Search NASA⌕ Search

SEARCH · Search NASA

Results for “ROC analysis”

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 55 records · Page 3

Relationship of computed tomography perfusion and positron emission tomography to tumour progression in malignant glioma

Introduction: This study aimed to explore the potential for computed tomography (CT) perfusion and 18-Fluorodeoxyglucose positron emission tomography (FDG-PET) in predicting sites of future progressive tumour on a voxel-by-voxel basis after radiotherapy and chemotherapy. Methods: Ten patients underwent pre-radiotherapy magnetic resonance (MR), FDG-PET and CT perfusion near the end of radiotherapy and repeated post-radiotherapy follow-up MR scans. The relationships between these images and tumour progression were assessed using logistic regression. Cross-validation with receiver operating characteristic (ROC) analysis was used to assess the value of these images in predicting sites of tumour progression. Results: Pre-radiotherapy MR-defined gross tumour; near-end-of-radiotherapy CT-defined enhancing lesion; CT perfusion blood flow (BF), blood volume (BV) and permeability-surface area (PS) product; FDG-PET standard uptake value (SUV); and SUV:BF showed significant associations with tumour progression on follow-up MR imaging (P < 0.0001). The mean sensitivity (±standard deviation), specificity and area under the ROC curve (AUC) of PS were 0.64 ± 0.15, 0.74 ± 0.07 and 0.72 ± 0.12 respectively. This mean AUC was higher than that of the pre-radiotherapy MR-defined gross tumour and near-end-of-radiotherapy CT-defined enhancing lesion (both AUCs = 0.6 ± 0.1, P ≤ 0.03). The multivariate model using BF, BV, PS and SUV had a mean AUC of 0.8 ± 0.1, but this was not significantly higher than the PS only model. Conclusion: PS is the single best predictor of tumour progression when compared to other parameters, but voxel-based prediction based on logistic regression had modest sensitivity and specificity.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Diffusion and perfusion MR parameters to assess preoperative short-course radiotherapy response in locally advanced rectal cancer: a comparative explorative study among Standardized Index of Shape by DCE-MRI, intravoxel incoherent motion- and diffusion kurtosis imaging-derived parameters

Purpose: To assess preoperative short-course radiotherapy (SCR) tumor response in locally advanced rectal cancer (LARC) by means of Standardized Index of Shape (SIS) by dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), apparent diffusion coefficient (ADC), intravoxel incoherent motion (IVIM) and diffusion kurtosis imaging (DKI) parameters derived from diffusion-weighted MRI (DW-MRI). Materials and methods: Thirty-four patients with LARC who underwent MRI scans before and after SCR followed by delayed surgery, retrospectively, were enrolled. SIS, ADC, IVIM parameters [tissue diffusion (D{sub t}), pseudo-diffusion (D{sub p}), perfusion fraction (f{sub p})] and DKI parameters [mean diffusivity (MD), mean of diffusional kurtosis (MK)] were calculated for each patient. IVIM parameters were estimated using two methods, namely conventional biexponential fitting (CBFM) and variable projection (VARPRO). After surgery, the pathological TNM and tumor regression grade (TRG) were estimated. For each parameter, percentage changes between before and after SCR were evaluated. Furthermore, an artificial neural network was trained for outcome prediction. Nonparametric sample tests and receiver operating characteristic curve (ROC) analysis were performed. Results: Fifteen patients were classified as responders (TRG ≤ 2) and 19 as not responders (TRG > 3). Seven patients had TRG 1 (pathological complete response, pCR). Mean and standard deviation values of pre-treatment CBFM D{sub p} and mean value of VARPRO D{sub p} pre-treatment showed statistically significant differences to predict pCR. (p value at Mann–Whitney test was 0.05, 0.03 and 0.008, respectively.) Exclusively SIS percentage change showed significant differences between responder and non-responder patients after SCR (p value << 0.001) and to assess pCR after SCR (p value << 0.001). The best results to predict pCR were obtained by VARPRO Fp mean value pre-treatment with area under ROC of 0.84, a sensitivity of 96.4%, a specificity of 71.4%, a positive predictive value (PPV) of 92.9%, a negative predictive value (NPV) of 83.3% and an accuracy of 91.2%. The best results to assess after treatment complete pathological response were obtained by SIS with an area under ROC of 0.89, a sensitivity of 85.7%, a specificity of 92.6%, a PPV of 75.0%, a NPV of 96.1% and an accuracy of 91.2%. Moreover, the best results to differentiate after treatment responders vs. non-responders were obtained by SIS with an area under ROC of 0.94, a sensitivity of 93.3%, a specificity of 84.2%, a PPV of 82.4%, a NPV of 94.1% and an accuracy of 88.2%. Promising initial results were obtained using a decision tree tested with all ADC, IVIM and DKI extracted parameter: we reached high accuracy to assess pathological complete response after SCR in LARC (an accuracy of 85.3% to assess pathological complete response after SCR using VARPRO D{sub p} mean value post-treatment, ADC standard deviation value pre-treatment, MD standard deviation value post-treatment). Conclusion: SIS is a hopeful DCE-MRI angiogenic biomarker to assess preoperative treatment response after SCR with delayed surgery. Furthermore, an important prognostic role was obtained by VARPRO F{sub p} mean value pre-treatment and by a decision tree composed by diffusion parameters derived by DWI and DKI to assess pathological complete response.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

