Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Quantification”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 685 records · Page 38

Performance of today’s dual energy CT and future multi energy CT in virtual non-contrast imaging and in iodine quantification: A simulation study

Purpose: To study the performance of different dual energy computed tomography (DECT) techniques, which are available today, and future multi energy CT (MECT) employing novel photon counting detectors in an image-based material decomposition task. Methods: The material decomposition performance of different energy-resolved CT acquisition techniques is assessed and compared in a simulation study of virtual non-contrast imaging and iodine quantification. The material-specific images are obtained via a statistically optimal image-based material decomposition. A projection-based maximum likelihood approach was used for comparison with the authors’ image-based method. The different dedicated dual energy CT techniques are simulated employing realistic noise models and x-ray spectra. The authors compare dual source DECT with fast kV switching DECT and the dual layer sandwich detector DECT approach. Subsequent scanning and a subtraction method are studied as well. Further, the authors benchmark future MECT with novel photon counting detectors in a dedicated DECT application against the performance of today’s DECT using a realistic model. Additionally, possible dual source concepts employing photon counting detectors are studied. Results: The DECT comparison study shows that dual source DECT has the best performance, followed by the fast kV switching technique and the sandwich detector approach. Comparing DECT with future MECT, the authors found noticeable material image quality improvements for an ideal photon counting detector; however, a realistic detector model with multiple energy bins predicts a performance on the level of dual source DECT at 100 kV/Sn 140 kV. Employing photon counting detectors in dual source concepts can improve the performance again above the level of a single realistic photon counting detector and also above the level of dual source DECT. Conclusions: Substantial differences in the performance of today’s DECT approaches were found for the application of virtual non-contrast and iodine imaging. Future MECT with realistic photon counting detectors currently can only perform comparably to dual source DECT at 100 kV/Sn 140 kV. Dual source concepts with photon counting detectors could be a solution to this problem, promising a better performance.

60 APPLIED LIFE SCIENCES↗

SU-C-201-04: Quantification of Perfusion Heterogeneity Based On Texture Analysis for Fully Automatic Detection of Ischemic Deficits From Myocardial Perfusion Imaging

Purpose: Texture-based quantification of image heterogeneity has been a popular topic for imaging studies in recent years. As previous studies mainly focus on oncological applications, we report our recent efforts of applying such techniques on cardiac perfusion imaging. A fully automated procedure has been developed to perform texture analysis for measuring the image heterogeneity. Clinical data were used to evaluate the preliminary performance of such methods. Methods: Myocardial perfusion images of Thallium-201 scans were collected from 293 patients with suspected coronary artery disease. Each subject underwent a Tl-201 scan and a percutaneous coronary intervention (PCI) within three months. The PCI Result was used as the gold standard of coronary ischemia of more than 70% stenosis. Each Tl-201 scan was spatially normalized to an image template for fully automatic segmentation of the LV. The segmented voxel intensities were then carried into the texture analysis with our open-source software Chang Gung Image Texture Analysis toolbox (CGITA). To evaluate the clinical performance of the image heterogeneity for detecting the coronary stenosis, receiver operating characteristic (ROC) analysis was used to compute the overall accuracy, sensitivity and specificity as well as the area under curve (AUC). Those indices were compared to those obtained from the commercially available semi-automatic software QPS. Results: With the fully automatic procedure to quantify heterogeneity from Tl-201 scans, we were able to achieve a good discrimination with good accuracy (74%), sensitivity (73%), specificity (77%) and AUC of 0.82. Such performance is similar to those obtained from the semi-automatic QPS software that gives a sensitivity of 71% and specificity of 77%. Conclusion: Based on fully automatic procedures of data processing, our preliminary data indicate that the image heterogeneity of myocardial perfusion imaging can provide useful information for automatic determination of the myocardial ischemia.

60 APPLIED LIFE SCIENCES↗

SU-D-303-03: Impact of Uncertainty in T1 Measurements On Quantification of Dynamic Contrast Enhanced MRI

