Engineering Papers⌕ Search

Engineering topics

Liu, H

Publications and source records attributed to Liu, H.

At least 19 records

Scalable Deep Learning-Based Microarchitecture Simulation on GPUs

Cycle-accurate microarchitecture simulators are essential tools for designers to architect, estimate, optimize, and manufacture new processors that meet specific design expectations. However, conventional simulators based on discrete-event methods often require an exceedingly long time-to-solution for the simulation of applications and architectures at full complexity and scale. Given the excitement around wielding the machine learning (ML) hammer to tackle various architecture problems, there have been attempts to employ ML to perform architecture simulations, such as Ithemal and SimNet. However, the direct application of existing ML approaches to architecture simulation may be even slower due to overwhelming memory traffic and stringent sequential computation logic. This work proposes the first graphics processing unit (GPU)-based microarchitecture simulator that fully unleashes the potential of GPUs to accelerate state-of-the-art ML-based simulators. First, considering the application traces are loaded from central processing unit (CPU) to GPU for simulation, we introduce various designs to reduce the data movement cost between CPUs and GPUs. Second, we propose a parallel simulation paradigm that partitions the application trace into sub-traces to simulate them in parallel with rigorous error analysis and effective error correction mechanisms. Combined, this scalable GPU-based simulator outperforms by orders of magnitude the traditional CPU-based simulators and the state-of-the-art ML-based simulators, i.e., SimNet and Ithemal.

97 MATHEMATICS AND COMPUTING↗

C-SAW: a framework for graph sampling and random walk on GPUs

Many applications require to learn, mine, analyze and visualize large-scale graphs. These graphs are often too large to be addressed efficiently using conventional graph processing technologies. Fortunately, recent research efforts find out graph sampling and random walk, which significantly reduce the size of original graphs, can benefit the tasks of learning, mining, analyzing and visualizing large graphs by capturing the desirable graph properties. This paper introduces C-SAW, the first framework that accelerates Sampling and Random Walk framework on GPUs. Particularly, C-SAW makes three contributions: First, our framework provides a generic API which allows users to implement a wide range of sampling and random walk algorithms with ease. Second, offloading this framework on GPU, we introduce warp-centric parallel selection, and two novel optimizations for collision migration. Third, towards supporting graphs that exceed the GPU memory capacity, we introduce efficient data transfer optimizations for out-of-memory and multi-GPU sampling, such as workload-aware scheduling and batched multi-instance sampling. Taken together, our framework constantly outperforms the state of the art projects in addition to the capability of supporting a wide range of sampling and random walk algorithms.

97 MATHEMATICS AND COMPUTING↗

SU-D-209-06: Study On the Dose Conversion Coefficients in Pediatric Radiography with the Development of Children Voxel Phantoms

Purpose: Conversion coefficients of organ dose normalized to entrance skin dose (ESD) are widely used to evaluate the organ doses directly using ESD without time-consuming dose measurement, this work aims to investigate the dose conversion coefficients in pediatric chest and abdomen radiography with the development of 5 years and 10 years old children voxel phantoms. Methods: After segmentation of organs and tissues from CT slice images of ATOM tissue-equivalent phantoms, a 5-year-old and a 10-year-old children computational voxel phantoms were developed for Monte Carlo simulation. The organ doses and the entrance skin dose for pediatric chest postero-anterior projection and abdominal antero-posterior projection were simulated at the same time, and then the organ dose conversion coefficients were calculated.To verify the simulated results, dose measurement was carried out with ATOM tissue-equivalent phantoms for 5 year chest radiography. Results: Simulated results and experimental results matched very well with each other, the result differences of all the organs covered in radiation field were below 16% for 5-year-old child in chest projection. I showed that the conversion coefficients of organs covered in the radiation field were much larger than organs out of the field for all the study cases, for example, the conversion coefficients of stomach, liver intestines, and pancreas are larger for abdomen radiography while conversion coefficients of lungs are larger for chest radiography. Conclusion: The voxel children phantoms were helpful to evaluate the radiation doses more accurately and efficiently. Radiation field was the essential factor that affects the organ dose, use reasonably small field should be encouraged for radiation protection. This work was supported by the National Natural Science Foundation of China(11475047)

60 APPLIED LIFE SCIENCES↗

SU-F-E-03: PET/CT Guided Dose Boost to Hypoxic Sub-Volume in Nasopharyngeal Carcinomas Using Self-Optimizing Non-Uniform VMAT