SU-E-E-07: An Adaptable Approach for Education On Medical Physics at Undergraduate and Postgraduate Levels

Purpose: To contribute to the professional profile of future medical physicists, technologists and physicians, and implement an adaptable educational strategy at both undergraduate and postgraduate levels. Methods: The Medical Physics Block of Electives (MPBE) designed was adapted to the Program of B.S. in Physics. The conferences and practical activities were developed with participatory methods, with interdisciplinary collaboration from research institutions and hospitals engaged on projects of Research, Development and Innovation (RDI). The scientific education was implemented by means of critical analysis of scientific papers and seminars where students debated on solutions for real research problems faced by medical physicists. This approach included courses for graduates not associated to educational programs of Medical Physics (MP). Results: The implementation of the MPBE began in September 2014, with the electives of Radiation MP and Introduction to Nuclear Magnetic Resonance. The students of second year received an Introduction to MP. This initiative was validated by the departmental Methodological Workshop, which promoted the full implementation of the MPBE. Both postgraduated and undergraduate trainees participated in practices with our DICOM viewer system, a local prototype for photoplethysmography and a home-made interface for ROC analysis, built with MATLAB. All these tools were designed and constructed in previous RDI projects. The collaborative supervision of University’s researchers with clinical medical physicists will allow to overcome the limitations of residency in hospitals, to reduce the workload for clinical supervisors and develop appropriate educational activities. Conclusion: We demonstrated the feasibility of adaptable educational strategies, considering available resources. This provides an innovative way for prospective medical physicists, technologists and radiation oncologists. This strategy can be implemented in several regions without formal programs of MP, like most of developing countries. Starting with undergraduate students would allow to reach appropriate certification faster than most of traditional or alternative approaches for education on MP. The authors acknowledge Radiation Consulting Group, LLC, an Arizona Corporation which promotes the use of ionizing radiation in the healing arts, for the “Oscar Luis Caballero” travel grant. The authors thanks to professors Meisbel Daudinot, David Adame and Alexander Pascau for the practices through imagis, imageROC and ANGIODIN PD3000 respectively.

