PET) were smaller than the GTVp contoured on CT (GTVp-CT) in 80% of the cases, leading to a statistically significant volume difference (p=0.001). Of the 60 lymph nodes suspicious on PET, 55 were also detected on CT. No volume change was observed (p=0.08). Ten biopsies were performed for lymph nodes that were discordant between modalities and all were of benign histology. Distant metastases were found in two patients and one had a newly diagnosed lung adenocarcinoma. CONCLUSIONS GTVp-CT was significantly larger when compared to GTVp-PET. No such change was observed for the lymph nodes. (18)F-FDG-PET modified treatment management in three patients, including two for which no curative radiotherapy was attempted. Larger multicenter studies are needed to ascertain whether combined (18)F-FDG-PET/CT in target delineation can influence the main clinical outcomes.

A convolution-superposition dose calculation engine for GPUs

PURPOSE Graphic processing units (GPUs) are increasingly used for scientific applications, where their parallel architecture and unprecedented computing power density can be exploited to accelerate calculations. In this paper, a new GPU implementation of a convolution/superposition (CS) algorithm is presented. METHODS This new GPU implementation has been designed from the ground-up to use the graphics cards strengths and to avoid its weaknesses. The CS GPU algorithm takes into account beam hardening, off-axis softening, kernel tilting, and relies heavily on raytracing through patient imaging data. Implementation details are reported as well as a multi-GPU solution. RESULTS An overall single-GPU acceleration factor of 908x was achieved when compared to a nonoptimized version of the CS algorithm implemented in PlanUNC in single threaded central processing unit (CPU) mode, resulting in approximatively 2.8 s per beam for a 3D dose computation on a 0.4 cm grid. A comparison to an established commercial system leads to an acceleration factor of approximately 29x or 0.58 versus 16.6 s per beam in single threaded mode. An acceleration factor of 46x has been obtained for the total energy released per mass (TERMA) calculation and a 943x acceleration factor for the CS calculation compared to PlanUNC. Dose distributions also have been obtained for a simple water-lung phantom to verify that the implementation gives accurate results. CONCLUSIONS These results suggest that GPUs are an attractive solution for radiation therapy applications and that careful design, taking the GPU architecture into account, is critical in obtaining significant acceleration factors. These results potentially can have a significant impact on complex dose delivery techniques requiring intensive dose calculations such as intensity-modulated radiation therapy (IMRT) and arc therapy. They also are relevant for adaptive radiation therapy where dose results must be obtained rapidly.

The Role of Computed Tomography in the Management of the Neck After Chemoradiotherapy in Patients With Head-and-Neck Cancer

PURPOSE The aim of this study was to describe the outcome in patients with head-and neck-squamous cell carcinoma (HNSCC) followed up without neck dissection (ND) after concomitant chemoradiotherapy (CRT) based on computed tomography (CT) response. The second objective was to establish CT characteristics that can predict which patients can safely avoid ND. METHODS AND MATERIALS Between 1998 and 2007, 369 patients with node-positive HNSCC were treated with primary CRT at our institution. After a clinical and a radiologic evaluation based on CT done 6 to 8 weeks after CRT, patients were labeled with a complete neck response (CR) or with a partial neck response (PR). RESULTS The median follow-up was 44 months. The number of patients presenting with N3, N2, or N1 disease were 54 (15%), 268 (72%), and 47 (13%), respectively. After CRT, 263 (71%) patients reached a CR, and 253 of them did not undergo ND. Ninety-six patients reached a PR and underwent ND. Of those, 34 (35%) had residual disease on pathologic evaluation. A regression of the diameter of ≥ 80% and a residual largest diameter of 15 mm of nodes had negative pathologic predictive values of 100% and 86%, respectively. The 3-year regional control and survival rates were not different between patients with CR who had no ND and patients with PR followed by ND. CONCLUSION Node-positive patients presenting a CR as determined by CT evaluation 6 to 8 weeks after CRT had a low rate of regional recurrence without ND. This study also suggests that lymph node residual size and percentage of regression on CT after CRT may be useful criteria to guide clinical decisions regarding neck surgery. Those results can help diminish the number of ND procedures with negative results and their associated surgical complications.

Simultaneous Integrated Boost Using Intensity-Modulated Radiotherapy Compared With Conventional Radiotherapy in Patients Treated With Concurrent Carboplatin and 5-Fluorouracil for Locally Advanced Oropharyngeal Carcinoma

PURPOSE To compare, in a retrospective study, the toxicity and efficacy of simultaneous integrated boost using intensity-modulated radiotherapy (IMRT) vs. conventional radiotherapy (CRT) in patients treated with concomitant carboplatin and 5-fluorouracil for locally advanced oropharyngeal cancer. METHODS AND MATERIALS Between January 2000 and December 2007, 249 patients were treated with definitive chemoradiation. One hundred patients had 70 Gy in 33 fractions using IMRT, and 149 received CRT at 70 Gy in 35 fractions. Overall survival, disease-free survival, and locoregional control were estimated using the Kaplan-Meier method. RESULTS Median follow-up was 42 months. Three-year actuarial rates for locoregional control, disease-free survival, and overall survival were 95.1% vs. 84.4% (p = 0.005), 85.3% vs. 69.3% (p = 0.001), and 92.1% vs. 75.2% (p < 0.001) for IMRT and CRT, respectively. The benefit of the radiotherapy regimen on outcomes was also observed with a Cox multivariate analysis. Intensity-modulated radiotherapy was associated with less acute dermatitis and less xerostomia at 6, 12, 24, and 36 months. CONCLUSIONS This study suggests that simultaneous integrated boost using IMRT is associated with favorable locoregional control and survival rates with less xerostomia and acute dermatitis than CRT when both are given concurrently with chemotherapy.