Purpose: This study is to evaluate the feasibility of simultaneously integrated boost (SIB) to hypoxic subvolume (HTV) in nasopharyngeal carcinomas under the guidance of 18F-Fluoromisonidazole (FMISO) PET/CT using a novel non-uniform volumetric modulated arc therapy (VMAT)technique. Methods: Eight nasopharyngeal carcinoma patients treated with conventional uniform VMAT were retrospectively analyzed. For each treatment, actual conventional uniform VMAT plan with two or more arcs (2–2.5 arcs, totally rotating angle < 1000o) was designed with dose boost to hopxic subvolume (total dose, 84Gy) in the gross tumor volme (GTV) under the guidance of 18F- FMISO PET/CT. Based on the same dataset, experimental single arc non-uniform VAMT plans were generated with the same dose prescription using customized software tools. Dosimetric parameters, quality assurance and the efficiency of the treatment delivery were compared between the uniform and non-uniform VMAT plans. Results: To develop the non-uniform VMAT technique, a specific optimization model was successfully established. Both techniques generate high-quality plans with pass rate (>98%) with the 3mm, 3% criterion. HTV received dose of 84.1±0.75Gy and 84.1±1.2Gy from uniform and non-uniform VMAT plans, respectively. In terms of target coverage and dose homogeneity, there was no significant statistical difference between actual and experimental plans for each case. However, for critical organs at risk (OAR), including the parotids, oral cavity and larynx, dosimetric difference was significant with better dose sparing form experimental plans. Regarding plan implementation efficiency, the average machine time was 3.5 minutes for the actual VMAT plans and 3.7 minutes for the experimental nonuniform VMAT plans (p>0.050). Conclusion: Compared to conventional VMAT technique, the proposed non-uniform VMAT technique has the potential to produce efficient and safe treatment plans, especially in cases with complicated anatomical structures and demanding dose boost to subvolumes.

60 APPLIED LIFE SCIENCES↗

SU-F-I-32: Organ Doses from Pediatric Head CT Scan

Purpose: To evaluate the organ doses of pediatric patients who undergoing head CT scan using Monte Carlo (MC) simulation and compare it with measurements in anthropomorphic child phantom.. Methods: A ten years old children voxel phantom was developed from CT images, the voxel size of the phantom was 2mm*2mm*2mm. Organ doses from head CT scan were simulated using MCNPX software, 180 detectors were placed in the voxel phantom to tally the doses of the represented tissues or organs. When performing the simulation, 120 kVp and 88 mA were selected as the scan parameters. The scan range covered from the top of the head to the end of the chain, this protocol was used at CT simulator for radiotherapy. To validate the simulated results, organ doses were measured with radiophotoluminescence (RPL) detectors, placed in the 28 organs of the 10 years old CIRS ATOM phantom. Results: The organ doses results matched well between MC simulation and phantom measurements. The eyes dose was showed to be as expected the highest organ dose: 28.11 mGy by simulation and 27.34 mGy by measurement respectively. Doses for organs not included in the scan volume were much lower than those included in the scan volume, thymus doses were observed more than 10 mGy due the CT protocol for radiotherapy covered more body part than routine head CT scan. Conclusion: As the eyes are superficial organs, they may receive the highest radiation dose during the CT scan. Considering the relatively high radio sensitivity, using shielding material or organ based tube current modulation technique should be encouraged to reduce the eye radiation risks. Scan range was one of the most important factors that affects the organ doses during the CT scan. Use as short as reasonably possible scan range should be helpful to reduce the patient radiation dose. This work was supported by the National Natural Science Foundation of China(11475047)

60 APPLIED LIFE SCIENCES↗

SU-F-I-69: Absorbed Dose Estimation for a Commercially Available MicroCT Scanner with Various Filtration Techniques