Purpose: Quantification of dynamic contrast enhanced (DCE) MRI requires native longitudinal relaxation time (T1) measurement. This study aimed to assess uncertainty in T1 measurements using two different methods. Methods and Materials: Brain MRI scans were performed on a 3T scanner in 9 patients who had low grade/benign tumors and partial brain radiotherapy without chemotherapy at pre-RT, week-3 during RT (wk-3), end-RT, and 1, 6 and 18 months after RT. T1-weighted images were acquired using gradient echo sequences with 1) 2 different flip angles (50 and 150), and 2) 5 variable TRs (100–2000ms). After creating quantitative T1 maps, average T1 was calculated in regions of interest (ROI), which were distant from tumors and received a total of accumulated radiation doses < 5 Gy at wk-3. ROIs included left and right normal Putamen and Thalamus (gray matter: GM), and frontal and parietal white matter (WM). Since there were no significant or even a trend of T1 changes from pre-RT to wk-3 in these ROIs, a relative repeatability coefficient (RC) of T1 as a measure of uncertainty was estimated in each ROI using the data pre-RT and at wk-3. The individual T1 changes at later time points were evaluated compared to the estimated RCs. Results: The 2-flip angle method produced small RCs in GM (9.7–11.7%) but large RCs in WM (12.2–13.6%) compared to the saturation-recovery (SR) method (11.0–17.7% for GM and 7.5–11.2% for WM). More than 81% of individual T1 changes were within T1 uncertainty ranges defined by RCs. Conclusion: Our study suggests that the impact of T1 uncertainty on physiological parameters derived from DCE MRI is not negligible. A short scan with 2 flip angles is able to achieve repeatability of T1 estimates similar to a long scan with 5 different TRs, and is desirable to be integrated in the DCE protocol. Present study was supported by National Institute of Health (NIH) under grant numbers; UO1 CA183848 and RO1 NS064973.

60 APPLIED LIFE SCIENCES↗

SU-E-I-25: Quantification of Coronary Artery Cross-Sectional Area in CT Angiography Using Integrated Density: A Simulation Study

Purpose: To develop a physics-based model for accurate quantification of the cross-sectional area (CSA) of coronary arteries in CT angiography by measuring the integrated density to account for the partial volume effect. Methods: In this technique the integrated density of the object as compared with its local background is measured to account for the partial volume effect. Normal vessels were simulated as circles with diameters in the range of 0.1–3mm. Diseased vessels were simulated as 2, 3, and 4mm diameter vessels with 10–90% area stenosis, created by inserting circular plaques. A simplified two material model was used with the lumen as 8mg/ml Iodine and background as lipid. The contrast-to-noise ratio between lumen and background was approximately 26. Linear fits to the known CSA were calculated. The precision and accuracy of the measurement were quantified using the root-mean-square fit deviations (RMSD) and errors to the known CSA (RMSE). Results compared to manual segmentation of the vessel lumen. To assess the impact of random variations, coefficients of variation (CV) from 10 simulations for each vessel were computed to determine reliability. Measurements with CVs less than 10% were considered reliable. Results: For normal vessels, the precision and accuracy of the integrated density technique were 0.12mm{sup 2} and 0.28mm{sup 2}, respectively. The corresponding results for manual segmentation were 0.27mm{sup 2} and 0.43mm{sup 2}. For diseased vessels, the precision and accuracy of the integrated density technique were 0.14mm{sup 2} and 0.19mm{sup 2}. Corresponding results for manual segmentation were 0.42mm{sup 2} and 0.71mm{sup 2}. Reliable CSAs were obtained for normal vessels with diameters larger than 1 mm and for diseased vessels with area as low as 1.26mm2. Conclusion: The CSA based on integrated density showed improved precision and accuracy as compared with manual segmentation in simulation. These results indicate the potential of using integrated density to quantify CSA of coronary arteries in CT angiography.

60 APPLIED LIFE SCIENCES↗

SU-E-J-216: A Sequence Independent Approach for Quantification of MR Image Deformations From Brachytherapy Applicators

Purpose: MRI is increasingly used as a single imaging modality for brachytherapy treatment planning. The presence of a brachytherapy applicator may cause distortions in the images, especially at higher field strengths. Our aim is to develop a procedure to quantify these distortions theoretically for any MR-sequence and to verify the estimated deformations for clinical sequences. Methods: Image distortions due to perturbation of the B0-field are proportional to the ratio of the induced frequency shift and the read-out bandwidth of the applied sequence. By reconstructing a frequency-shift map from the phase data from a multi-echo sequence, distortions can be calculated for any MR-sequence. Verification of this method for estimating distortions was performed by acquiring images with opposing read-out directions and consequently opposing distortions. The applicator shift can be determined by rigidly matching these images. Clinically, T2W-TSE-images are used for this purpose. For pre-clinical tests, EPI-sequences with narrow read-out bandwidth (19.5–47.5Hz), consequently large distortions, were added to the set of clinical MRsequences. To quantify deformations of the Utrecht Interstitial CT/MR applicator (Elekta Brachytherapy) on a Philips Ingenia 3T MRI, pre-clinical tests were performed in a phantom with the applicator in water, followed by clinical validation. Results: Deformations observed in the narrow bandwidth EPI-images were well predicted using the frequency-shift, the latter giving an overestimation up to 30%/up to 1 voxel. For clinically applied MR-sequences distortions were well below the voxel size. In patient setup distortions determined from the frequency-shift map were at sub-voxel level (<0.7mm). Using T2W-images larger distortions were found (1–2mm). This discrepancy was caused by patient movement between/during acquisition of the T2W-images with opposing read-out directions. Conclusion: Phantom experiments demonstrated the feasibility of a clinical procedure for quantification of MR-image distortions for any MR-sequence. In a clinical set-up the distortions from a Utrecht interstitial CT/MR applicator are sub-voxel level. This work was partially funded by Elekta Brachytherpy.