60 APPLIED LIFE SCIENCES↗

MO-FG-202-06: Improving the Performance of Gamma Analysis QA with Radiomics- Based Image Analysis

Purpose: The use of gamma analysis for IMRT quality assurance has well-known limitations. Traditionally, a simple thresholding technique is used to evaluated passing criteria. However, like any image the gamma distribution is rich in information which thresholding mostly discards. We therefore propose a novel method of analyzing gamma images that uses quantitative image features borrowed from radiomics, with the goal of improving error detection. Methods: 368 gamma images were generated from 184 clinical IMRT beams. For each beam the dose to a phantom was measured with EPID dosimetry and compared to the TPS dose calculated with and without normally distributed (2mm sigma) errors in MLC positions. The magnitude of 17 intensity histogram and size-zone radiomic features were derived from each image. The features that differed most significantly between image sets were determined with ROC analysis. A linear machine-learning model was trained on these features to classify images as with or without errors on 180 gamma images.The model was then applied to an independent validation set of 188 additional gamma distributions, half with and half without errors. Results: The most significant features for detecting errors were histogram kurtosis (p=0.007) and three size-zone metrics (p<1e-6 for each). The sizezone metrics detected clusters of high gamma-value pixels under mispositioned MLCs. The model applied to the validation set had an AUC of 0.8, compared to 0.56 for traditional gamma analysis with the decision threshold restricted to 98% or less. Conclusion: A radiomics-based image analysis method was developed that is more effective in detecting error than traditional gamma analysis. Though the pilot study here considers only MLC position errors, radiomics-based methods for other error types are being developed, which may provide better error detection and useful information on the source of detected errors. This work was partially supported by a grant from the Agency for Healthcare Research and Quality, grant number R18 HS022244-01.

60 APPLIED LIFE SCIENCES↗

SU-E-T-363: Error Detection Comparison of EPID and MLC Log File Based IMRT QA Systems

Purpose: In this study we will compare the ability of three QA methods (Delta4, MU-EPID, Dynalog QA) to detect specific errors. Methods: A Varian Novalis Tx with a HD120 MLC and aS1000 Electronic Portal Imaging Device (EPID) was used in our study. Multi-leaf collimator (MLC) errors, gantry angle and dose errors were introduced into 5 volumetric arc therapy (VMAT) plans. 3D dose distributions calculated with data from the EPID and Dynalog QA methods were compared with the planned dose distribution. The gamma passing percentages as well as percentage error of planning target volume (PTV) dose were used for passing determination. Baselines for gamma passing percentages and PTV dose were established by measuring the original plan 5 times consecutively. Standard passing thresholds as well as thresholds derived from receiver operator characteristic (ROC) analysis and 2 standard deviation (SD) criteria were used. Results: When applying the standard 95% pass rate at 3%/3mm gamma analysis 14, 21 and 8 of 30 errors were detected by the Delta4, MU-EPID and Dynalog QA methods respectively. Thresholds set at 2 SD from our base line measurements resulted in the detection of 18, 9 and 14 of 30 errors for the Delta4, MU-EPID and Dynalog QA methods respectively. When using D2 of the PTV as a metric the Dynalog QA detected 20 of 30 errors while the EPID method detected 14 of 30 errors. Using D98 of the PTV, Dynalog QA detected 13 of 30 while the EPID detected 3 of 30 errors. Conclusion: Although MU-EPID detected the most errors at the standard 95% cutoff it also produced the most false detections in the baseline data. The Dynalog QA was the most effective when the ROC adjusted passing threshold was used. D2 was more effective as a metric for detecting errors than D98.