Purpose: To quantify the simulated mean absorbed dose per technique (cGy/mAs) from a commercially available microCT scanner using various filtration techniques. Methods: Monte Carlo simulations using the Geant4 toolkit (version 10) was used utilizing the standard electromagnetic physics model. The Quantum FX microCT scanner (PerkinElmer, Waltham, MA) was modeled incorporating measured energy spectra and spatial dimensions of nominal source-to-object (SOD) distances. The energy distribution was measured using a spectrometer (X-123CdTe, Amptek Inc., Bedford, USA) for the 90 kVp X-ray beams with various filters (including no filter, 1 mm, 2 mm, 3 mm, 4 mm Al and 0.2 mm Cu + 2.5 mm Al). The SOD was set to 154 mm, 104 mm, and 52 mm. A total of 10 million incident particles were processed per simulation. Cutout value was set to 0.1 mm for both photon and electron. The mean dose absorbed (cGy/per incident particle) in a PMMA phantom (length of 2 cm and radius of 3 cm) were recorded. Exposure measurements were taken using a Radcal 9095 system with a protocol of 90 kVp, 200 µA, and ∼12 s beam-On time for the various filters. Results: The mean absorbed dose per mAs for various filtrations and different SOD setups indicated that the dose decreased as the SOD increased and as the amount of filtration increased. For a given SOD, the dose was reduced by as much as ∼13.7% by varying the filter (from 0.2 mm Cu + 2.5 mm Al to no filter). The maximum dose was found to be 0.39 cGy/mAs (SOD of 5.196 cm, no filter) while the minimum dose value was 0.077 cGy/mAs (SOD of 15.4 cm, .2mm Cu + 2mm Al filter). Conclusion: This study estimates easily the mean dose for objects scanned with a microCT scanner with different filtration.

60 APPLIED LIFE SCIENCES↗

SU-F-J-12: Dosimetrical Characteristics of a 2.5 MV Megavoltage Photon Beam

Purpose: To evaluate the accuracy of modeling scatter factor (SF) and primary off-axis ratio (POAR) for 2.5 MV megavoltage photon beams using an empirical model, which can be used as a dose calculation model for low energy EPID imaging. Methods: Scatter photon parameters were calculated for 2.5 MV photon beam from a Varian TrueBeam by fitting the product of depth dose, PDD, and the phantom scatter factor, Sp, for a range of square field sizes between 2 and 40 cm and depths between 0 and 30 cm. The model is then applied to off-axis profiles measured for a slit field (3cm ×35 cm) for 5 depths: 0.6, 5, 10, and 20, 30 cm. Primary off-axis ratio (POAR) values were determined based on the fitting to data using the empirical model. Results: The fitting to PDD*Sp has a maximum error of 14%, mostly near surface, where backscattering of photons dominate and has a standard deviation of 4.7%. The attenuation coefficient and beam-hardening coefficient of the beam is determined to be 0.095 1/cm and 0.0038 1/cm, respectively. POAR for 2.5 MV has a peak in the center, corresponding to a beam without sufficient material to flatten the beam. Conclusion: This work illustrates a deficiency of the current empirical model to fit SF for a 2.5 MV photon beams. Most of the errors are near the surface region and an improved model that incorporate better modeling of backscattering can be developed to improve the model accuracy.

60 APPLIED LIFE SCIENCES↗

SU-F-J-42: Comparison of Varian TrueBeam Cone-Beam CT and BrainLab ExacTrac X-Ray for Cranial Radiotherapy

Purpose: To compare online image registrations of TrueBeam cone-beam CT (CBCT) and BrainLab ExacTrac x-ray imaging systems for cranial radiotherapy. Method: Phantom and patient studies were performed on a Varian TrueBeam STx linear accelerator (Version 2.5), which is integrated with a BrainLab ExacTrac imaging system (Version 6.1.1). The phantom study was based on a Rando head phantom, which was designed to evaluate isocenter-location dependence of the image registrations. Ten isocenters were selected at various locations in the phantom, which represented clinical treatment sites. CBCT and ExacTrac x-ray images were taken when the phantom was located at each isocenter. The patient study included thirteen patients. CBCT and ExacTrac x-ray images were taken at each patient’s treatment position. Six-dimensional image registrations were performed on CBCT and ExacTrac, and residual errors calculated from CBCT and ExacTrac were compared. Results: In the phantom study, the average residual-error differences between CBCT and ExacTrac image registrations were: 0.16±0.10 mm, 0.35±0.20 mm, and 0.21±0.15 mm, in the vertical, longitudinal, and lateral directions, respectively. The average residual-error differences in the rotation, roll, and pitch were: 0.36±0.11 degree, 0.14±0.10 degree, and 0.12±0.10 degree, respectively. In the patient study, the average residual-error differences in the vertical, longitudinal, and lateral directions were: 0.13±0.13 mm, 0.37±0.21 mm, 0.22±0.17 mm, respectively. The average residual-error differences in the rotation, roll, and pitch were: 0.30±0.10 degree, 0.18±0.11 degree, and 0.22±0.13 degree, respectively. Larger residual-error differences (up to 0.79 mm) were observed in the longitudinal direction in the phantom and patient studies where isocenters were located in or close to frontal lobes, i.e., located superficially. Conclusion: Overall, the average residual-error differences were within 0.4 mm in the translational directions and were within 0.4 degree in the rotational directions.