Physical characteristics of a low-dose gas microstrip detector for orthopedic x-ray imaging.

A new scanning slit gas detector dedicated to orthopedic x-ray imaging is presented and evaluated in terms of its fundamental imaging characteristics. The system is based on the micromesh gaseous structure detector and achieves primary signal amplification through electronic avalanche in the gas. This feature, together with high quantum detection efficiency and fan-beam geometry, allows for imaging at low radiation levels. The system is composed of 1764 channels spanning a width of 44.8 cm and is capable of imaging an entire patient at speeds of up to 15 cm/s. The resolution was found to be anisotropic and significantly affected by the beam quality in the horizontal direction, but otherwise sufficient for orthopedic studies. As a consequence of line-by-line acquisition, the images contain some ripple components due to mechanical vibrations combined with variations in the x-ray tube output power. The reported detective quantum efficiency (DQE) values are relatively low (0.14 to 0.20 at 0.5 mm(-1)) as a consequence of a suboptimal collimation geometry. The DQE values were found to be unaffected by the exposure down to 7 microGy, suggesting that the system is quantum limited even for low radiation levels. A system composed of two orthogonal detectors is already in use and can produce dual-view full body scans at low doses. This device could contribute to reduce the risk of radiation induced cancer in sensitive clientele undergoing intensive x-ray procedures, like young scoliotic women.

Cervical Lymph Node Metastases From Unknown Primary Cancer: A Single-Institution Experience With Intensity-Modulated Radiotherapy

PURPOSE To determine the effectiveness and rate of complications of intensity-modulated radiotherapy (IMRT) in the treatment of cervical lymph node metastases from unknown primary cancer. METHODS AND MATERIALS Between February 2005 and November 2008, 25 patients with an unknown primary cancer underwent IMRT, with a median radiation dose of 70 Gy. The bilateral neck and ipsilateral putative pharyngeal mucosa were included in the target volume. All patients had squamous cell carcinoma, except for 1 patient who had adenosquamous differentiation. They were all treated with curative intent. Of the 25 included patients, 20 were men and 5 were women, with a median age of 54 years. Of these patients, 3 had Stage III, 18 had Stage IVa, and 4 had Stage IVb. Of the 25 patients, 18 (72%) received platinum-based chemotherapy in a combined-modality setting. Neck dissection was reserved for residual disease after definitive IMRT. Overall survival, disease-free survival, and locoregional control were calculated using the Kaplan-Meier method. RESULTS With a median follow-up of 38 months, the overall survival, disease-free survival, and locoregional control rates were all 100% at 3 years. No occurrence of primary cancer was observed during the follow-up period. The reported rates of xerostomia reduced with the interval from the completion of treatment. Nine patients (36%) reported Grade 2 or greater xerostomia at 6 months, and only 2 (8%) of them reported the same grade of salivary function toxicity after 24 months of follow-up. CONCLUSION In our institution, IMRT for unknown primary cancer has provided good overall and disease-free survival in all the patients with an acceptable rate of complications. IMRT allowed us to address the bilateral neck and ipsilateral putative pharyngeal mucosa with minimal late salivary function toxicity. The use of concurrent chemotherapy and IMRT for more advanced disease led to good clinical results with reasonable toxicities.

Real-time processing in dynamic ultrasound elastography: A GPU-based implementation using CUDA

This paper addresses the computational cost of the normalized cross-correlation (NCC) algorithm in ultrasound elastography. Parallel implementations of the NCC algorithm based on multicore architectures and a graphical processor unit (GPU) are formulated and applied to radio-frequency (RF) data from dynamic elastography experiments. Compared to single computer processor unit (CPU) performances, results show that parallel implementation of the NCC algorithm allows speedups of less than 5 for multi-threaded execution on CPU and up to 85 using a GPU. Processing frame rates from 80 to 173 sec-1 have been achieved for large fields of view with good spatial resolution. The trade-off between accuracy, spatial resolution and computational cost in displacement estimation using the NCC algorithm therefore appears obsolete. Open source codes for implementing the NCC algorithm on GPU are made available at www.lbum-crchum.com.

The importance of an exponential prostate-specific antigen decline after external beam radiotherapy for intermediate risk prostate cancer

BACKGROUND To study the influence of an exponential prostate-specific antigen (PSA) decline on biochemical failure after external-beam radiotherapy (EBRT). METHODS We analyzed 114 patients with intermediate risk prostate cancer (Gleason≤6 and PSA 10-20 or Gleason 7 and PSA <10). Patients were randomized between EBRT doses of either 70.2 Gy or 79.2 Gy (1.8 Gy per day). All patients had a follow up of at least six PSA measurements post-EBRT. Exponential decline and PSA half life were included in a Cox regression analysis for factors associated with biochemical failure. RESULTS A total of 80/114 (70.2%) patterns were classified as having an exponential PSA decline. Both exponential decline (HR 0.115, 95%CI 0.03-0.44, p=0.0016) and PSA half life ratio were statistically significant predictors (HR 1.03 (95% CI 1.01-1.06)) of biochemical failure. In the model predicting for exponential decline, none of the factors were significant. CONCLUSION Patients with an exponential PSA decline show a better biochemical outcome in the long term.