60 APPLIED LIFE SCIENCES↗

SU-E-T-650: Quantification and Modeling of the Dosimetric Impact of the IBEAM Evo Treatment Couchtop EP (Elekta) in VMAT

Purpose: quantification and modelling of the dosimetric impact of the treatment couch in Monaco Treatment Planning System. Methods: The attenuation characteristics of couchtop EP was evaluated for two different photon acceleration potentials (6MV and 10MV) for a field size of (10×10) cm2. Phantom positions in A-B direction: on the left half, in the center and on the right half of the couch. Dose measurements of couch attenuation were performed at gantry angles from 180° to 122°, using a 0.125cc semiflex ionization chamber isocentrically placed in the center of a homogeneous cylindric sliced RW3 phantom. Each experimental setup was first measured on the LINAC and then reproduced in the TPS. By adjusting the relative-to-water electron density (ED) values of the couch, the measured attenuation was replicated. The simulated results were evaluated by comparing the measurements and simulations. Results: Without the couch model included the maximum difference between measured and calculated dose was 5.5% (5.1%) and 6.6% (6.1%) for 2 mm and 5 mm voxel size, when the phantom was positioned on the left (center). The couch model was included in the TPS with a uniform ED of 0.18 or a 2 component model with a fiber ED= 0.6 and foam core ED= 0.1. After including the treatment couch, the mean dose attenuation was reduced from 2.8% without couch included to (0.0, 0.8, −0.2, 0.6)%. The 4 different values represent the 1 and 2 components model and 2 and 5 mm voxel grid size. Conclusion: For a uniform relative-to-water couch electron density of 0.18 a good agreement between measured and calculated dose distributions was obtained for all different energies, voxel grid spacings and gantry angles. Therefore, we conclude that the Monaco couch model accurately describes the dose perturbations due to the presence of the patient couch and should therefore be used during treatment planning. This project is supported by Technology Foundation for Selected Overseas Chinese Scholar, Ministry of Hebei Personnel of China.

61 RADIATION PROTECTION AND DOSIMETRY↗

TU-EF-210-03: Real-Time Ablation Monitoring and Lesion Quantification Using Harmonic Motion Imaging

The use of therapeutic ultrasound to provide targeted therapy is an active research area that has a broad application scope. The invited talks in this session will address currently implemented strategies and protocols for both hyperthermia and ablation applications using therapeutic ultrasound. The role of both ultrasound and MRI in the monitoring and assessment of these therapies will be explored in both pre-clinical and clinical applications. Katherine Ferrara: High Intensity Focused Ultrasound, Drug Delivery, and Immunotherapy Rajiv Chopra: Translating Localized Doxorubicin Delivery to Pediatric Oncology using MRI-guided HIFU Elisa Konofagou: Real-time Ablation Monitoring and Lesion Quantification using Harmonic Motion Imaging Keyvan Farahani: AAPM Task Groups in Interventional Ultrasound Imaging and Therapy Learning Objectives: Understand the role of ultrasound in localized drug delivery and the effects of immunotherapy when used in conjunction with ultrasound therapy. Understand potential targeted drug delivery clinical applications including pediatric oncology. Understand the technical requirements for performing targeted drug delivery. Understand how radiation-force approaches can be used to both monitor and assess high intensity focused ultrasound ablation therapy. Understand the role of AAPM task groups in ultrasound imaging and therapies. Chopra: Funding from Cancer Prevention and Research Initiative of Texas (CPRIT), Award R1308 Evelyn and M.R. Hudson Foundation; Research Support from Research Contract with Philips Healthcare; COI are Co-founder of FUS Instruments Inc Ferrara: Supported by NIH, UCDavis and California (CIRM and BHCE) Farahani: In-kind research support from Philips Healthcare.