60 APPLIED LIFE SCIENCES↗

SU-F-T-468: Efficient Scanning Data Analysis for TomoTherapy

Purpose: The standard TomoTherapy 2D scanning system requires manual analysis using spreadsheets or similar software. Manual analysis is time consuming and does not allow efficient use of the medical physicist or timely results that may indicate issues during the acquisition of data. The purpose of this study is to determine if an in-house solution can provide timely analysis of TomoTherapy scan data. Methods: A custom MATLAB program utilizing a streamlined graphical interface was developed to interpret the native scanning data acquired by the TomoTherapy Electrometer Measurement System (TEMS). The gold beam data provided by the manufacturer was used as a baseline for the acquired data. Two separate sets of scan data were used including longitudinal, transverse, and percent depth dose (PDD). The scan data were compared visually in an overlay and analytically for each beam width using the full width half max (FWHM) for longitudinal profiles and difference inside the field core for the transverse profiles as recommended by TG-148. A plot of the PDD percent error at each point along the curve was generated to show agreement. The time to analysis was measured on a laptop running an Intel i5 processor with 8GB of RAM. Results: The software program required on average less than 5 seconds to load and analyze the results. No conversion or reformatting of the data was required. This allowed real-time visualization of the comparison to the gold beam data as well as the parameters recommended by TG-148. The average percent difference was 0.3% for the longitudinal profiles and 0.1% for the transverse profiles. All PDD data was less than 1% different beyond dmax. Conclusion: We have developed a tool to analyze TomoTherapy scan data in real-time. This allows an efficient method to perform beam commissioning, annual QA, and adjust the beam parameters if necessary.

61 RADIATION PROTECTION AND DOSIMETRY↗

SU-F-T-510: Single Isocenter Radiosurgery for the Simultaneous Treatment of Multiple Brain Metastases: Volumetric Modulated Arc Therapy Or Dynamic Conformal Arc?

Purpose: This study evaluates a novel algorithm that can be used with any treatment planning system for simple and rapid generation of stereotactic radiosurgery (SRS) plans for treating multiple brain metastases using a single isocenter dynamic conformal arc (DCA) approach. This technique is compared with a single isocenter volumetric modulated arc therapy (VMAT) technique in terms of delivery time, conformity, low dose spread and delivery accuracy. Methods: Five patients, with a total of 37 (5 – 11) targets were planned using a previously published method for generating optimal VMAT plans and using the proposed DCA algorithm. All planning target volumes (PTVs) were planned to 20 Gy, meeting a minimum 99% coverage and maximum 135 % hot spot for both techniques. Quality assurance was performed using radiochromic film, with films placed in the high dose regions of each PTV. Normal tissue volumes receiving 12 Gy and 6 Gy (V12 and V6) were computed for each plan. Conformity index (CI) and gamma evaluations (95% of points passing 4%/0.5mm) were computed for each PTV. Results: Delivery times, including beam on and table rotation times, were comparable: 17 – 22 minutes for all deliveries. V12s for DCA plans were (18.5±15.2 cc) vs. VMAT (19.7±14.4 cc). V6s were significantly lower for DCA (69.0±52.0 cc) compared with VMAT (154.0±91.0 cc) (p <<0.05). CIs for VMAT targets were (1.38±0.50) vs. DCA (1.61±0.41). 36 of 37 DCA planned targets passed gamma tests, while 29 of 37 VMAT planned targets passed. Conclusion: Single isocenter DCA plans were easily achieved. The evaluation suggests that DCA may represent a favorable technique compared with VMAT for multiple target SRS by reducing dose to normal tissue and more accurately depicting deliverable dose.