60 APPLIED LIFE SCIENCES↗

Improvement of prostate cancer detection combining a computer-aided diagnostic system with TRUS-MRI targeted biopsy

Purpose: To validate a novel consensus method, called target-in-target, combining human analysis of mpMRI with automated CAD system analysis, with the aim to increasing the prostate cancer detection rate of targeted biopsies. Methods: A cohort of 420 patients was enrolled and 253 patients were rolled out, due to exclusion criteria. 167 patients, underwent diagnostic 3T MpMRI. Two expert radiologists evaluated the exams adopting PI-RADSv2 and CAD system. When a CAD target overlapped with a radiologic one, we performed the biopsy in the overlapping area which we defined as target-in-target. Targeted TRUS-MRI fusion biopsy was performed in 63 patients with a total of 212 targets. The MRI data of all targets were quantitatively analyzed, and diagnostic findings were compared to pathologist’s biopsy reports. Results: CAD system diagnostic performance exhibited sensitivity and specificity scores of 55.2% and 74.1% [AUC = 0.63 (0.54 ÷ 0.71)] , respectively. Human readers achieved an AUC value, in ROC analysis, of 0.71 (0.63 ÷ 0.79). The target-in-target method provided a detection rate per targeted biopsy core of 81.8 % vs. a detection rate per targeted biopsy core of 68.6 % for pure PI-RADS based on target definitions. The higher per-core detection rate of the target-in-target approach was achieved irrespective of the presence of technical flaws and artifacts. Conclusions: A novel consensus method combining human reader evaluation with automated CAD system analysis of mpMRI to define prostate biopsy targets was shown to improve the detection rate per biopsy core of TRUS-MRI fusion biopsies. Results suggest that the combination of CAD system analysis and human reader evaluation is a winning strategy to improve targeted biopsy efficiency.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

{sup 18}F-FDG PET/CT-imaging of left ventricular assist device infection: a retrospective quantitative intrapatient analysis

Background: Despite the use of {sup 18}F-fluorodeoxyglucose ({sup 18}F-FDG) positron emission tomography/computed tomography (PET/CT), diagnosis of a driveline infection in ventricular assist device (LVAD) recipients remains challenging. Our aim was to evaluate the potential of a baseline {sup 18}F-FDG PET/CT (prior to an infection) for the diagnosis of an LVAD-related infection. Methods: We retrospectively selected all LVAD recipients who had undergone two consecutive whole-body {sup 18}F-FDG PET/CT examinations between January 2010 and December 2016. PET/CT was analyzed qualitatively (uptake pattern) and semi-quantitatively (SUV{sub max} and ∆SUV{sub max}). SUV{sub max} was measured and compared in five distinctive volumes of interest along the LVAD driveline. An SUV{sub max} threshold was calculated. Final diagnosis was made by clinical examination, microbiological parameters, and molecular imaging. Results: Thirty patients were enrolled (mean age 54 ± 12 years; 26 male). Mean difference in SUV{sub max} for all five positions between the first and the second PET/CT along the driveline was significantly higher in patients with an LVAD-related infection (mean ∆SUV{sub max} = 4.38 ± 1.44) compared to those without a driveline infection (mean ∆SUV{sub max} = 0.03 ± 0.43), P < 0.05. Applying ROC analysis, an SUV{sub max} threshold of 3.88 resulted in a sensitivity and specificity of 100%, respectively. There were three distinctive uptake patterns in patients with a driveline infection. Conclusion: PET/CT diagnosis in the context of an LVAD-related infection can be improved by comparison to a baseline examination using a distinctive SUV{sub max} threshold.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Automated contouring error detection based on supervised geometric attribute distribution models for radiation therapy: A general strategy