60 APPLIED LIFE SCIENCES↗

WE-D-204-07: Development of An ImageJ Plugin for Renal Function Quantification: RenalQuant

Purpose: Commercial workstations usually have their own software to calculate dynamic renal functions. However, usually they have low flexibility and subjectivity on delimiting kidney and background areas. The aim of this paper is to present a public domain software, called RenalQuant, capable to semi-automatically draw regions of interest on dynamic renal scintigraphies, extracting data and generating renal function quantification parameters. Methods: The software was developed in Java and written as an ImageJ-based plugin. The preprocessing and segmentation steps include the user’s selection of one time frame with higher activity in kidney’s region, compared with background, and low activity in the liver. Next, the chosen time frame is smoothed using a Gaussian low pass spatial filter (σ = 3) for noise reduction and better delimitation of kidneys. The maximum entropy thresholding method is used for segmentation. A background area is automatically placed below each kidney, and the user confirms if these regions are correctly segmented and positioned. Quantitative data are extracted and each renogram and relative renal function (RRF) value is calculated and displayed. Results: RenalQuant plugin was validated using retrospective 20 patients’ 99mTc-DTPA exams, and compared with results produced by commercial workstation software, referred as reference. The renograms intraclass correlation coefficients (ICC) were calculated and false-negative and false-positive RRF values were analyzed. The results showed that ICC values between RenalQuant plugin and reference software for both kidneys’ renograms were higher than 0.75, showing excellent reliability. Conclusion: Our results indicated RenalQuant plugin can be trustingly used to generate renograms, using DICOM dynamic renal scintigraphy exams as input. It is user friendly and user’s interaction occurs at a minimum level. Further studies have to investigate how to increase RRF accuracy and explore how to solve limitations in the segmentation step, mainly when background region has higher activity compared to kidneys. Financial support by CAPES.

60 APPLIED LIFE SCIENCES↗

Evaluation of in vivo quantification accuracy of the Ingenuity-TF PET/MR

Purpose: The quantitative accuracy of standardized uptake values (SUVs) and tracer kinetic uptake parameters in patient investigations strongly depends on accurate determination of regional activity concentrations in positron emission tomography (PET) data. This determination rests on the assumption that the given scanner calibration is valid in vivo. In a previous study, we introduced a method to test this assumption. This method allows to identify discrepancies in quantitative accuracy in vivo by comparison of activity concentrations of urine samples measured in a well-counter with activity concentrations extracted from PET images of the bladder. In the present study, we have applied this method to the Philips Ingenuity-TF PET/MR since at the present stage, absolute quantitative accuracy of combined PET/MR systems is still under investigation. Methods: Twenty one clinical whole-body F18-FDG scans were included in this study. The bladder region was imaged as the last bed position and urine samples were collected afterward. PET images were reconstructed including MR-based attenuation correction with and without truncation compensation and 3D regions-of-interest (ROIs) of the bladder were delineated by three observers. To exclude partial volume effects, ROIs were concentrically shrunk by 8–10 mm. Then, activity concentrations were determined in the PET images for the bladder and for the urine by measuring the samples in a calibrated well-counter. In addition, linearity measurements of SUV vs singles rate and measurements of the stability of the coincidence rate of “true” events of the PET/MR system were performed over a period of 4 months. Results: The measured in vivo activity concentrations were significantly lower in PET/MR than in the well-counter with a ratio of the former to the latter of 0.756 ± 0.060 (mean ± std. dev.), a range of 0.604–0.858, and a P value of 3.9 ⋅ 10{sup −14}. While the stability measurements of the coincidence rate of “true” events showed no relevant deviation over time, the linearity scans revealed a systematic error of 8%–11% (avg. 9%) for the range of singles rates present in the bladder scans. After correcting for this systematic bias caused by shortcomings of the manufacturers calibration procedure, the PET to well-counter ratio increased to 0.832 ± 0.064 (0.668 –0.941), P = 1.1 ⋅ 10{sup −10}. After compensating for truncation of the upper extremities in the MR-based attenuation maps, the ratio further improved to 0.871 ± 0.069 (0.693–0.992), P = 3.9 ⋅ 10{sup −8}. Conclusions: Our results show that the Philips PET/MR underestimates activity concentrations in the bladder by 17%, which is 7 percentage points (pp.) larger than in the previously investigated PET and PET/CT systems. We attribute this increased underestimation to remaining limitations of the MR-based attenuation correction. Our results suggest that only a 2 pp. larger underestimation of activity concentrations compared to PET/CT can be observed if compensation of attenuation truncation of the upper extremities is applied. Thus, quantification accuracy of the Philips Ingenuity-TF PET/MR can be considered acceptable for clinical purposes given the ±10% error margin in the EANM guidelines. The comparison of PET images from the bladder region with urine samples has proven a useful method. It might be interesting for evaluation and comparison of the in vivo quantitative accuracy of PET, PET/CT, and especially PET/MR systems from different manufacturers or in multicenter trials.