60 APPLIED LIFE SCIENCES↗

SU-F-T-626: Intracranial SRS Re-Treatment Without Acquisition of New CT Images

Purpose: Linear accelerator based stereotactic radiosurgery (SRS) for multiple intracranial lesions with frequent surveillance is becoming a popular treatment option. This strategy leads to retreatment with SRS as new lesions arise. Currently, each course of treatment uses magnetic resonance (MR) and computed tomography (CT) images for treatment planning. We propose that new MR images, with course 1 CT images, may be used for future treatment plans with negligible loss of dosimetric accuracy. Methods: Ten patients that received multiple courses of SRS were retrospectively reviewed. The treatment plans and contours from non-initial courses were copied to the initial CTs and recalculated. Doses metrics for the plans calculated on the initial CTs and later CTs were compared. All CT scans were acquired on a Philips CT scanner with a 600 mm field of view and 1 mm slice thickness (Philips Healthcare, Andover, MA). All targets were planned to 20 Gy and calculated in Eclipse V. 13.6 (Varian, Palo Alto, CA) using analytic anisotropic algorithm with 1 mm calculation grid. Results: Sixteen lesions were evaluated. The mean time between courses was 250 +/− 215 days (range 103–979). The mean target volume was 2.0 +/− 2.9 cc (range 0.1–10.1). The average difference in mean target dose between the two calculations was 0.2 +/− 0.3 Gy (range 0.0 – 1.0). The mean conformity index (CI) was 1.28 +/− 0.14 (range 1.07 – 1.82). The average difference in CI was 0.03 +/− 0.16 (range 0.00 – 0.44). Targets volumes < 0.5 cc showed the largest changes in both metrics. Conclusion: Continued treatment based on initial CT images is feasible. Dose calculation on the initial CT for future treatments provides reasonable dosimetric accuracy. Changes in dose metrics are largest for small volumes, and are likely dominated by partial volume effects in target definition.

60 APPLIED LIFE SCIENCES↗

SU-G-IeP1-07: Inaccuracy of Lesion Blood Flow Quantification Related to the Proton Density Reference Image in Arterial Spin Labeling MRI of Brain Tumors

Purpose: Cerebral blood flow quantification in arterial spin labeling (ASL) MRI requires an estimate of the equilibrium magnetization of blood, which is often obtained by a set of proton density (PD) reference image. Normally, a constant blood-brain partition coefficient is assumed across the brain. However, this assumption may not be valid for brain lesions. This study aimed to evaluate the impact of lesion-related PD variations on ASL quantification in patients with brain tumors. Methods: MR images for posttreatment evaluation of 42 patients with brain tumors were retrospectively analyzed. These images were acquired on a 3T MRI scanner, including T2-weighted FLAIR, 3D pseudo-continuous ASL and post-contrast T1-weighted images. Anatomical images were coregistered with ASL images using the SPM software. Regions of interest (ROIs) of the enhancing and FLAIR lesions were manually drawn on the coregistered images. ROIs of the contralateral normal appearing tissues were also determined, with the consideration of approximating coil sensitivity patterns in lesion ROIs. Relative lesion blood flow (lesion/contralateral tissue) was calculated from both the CBF map (dependent on the PD) and the ΔM map for comparison. Results: The signal intensities in both enhancing and FLAIR lesions were significantly different than contralateral tissues on the PD reference image (p<0.001). The percent signal difference ranged from −15.9 to 19.2%, with a mean of 5.4% for the enhancing lesion, and from −2.8 to 22.9% with a mean of 10.1% for the FLAIR lesion. The high/low lesion-related PD signal resulted in inversely proportional under-/over-estimation of blood flow in both enhancing and FLAIR lesions. Conclusion: Significant signal differences were found between lesions and contralateral tissues in the PD reference image, which introduced errors in blood flow quantification in ASL. The error can be up to 20% in individual patients with an average of 5- 10% for the group of patients with brain tumors.