Purpose: One of the most critical steps in radiation therapy treatment is accurate tumor and critical organ-at-risk (OAR) contouring. Both manual and automated contouring processes are prone to errors and to a large degree of inter- and intraobserver variability. These are often due to the limitations of imaging techniques in visualizing human anatomy as well as to inherent anatomical variability among individuals. Physicians/physicists have to reverify all the radiation therapy contours of every patient before using them for treatment planning, which is tedious, laborious, and still not an error-free process. In this study, the authors developed a general strategy based on novel geometric attribute distribution (GAD) models to automatically detect radiation therapy OAR contouring errors and facilitate the current clinical workflow. Methods: Considering the radiation therapy structures’ geometric attributes (centroid, volume, and shape), the spatial relationship of neighboring structures, as well as anatomical similarity of individual contours among patients, the authors established GAD models to characterize the interstructural centroid and volume variations, and the intrastructural shape variations of each individual structure. The GAD models are scalable and deformable, and constrained by their respective principal attribute variations calculated from training sets with verified OAR contours. A new iterative weighted GAD model-fitting algorithm was developed for contouring error detection. Receiver operating characteristic (ROC) analysis was employed in a unique way to optimize the model parameters to satisfy clinical requirements. A total of forty-four head-and-neck patient cases, each of which includes nine critical OAR contours, were utilized to demonstrate the proposed strategy. Twenty-nine out of these forty-four patient cases were utilized to train the inter- and intrastructural GAD models. These training data and the remaining fifteen testing data sets were separately employed to test the effectiveness of the proposed contouring error detection strategy. Results: An evaluation tool was implemented to illustrate how the proposed strategy automatically detects the radiation therapy contouring errors for a given patient and provides 3D graphical visualization of error detection results as well. The contouring error detection results were achieved with an average sensitivity of 0.954/0.906 and an average specificity of 0.901/0.909 on the centroid/volume related contouring errors of all the tested samples. As for the detection results on structural shape related contouring errors, an average sensitivity of 0.816 and an average specificity of 0.94 on all the tested samples were obtained. The promising results indicated the feasibility of the proposed strategy for the detection of contouring errors with low false detection rate. Conclusions: The proposed strategy can reliably identify contouring errors based upon inter- and intrastructural constraints derived from clinically approved contours. It holds great potential for improving the radiation therapy workflow. ROC and box plot analyses allow for analytically tuning of the system parameters to satisfy clinical requirements. Future work will focus on the improvement of strategy reliability by utilizing more training sets and additional geometric attribute constraints.

60 APPLIED LIFE SCIENCES↗

BCL::SAXS: GPU accelerated Debye method for computation of small angle X-ray scattering profiles

Small angle X-ray scattering (SAXS) is an experimental technique used for structural characterization of macromolecules in solution. In this work, we present BCL::SAXS--an algorithm designed to replicate SAXS profiles from rigid protein models at different levels of detail. We first show our derivation of BCL::SAXS and compare our results with the experimental scattering profile of hen egg white lysozyme. Using this protein we show how to generate SAXS profiles representing: (1) complete models, (2) models with approximated side chain coordinates, and (3) models with approximated side chain and loop region coordinates. We evaluated the ability of SAXS profiles to identify a correct protein topology from a non-redundant benchmark set of proteins. We find that complete SAXS profiles can be used to identify the correct protein by receiver operating characteristic (ROC) analysis with an area under the curve (AUC)>99%. We demonstrate how our approximation of loop coordinates between secondary structure elements improves protein recognition by SAχS for protein models without loop regions and side chains. Agreement with SAXS data is a necessary but not sufficient condition for structure determination. We conclude that experimental SAXS data can be used as a filter to exclude protein models with large structural differences from the native.

59 BASIC BIOLOGICAL SCIENCES↗

Prenatal planning of placenta previa: diagnostic accuracy of a novel MRI-based prediction model for placenta accreta spectrum (PAS) and clinical outcome