60 APPLIED LIFE SCIENCES↗

Correction of quantification errors in pelvic and spinal lesions caused by ignoring higher photon attenuation of bone in [{sup 18}F]NaF PET/MR

Purpose: MR-based attenuation correction (MRAC) in routine clinical whole-body positron emission tomography and magnetic resonance imaging (PET/MRI) is based on tissue type segmentation. Due to lack of MR signal in cortical bone and the varying signal of spongeous bone, standard whole-body segmentation-based MRAC ignores the higher attenuation of bone compared to the one of soft tissue (MRAC{sub nobone}). The authors aim to quantify and reduce the bias introduced by MRAC{sub nobone} in the standard uptake value (SUV) of spinal and pelvic lesions in 20 PET/MRI examinations with [{sup 18}F]NaF. Methods: The authors reconstructed 20 PET/MR [{sup 18}F]NaF patient data sets acquired with a Philips Ingenuity TF PET/MRI. The PET raw data were reconstructed with two different attenuation images. First, the authors used the vendor-provided MRAC algorithm that ignores the higher attenuation of bone to reconstruct PET{sub nobone}. Second, the authors used a threshold-based algorithm developed in their group to automatically segment bone structures in the [{sup 18}F]NaF PET images. Subsequently, an attenuation coefficient of 0.11 cm{sup −1} was assigned to the segmented bone regions in the MRI-based attenuation image (MRAC{sub bone}) which was used to reconstruct PET{sub bone}. The automatic bone segmentation algorithm was validated in six PET/CT [{sup 18}F]NaF examinations. Relative SUV{sub mean} and SUV{sub max} differences between PET{sub bone} and PET{sub nobone} of 8 pelvic and 41 spinal lesions, and of other regions such as lung, liver, and bladder, were calculated. By varying the assigned bone attenuation coefficient from 0.11 to 0.13 cm{sup −1}, the authors investigated its influence on the reconstructed SUVs of the lesions. Results: The comparison of [{sup 18}F]NaF-based and CT-based bone segmentation in the six PET/CT patients showed a Dice similarity of 0.7 with a true positive rate of 0.72 and a false discovery rate of 0.33. The [{sup 18}F]NaF-based bone segmentation worked well in the pelvis and spine. However, it showed artifacts in the skull and in the extremities. The analysis of the 20 [{sup 18}F]NaF PET/MRI examinations revealed relative SUV{sub max} differences between PET{sub nobone} and PET{sub bone} of (−8.8% ± 2.7%, p = 0.01) and (−8.1% ± 1.9%, p = 2.4 × 10{sup −8}) in pelvic and spinal lesions, respectively. A maximum SUV{sub max} underestimation of −13.7% was found in lesion in the third cervical spine. The averaged SUV{sub mean} differences in volumes of interests in lung, liver, and bladder were below 3%. The average SUV{sub max} differences in pelvic and spinal lesions increased from −9% to −18% and −8% to −17%, respectively, when increasing the assigned bone attenuation coefficient from 0.11 to 0.13 cm{sup −1}. Conclusions: The developed automatic [{sup 18}F]NaF PET-based bone segmentation allows to include higher bone attenuation in whole-body MRAC and thus improves quantification accuracy for pelvic and spinal lesions in [{sup 18}F]NaF PET/MRI examinations. In nonbone structures (e.g., lung, liver, and bladder), MRAC{sub nobone} yields clinically acceptable accuracy.

60 APPLIED LIFE SCIENCES↗

SU-F-R-28: Correction of FCh-PET Bladder Uptake Using Virtual Sinograms and Investigation of Its Impact On the Quantification of Prostate Textural Characteristics