60 APPLIED LIFE SCIENCES↗

SU-G-IeP1-12: Size Selective Arterial Cerebral Blood Volume Mapping Using Multiple Inversion Time Arterial Spin Labeling

Purpose: To demonstrate the feasibility of a novel method for size specific arterial cerebral blood volume (aCBV) mapping using pseudo-continuous arterial spin labeling (PCASL), with multiple TI. Methods: Multiple PCASL images were obtained from a subject with TI of [300, 400, 500, 600, 700, 800, 900, 1000, 1500, 2000, 2500, 3000, 3500, 4000] ms. Each TI pair was averaged six times. Two scans were performed: one without a flow crusher gradient and the other with a crusher gradient (10cm/s in three directions) to remove signals from large arteries. Scan times were 5min. without a crusher gradient and 5.5 min with a crusher gradient. Non-linear fitting algorithm finds the minimum mean squared solution of per-voxel based aCBV, cerebral blood flow, and arterial transit time, and fits the data into a hemodynamic model that represents superposition of blood volume and flow components within a single voxel. Results: aCBV maps with a crusher gradient represent signals from medium and small sized arteries, while those without a crusher gradient represent signals from all sized arteries, indicating that flow crusher gradients can be effectively employed to achieve size-specific aCBV mapping. Regardless of flow crusher, the CBF and ATT maps are very similar in appearance. Conclusion: Quantitative size selective blood volume mapping controlled by a flow crusher is feasible without additional information because the ASL quantification process doesn’t require an arterial input function measured from a large artery. The size specific blood volume mapping is not interfered by sSignals from large arteries do not interfere with size specific aCBV mapping in the applications of interest in for applications in which only medium or small arteries are of interest.

60 APPLIED LIFE SCIENCES↗

MO-DE-207A-05: Dictionary Learning Based Reconstruction with Low-Rank Constraint for Low-Dose Spectral CT

Purpose: Spectral CT enabled by an energy-resolved photon-counting detector outperforms conventional CT in terms of material discrimination, contrast resolution, etc. One reconstruction method for spectral CT is to generate a color image from a reconstructed component in each energy channel. However, given the radiation dose, the number of photons in each channel is limited, which will result in strong noise in each channel and affect the final color reconstruction. Here we propose a novel dictionary learning method for spectral CT that combines dictionary-based sparse representation method and the patch based low-rank constraint to simultaneously improve the reconstruction in each channel and to address the inter-channel correlations to further improve the reconstruction. Methods: The proposed method has two important features: (1) guarantee of the patch based sparsity in each energy channel, which is the result of the dictionary based sparse representation constraint; (2) the explicit consideration of the correlations among different energy channels, which is realized by patch-by-patch nuclear norm-based low-rank constraint. For each channel, the dictionary consists of two sub-dictionaries. One is learned from the average of the images in all energy channels, and the other is learned from the average of the images in all energy channels except the current channel. With average operation to reduce noise, these two dictionaries can effectively preserve the structural details and get rid of artifacts caused by noise. Combining them together can express all structural information in current channel. Results: Dictionary learning based methods can obtain better results than FBP and the TV-based method. With low-rank constraint, the image quality can be further improved in the channel with more noise. The final color result by the proposed method has the best visual quality. Conclusion: The proposed method can effectively improve the image quality of low-dose spectral CT. This work is partially supported by the National Natural Science Foundation of China (No. 61302136), and the Natural Science Basic Research Plan in Shaanxi Province of China (No. 2014JQ8317).

60 APPLIED LIFE SCIENCES↗

MO-FG-CAMPUS-TeP1-03: Pre-Treatment Surface Imaging Based Collision Detection

Purpose: Modern radiotherapy increasingly employs large immobilization devices, gantry attachments, and couch rotations for treatments. All of which raise the risk of collisions between the patient and the gantry / couch. Collision detection is often achieved by manually checking each couch position in the treatment room and sometimes results in extraneous imaging if collisions are detected after image based setup has begun. In the interest of improving efficiency and avoiding extra imaging, we explore the use of a surface imaging based collision detection model. Methods: Surfaces acquired from AlignRT (VisionRT, London, UK) were transferred in wavefront format to a custom Matlab (Mathworks, Natick, MA) software package (CCHECK). Computed tomography (CT) scans acquired at the same time were sent to CCHECK in DICOM format. In CCHECK, binary maps of the surfaces were created and overlaid on the CT images based on the fixed relationship of the AlignRT and CT coordinate systems. Isocenters were added through a graphical user interface (GUI). CCHECK then compares the inputted surfaces to a model of the linear accelerator (linac) to check for collisions at defined gantry and couch positions. Note, CCHECK may be used with or without a CT. Results: The nominal surface image field of view is 650 mm × 900 mm, with variance based on patient position and size. The accuracy of collision detections is primarily based on the linac model and the surface mapping process. The current linac model and mapping process yield detection accuracies on the order of 5 mm, assuming no change in patient posture between surface acquisition and treatment. Conclusions: CCHECK provides a non-ionizing method to check for collisions without the patient in the treatment room. Collision detection accuracy may be improved with more robust linac modeling. Additional gantry attachments (e.g. conical collimators) can be easily added to the model.