Purpose: To investigate the diagnostic accuracy of MRI for placenta accreta spectrum (PAS) and clinical outcome prediction in women with placenta previa, using a novel MRI-based predictive model. Methods: Thirty-eight placental MRI exams performed on a 1.5T scanner were retrospectively reviewed by two radiologists in consensus. The presence of T2 dark bands, myometrial thinning, abnormal vascularity, uterine bulging, placental heterogeneity, placental protrusion sign, placental recess, and percretism signs was scored using a 5-point scale. Pathology and clinical intrapartum findings were the standard of reference for PAS, while intrapartum/peripartum bleeding and emergency hysterectomy defined the clinical outcome. Receiver-operating characteristic (ROC) analysis and discriminant function analysis were performed to test the predictive power of MRI findings for both PAS and clinical outcome prediction. Results: Abnormal vascularity and percretism signs were the two most predictive MRI features of PAS. The area under the curve (AUC) of the predictive function was 0.833 (cutoff 0.39, 67% sensitivity, 100% specificity, p = 0.001). Percretism signs and myometrial thinning were the two most predictive MRI features of poor outcome. AUC of the predictive function was 0.971 (cutoff − 0.55, 100% sensitivity, 77% specificity, p < 0.001). Conclusion: The diagnostic accuracy of MRI, especially considering the combination of the most predictive MRI findings, is higher when the target of the prediction is the clinical outcome rather than the PAS.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Analyzing the post-contrast attenuation of the esophageal wall on routine contrast-enhanced MDCT examination can improve the diagnostic accuracy in response evaluation of the squamous cell esophageal carcinoma to neoadjuvant chemoradiotherapy in comparison with the esophageal wall thickness

Purpose: To evaluate the accuracy of the multidetector computed tomography (MDCT) in the response evaluation of the esophageal squamous cell carcinoma (ESCC) to neoadjuvant chemoradiotherapy (nCRT) by analyzing the thickness and post-contrast attenuation of the esophageal wall after the nCRT. Methods: Contrast-enhanced (CE)-MDCT examinations in portal venous phase of one hundred patients with locally advanced ESCC who received nCRT and underwent esophageal resection and histopathology assessment of tumor regression grade (TRG) were retrospectively analyzed by measuring the maximal thickness and mean density of the esophageal wall in the segment involved by tumor and visually searching for hyperdense foci within it. Diagnostic performance was evaluated using the ROC analysis. Results: Average attenuation of the esophageal wall had stronger diagnostic performance for predicting pathologic complete regression (pCR) (AUC = 0.994; p < 0.001) in relation to maximal esophageal wall thickness (AUC = 0.731; p < 0.001). Maximal esophageal wall thickness ≤ 9 mm and average attenuation of the esophageal wall ≤ 64 HU predicted pCR with the sensitivity, specificity, and overall accuracy of 62.5%, 77.9%, and 73%, and 96.9%, 98.5%, and 98%, respectively. Combination of both cutoff values enabled correct assessment of pCR with the 100% accuracy. Visual detection of the hyperdense focus within the esophageal wall predicted pCR with the sensitivity, specificity, and overall accuracy values of 100%, 94.1%, and 96%, respectively. Conclusion: Visual analysis and measurement of post-contrast attenuation of the esophageal wall after the nCRT can improve diagnostic accuracy of MDCT in the response evaluation of the ESCC to nCRT in comparison with measuring the esophageal wall thickness.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Fast myocardial perfusion imaging with {sup 99m}Tc in challenging patients using conventional SPECT cameras

Background: We attempted to validate the performance of a fast myocardial perfusion imaging (MPI) protocol in diagnostically challenging patients. Methods: 78 patients with ΒΜΙ > 24.9, LVH or three vessels disease underwent two sequential gated-MPI studies. The first at 15 (Early Imaging, EI) and the second at 45 (Late Imaging, LI) minutes post {sup 99m}Tc-injection, at both stress and rest. Counts over heart (H), liver (Liv) and subdiaphragmatic space (Sub) and image quality, and myocardial perfusion and function parameters were compared between the two protocols. Coronary angiography was performed within 2 months from MPI, and ROC analysis was used to compare the diagnostic accuracy for the detection of ≥50% diameter luminal stenosis. Results: Quality was optimal-good in 93% of EI and 98% of LI studies (P = .12), H/Liv and stress H/Sub ratios were similar, but rest H/Sub ratio was lower in EI (P = .009). SSS [10 (0 to 46) vs 9 (0 to 36), P = .006] and SDS [3 (0 to 35) vs 2 (0 to 34), P = .02] were higher in EI protocol. LVEF, motion and thickening scores did not differ between the two protocols. A highly significant (P < .001) linear relationship with clinically negligible mean differences in Bland-Altman analysis was observed for all perfusion and function-related data. Sensitivity (EI 81%, LI 80%) and specificity (65% for both) did not differ (P = .23) between the two protocols. Conclusion: The fast protocol is technically feasible and diagnostically accurate compared to the established protocol in diagnostically challenging patients.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Exosomal miR-93 promotes proliferation and invasion in hepatocellular carcinoma by directly inhibiting TIMP2/TP53INP1/CDKN1A