Purpose: Textural biomarkers as a tool for quantifying intratumoral heterogeneity hold great promise for diagnosis and early assessment of treatment response in prostate cancer. However, spill-in counts from the bladder uptake are suspected to have an impact on the textural measurements of the prostate volume. This work proposes a correction method for the FCh-PET bladder uptake and investigates its impact on intraprostatic textural properties. Methods: Two patients with PC received pre-treatment dynamic FCh-PET scans reconstructed at four time points (interval: 2 min), for which prostate and bladder contours were obtained. Projection bins affected by bladder uptake were determined by forward-projection. For each time point and axial position, virtual sinograms were obtained and affected bins replaced by a weighted combination of original values and values interpolated using cubic spline from non-affected bins of the current and adjacent projection angles. The process was optimized using a genetic algorithm in terms of minimization of the root-mean-square error (RMSE) within the bladder between the corrected dynamic time point volume and a reference initial uptake volume. Finally, the impact of the bladder uptake correction on the prostate region was investigated using two standard SUV metrics (1) and three texture metrics (2): 1) SUVmax, SUVmean; 2) Contrast, Homogeneity, Coarseness. Results: Without bladder uptake correction, SUVmax and SUVmean were on average overestimated in the prostate by 0%, 0%, 33.2%, 51.2%, and 3.6%, 6.0%, 2.9%, 3.2%, for each time point respectively. Contrast varied by −9.1%, −6.7%, +40.4%, +107.7%, and Homogeneity and Coarseness by +4.5%, +1.8%, −8.8%, −14.8% and +1.0%, +0.5%, −9.5%, +0.9%. Conclusion: We proposed a method for FCh-PET bladder uptake correction and showed an impact on the quantification of the prostate signal. This method achieved a large reduction of intra-prostatic SUVmax while minimizing the impact on SUVmean. Further investigation is necessary to interpret changes in textural features. SL acknowledges partial support by the CREATE Medical Physics Research Training Network grant of the Natural Sciences and Engineering Research Council (Grant number: 432290).

60 APPLIED LIFE SCIENCES↗

SU-G-IeP1-15: Towards Accurate Cerebral Blood Flow Quantification with Distortion- Corrected Pseudo-Continuous Arterial Spin Labeling

Purpose: Traditional arterial spin labeling (ASL) acquisitions with echo planar imaging (EPI) readouts suffer from image distortion due to susceptibility effects, compromising ASL’s ability to accurately quantify cerebral blood flow (CBF) and assess disease-specific patterns associated with CBF abnormalities. Phase labeling for additional coordinate encoding (PLACE) can remove image distortion; our goal is to apply PLACE to improve the quantitative accuracy of ASL CBF in humans. Methods: Four subjects were imaged on a 3T Philips Ingenia scanner using a 16-channel receive coil with a 21/21/10cm (frequency/phase/slice direction) field-of-view. An ASL sequence with a pseudo-continuous ASL (pCASL) labeling scheme was employed to acquire thirty dynamics of single-shot EPI data, with control and label datasets for all dynamics, and PLACE gradients applied on odd dynamics. Parameters included a post-labeling delay = 2s, label duration = 1.8s, flip angle = 90°, TR/TE = 5000/23.5ms, and 2.9/2.9/5.0mm (frequency/phase/slice direction) voxel size. “M0” EPI-reference images and T1-weighted spin-echo images with 0.8/1.0/3.3mm (frequency/phase/slice directions) voxel size were also acquired. Complex conjugate image products of pCASL odd and even dynamics were formed, a linear phase ramp applied, and data expanded and smoothed. Data phase was extracted to map control, label, and M0 magnitude image pixels to their undistorted locations, and images were rebinned to original size. All images were corrected for motion artifacts in FSL 5.0. pCASL images were registered to M0 images, and control and label images were subtracted to compute quantitative CBF maps. Results: pCASL image and CBF map distortions were removed by PLACE in all subjects. Corrected images conformed well to the anatomical T1-weighted reference image, and deviations in corrected CBF maps were evident. Conclusion: Eliminating pCASL distortion with PLACE can improve CBF quantification accuracy using minimal pulse sequence modifications and no additional scan time, improving ASL’s clinical applicability.

60 APPLIED LIFE SCIENCES↗

SU-G-JeP2-12: Quantification of 3D Geometric Distortion for 1.5T and 3T MRI Scanners Used for Radiation Therapy