60 APPLIED LIFE SCIENCES↗

TU-H-CAMPUS-IeP2-01: Quantitative Evaluation of PROPELLER DWI Using QIBA Diffusion Phantom

Purpose: The purpose of this study is to determine the quantitative variability of apparent diffusion coefficient (ADC) values when varying imaging parameters in a diffusion-weighted (DW) fast spin echo (FSE) sequence with Periodically Rotated Overlapping ParallEL Lines with Enhanced Reconstruction (PROPELLER) k-space trajectory. Methods: Using a 3T MRI scanner, a NIST traceable, quantitative magnetic resonance imaging (MRI) diffusion phantom (High Precision Devices, Inc, Boulder, Colorado) consisting of 13 vials filled with various concentrations of polymer polyvinylpyrrolidone (PVP) in aqueous solution was imaged with a standard Quantitative Imaging Biomarkers Alliance (QIBA) DWI spin echo, echo planar imaging (SE EPI) acquisition. The same phantom was then imaged with a DWI PROPELLER sequence at varying echo train lengths (ETL) of 8, 20, and 32, as well as b-values of 400, 900, and 2000. QIBA DWI phantom analysis software was used to generate ADC maps and create region of interests (ROIs) for quantitative measurements of each vial. Mean and standard deviations of the ROIs were compared. Results: The SE EPI sequence generated ADC values that showed very good agreement with the known ADC values of the phantom (r2 = 0.9995, slope = 1.0061). The ADC values measured from the PROPELLER sequences were inflated, but were highly correlated with an r2 range from 0.8754 to 0.9880. The PROPELLER sequence with an ETL=20 and b-value of 0 and 2000 showed the closest agreement (r2 = 0.9034, slope = 0.9880). Conclusion: The DW PROPELLER sequence is promising for quantitative evaluation of ADC values. A drawback of the PROPELLER sequence is the longer acquisition time. The 180° refocusing pulses may also cause the observed increase in ADC values compared to the standard SE EPI DW sequence. However, the FSE sequence offers an advantage with in-plane motion and geometric distortion which will be investigated in future studies.

60 APPLIED LIFE SCIENCES↗

WE-AB-209-02: A New Inverse Planning Framework with Principle-Based Modeling of Inter-Structural Dosimetric Tradeoffs

Purpose: Traditional radiotherapy inverse planning relies on the weighting factors to phenomenologically balance the conflicting criteria for different structures. The resulting manual trial-and-error determination of the weights has long been recognized as the most time-consuming part of treatment planning. The purpose of this work is to develop an inverse planning framework that parameterizes the inter-structural dosimetric tradeoff among with physically more meaningful quantities to simplify the search for a clinically sensible plan. Methods: A permissible dosimetric uncertainty is introduced for each of the structures to balance their conflicting dosimetric requirements. The inverse planning is then formulated as a convex feasibility problem, which aims to generate plans with acceptable dosimetric uncertainties. A sequential procedure (SP) is derived to decompose the model into three submodels to constrain the uncertainty in the planning target volume (PTV), the critical structures, and all other structures to spare, sequentially. The proposed technique is applied to plan a liver case and a head-and-neck case and compared with a conventional approach. Results: Our results show that the strategy is able to generate clinically sensible plans with little trial-and-error. In the case of liver IMRT, the fractional volumes to liver and heart above 20Gy are found to be 22% and 10%, respectively, which are 15.1% and 33.3% lower than that of the counterpart conventional plan while maintaining the same PTV coverage. The planning of the head and neck IMRT show the same level of success, with the DVHs for all organs at risk and PTV very competitive to a counterpart plan. Conclusion: A new inverse planning framework has been established. With physically more meaningful modeling of the inter-structural tradeoff, the technique enables us to substantially reduce the need for trial-and-error adjustment of the model parameters and opens new opportunities of incorporating prior knowledge to facilitate the treatment planning process.

60 APPLIED LIFE SCIENCES↗