Purpose: To quantify the magnitude of geometric distortion for MRI scanners and provide recommendations for MRI imaging for radiation therapy Methods: A novel phantom, QUASAR MRID3D [Modus Medical Devices Inc.], was scanned to evaluate the level of 3D geometric distortion present in five MRI scanners used for radiation therapy in our department. The phantom was scanned using the body coil with 1mm image slice thickness to acquire 3D images of the phantom body. The phantom was aligned to its geometric center for each scan, and the field of view was set to visualize the entire phantom. The dependence of distortion magnitude with distance from imaging isocenter and with magnetic field strength (1.5T and 3T) was investigated. Additionally, the characteristics of distortion for Siemens and GE machines were compared. The image distortion for each scanner was quantified in terms of mean, standard deviation (STD), maximum distortion, and skewness. Results: The 3T and 1.5T scans show a similar absolute distortion with a mean of 1.38mm (0.33mm STD) for 3T and 1.39mm (0.34mm STD) for 1.5T for a 100mm radius distance from isocenter. Some machines can have a distortion larger than 10mm at a distance of 200mm from the isocenter. The distortions are presented with plots of the x, y, and z directional components. Conclusion: The results indicate that quantification of MRI image distortion is crucial in radiation oncology for target and organ delineation and treatment planning. The magnitude of geometric distortion determines the margin needed for target contouring which is usually neglected in treatment planning process, especially for SRS/SBRT treatments. Understanding the 3D distribution of the MRI image distortion will improve the accuracy of target delineation and, hence, treatment efficacy. MRI imaging with proper patient alignment to the isocenter is vital to reducing the effects of MRI distortion in treatment planning.

60 APPLIED LIFE SCIENCES↗

SU-G-IeP3-07: High-Resolution, High-Sensitivity Imaging and Quantification of Intratumoral Distributions of Gold Nanoparticles Using a Benchtop L-Shell XRF Imaging System

Purpose: To demonstrate the ability to perform high-resolution imaging and quantification of sparse distributions of gold nanoparticles (GNPs) within ex vivo tumor samples using a highly-sensitive benchtop L-shell x-ray fluorescence (XRF) imaging system. Methods: An optimized L-shell XRF imaging system was assembled using a tungsten-target x-ray source (operated at 62 kVp and 45 mA). The x-rays were filtered (copper: 0.08 mm & aluminum: 0.04 mm) and collimated (lead: 5 cm thickness, 3 cm aperture diameter) into a cone-beam in order to irradiate small samples or objects. A collimated (stainless steel: 4 cm thickness, 2 mm aperture diameter) silicon drift detector, capable of 2D translation, was placed at 90° with respect to the beam to acquire XRF/scatter spectra from regions of interest. Spectral processing involved extracting XRF signal from background, followed by attenuation correction using a Compton scatter-based normalization algorithm. Calibration phantoms with water/GNPs (0 and 0.00001–10 mg/cm{sup 3}) were used to determine the detection limit of the system at a 10-second acquisition time. The system was then used to map the distribution of GNPs within a 12×11×2 mm{sup 3} slice excised from the center of a GNP-loaded ex vivo murine tumor sample; a total of 110 voxels (2.65×10{sup −3} cm{sup 3}) were imaged with 1.3-mm spatial resolution. Results: The detection limit of the current cone-beam benchtop L-shell XRF system was 0.003 mg/cm{sup 3} (3 ppm). Intratumoral GNP concentrations ranging from 0.003 mg/cm{sup 3} (3 ppm) to a maximum of 0.055 mg/cm{sup 3} (55 ppm) and average of 0.0093 mg/cm{sup 3} (9.3 ppm) were imaged successfully within the ex vivo tumor slice. Conclusion: The developed cone-beam benchtop L-shell XRF imaging system can immediately be used for imaging of ex vivo tumor samples containing low concentrations of GNPs. With minor finetuning/optimization, the system can be directly adapted for performing routine preclinical in vivo imaging tasks. Supported by NIH/NCI grant R01CA155446 This investigation was supported by NIH/NCI grant R01CA155446.

60 APPLIED LIFE SCIENCES↗

TH-CD-209-08: Quantification of the Interplay Effect in Proton Pencil Beam Scanning Treatment of Lung

Purpose: To quantify the dose degradation caused by the interplay effect based on a beam specific motion analysis in proton pencil beam scanning (PBS) treatment of lung tumors Methods: PBS plans were optimized on average CT using a beam-specific PTV method for 10 consecutive patients with locally advanced non-small-cell-lung-cancer (NSCLC) treated with proton therapy to 6660/180 cGy. End inhalation (CT0) and end exhalation (CT50) were selected as the two extreme scenarios to acquire the relative stopping power ratio difference (Δrsp) for a respiration cycle. The water equivalent difference (ΔWET) per radiological path was calculated from the surface of patient to the iCTV by integrating the Δrsp of each voxel. The motion magnitude of each voxel within the target follows a quasi-Gaussian distribution. A motion index (MI (>5mm WET)), defined as the percentage of target voxels with an absolute integral ΔWET larger than 5 mm, was adopted as a metric to characterize interplay. To simulate the treatment process, 4D dose was calculated by accumulating the spot dose on the corresponding respiration phase to the reference phase CT50 by deformable image registration based on spot timing and patient breathing phase. Results: The study indicated that the magnitude of target underdose in a single fraction plan is proportional to the MI (p<0.001), with larger motion equating to greater dose degradation and standard deviations. The target homogeneity, minimum, maximum and mean dose in the 4D dose accumulations of 37 fractions varied as a function of MI. Conclusion: The MI quantification metric can predict the level of dose degradation in PBS lung cancer treatment, which potentially serves as a clinical decision tool to assess whether patients are suitable to receive PBS treatment.

60 APPLIED LIFE SCIENCES↗

Plateau Phenomenon in Gradient Descent Training of RELU Networks: Explanation, Quantification, and Avoidance

The ability of neural networks to provide ‘best in class’ approximation across a wide range of applications is well-documented. Nevertheless, the powerful expressivity of neural networks comes to naught if one is unable to effectively train (choose) the parameters defining the network. In general, neural networks are trained by gradient descent type optimization methods,a stochastic variant thereof. In practice, such methods result in the loss function decreases rapidly at the beginning of training but then, after a relatively small number of steps, significantly slow down. The loss may even appear to stagnate over the period of a large number of epochs, only to then suddenly start to decrease fast again for no apparent reason. This so-called plateau phenomenon manifests itself in many learning tasks. The present work aims to identify and quantify the root causes of plateau phenomenon.analysis is carried out in the setting of univariate ReLU networks. No assumptions are made on the number of neurons relative to the number of training data, and our results hold for both the lazy and adaptive regimes. Here, the main findings are: plateaux correspond to periods during which activation patterns remain constant, where activation pattern refers to the number of data points that activate a given neuron; quantification of convergence of the gradient flow dynamics; and, characterization stationary points in terms solutions of local least squares regression lines over subsets of the training data. Based on these conclusions, we propose a new iterative training method, the Active Neuron Least Squares (ANLS), characterised by the explicit adjustment of the activation pattern at each step, which is designed to enable a quick exit from a plateau. Illustrative numerical examples are included throughout.

97 MATHEMATICS AND COMPUTING↗

Field Testing of a Mixed Potential IoT Sensor Platform for Methane Quantification

Emissions of CH 4 from natural gas infrastructure must urgently be addressed to mitigate its effect on global climate. With hundreds of thousands of miles of pipeline in the US used to transport natural gas, current methods of surveying for leaks are inadequate. Mixed potential sensors are a low cost, field deployable technology for remote and continuous monitoring of natural gas infrastructure. We demonstrate for the first time a field trial of a mixed potential sensor device coupled with machine learning and internet-of-things platform at Colorado State University’s Methane Emissions Technology Evaluation Center (METEC). Emissions were detected from a simulated buried underground pipeline source. Sensor data was acquired and transmitted from the field test site to a remote cloud server. Quantification of concentration as a function of vertical distance is consistent with previously reported transport modelling efforts and experimental surveys of methane emissions by more sophisticated CH 4 analyzers.

03 NATURAL GAS↗

Evaluating Probabilistic Deep Learning Methods for Uncertainty Quantification of Precipitation Bias Correction

Climate models often exhibit biases in their precipitation predictions, particularly underestimating high-intensity events and overestimating low precipitation. Deep learning approaches offer promising solutions, but their epistemic uncertainty associated with a deep learning–based bias correction method has not previously been quantified for reliable downstream climate impact studies. While methods for capturing the epistemic uncertainty in deep learning frameworks exist, there is currently no consensus on the best method. In this work, we compare three uncertainty quantification (UQ) methods—Deep Ensembles (DEns), Monte Carlo Dropout (MCD), and Flipout—by assessing the reliability of their uncertainty estimates using standard measures such as sharpness and calibration. These UQ methods are applied to an existing deep learning precipitation bias correction model known as UFNet: a coupled U-Net and fully connected neural network. The methods utilized to assess the models’ uncertainties are 1) calibration, which ensures that the expected probabilities of the model align with reality and 2) sharpness, which is a measure of the precision of the model’s probabilistic predictions. Of the three UQ methods evaluated, the DEns and MCD methods demonstrated the best-calibrated performance (expected calibration error of 0.36 and 0.35, respectively), compared to Flipout (0.58). In contrast, Flipout had the sharpest predictions and the highest metric performance in bias correcting precipitation—especially for higher-order moments such as kurtosis with a spatial correlation of 72% compared to 32% and 55% spatial correlation for DEns and MCD, respectively. Of the three UQ methods, MCD was found to be the most suitable method for UQ purposes based on its calibration, sharpness, and computational requirements.

Bayesian methods↗