Highlights: • Exosomal miR-93 is up-regulated both in HCC cell lines' media and HCC patients' serum. • Exosomal miR-93 promotes HCC proliferation and invasion by regulating TIMP2/TP53INP1/CDKN1A. • Exosomal miR-93 is correlated with clinical information including stage, tumor size and predict patients' survival rate. Hepatocellular carcinoma (HCC) is a malignant cancer worldwide; lacking biomarkers for early prognostication contributes to its high lethality. Herein, we report a novel biomarker, exosome delivered miR-93, is up-regulated in HCC cell line media and serum samples of HCC patients. We measured the proliferation and invasion ability of HCC cell lines following exosomal miR-93 treatment. After prediction with online algorithms, we further confirmed that TP53INP1, TIMP2 and CDKN1A are direct targets of miR-93 by dual-luciferase reporter assay. In addition, the diagnostic value of exosomal miR-93 was evaluated by qPCR and ROC analysis. The significant correlation between serum exosomal miR-93 and clinical information including stage, tumor size were observed. Furthermore, the survival differences of HCC patients with high or low miR-93 were statistically significant using Kaplan-Meier analysis. In summary, our work identified exosomal miR-93 as a novel biomarker for both diagnosis and prognosis in HCC.

60 APPLIED LIFE SCIENCES↗

Computerized method for evaluating diagnostic image quality of calcified plaque images in cardiac CT: Validation on a physical dynamic cardiac phantom

Purpose: In cardiac computed tomography (CT), important clinical indices, such as the coronary calcium score and the percentage of coronary artery stenosis, are often adversely affected by motion artifacts. As a result, the expert observer must decide whether or not to use these indices during image interpretation. Computerized methods potentially can be used to assist in these decisions. In a previous study, an artificial neural network (ANN) regression model provided assessability (image quality) indices of calcified plaque images from the software NCAT phantom that were highly agreeable with those provided by expert observers. The method predicted assessability indices based on computer-extracted features of the plaque. In the current study, the ANN-predicted assessability indices were used to identify calcified plaque images with diagnostic calcium scores (based on mass) from a physical dynamic cardiac phantom. The basic assumption was that better quality images were associated with more accurate calcium scores. Methods: A 64-channel CT scanner was used to obtain 500 calcified plaque images from a physical dynamic cardiac phantom at different heart rates, cardiac phases, and plaque locations. Two expert observers independently provided separate sets of assessability indices for each of these images. Separate sets of ANN-predicted assessability indices tailored to each observer were then generated within the framework of a bootstrap resampling scheme. For each resampling iteration, the absolute calcium score error between the calcium scores of the motion-contaminated plaque image and its corresponding stationary image served as the ground truth in terms of indicating images with diagnostic calcium scores. The performances of the ANN-predicted and observer-assigned indices in identifying images with diagnostic calcium scores were then evaluated using ROC analysis. Results: Assessability indices provided by the first observer and the corresponding ANN performed similarly (AUC{sub OBS1}=0.80 [0.73,0.86] vs AUC{sub ANN1}=0.88 [0.82,0.92]) as that of the second observer and the corresponding ANN (AUC{sub OBS2}=0.87 [0.83,0.91] vs AUC{sub ANN2}=0.90 [0.85,0.94]). Moreover, the ANN-predicted indices were generated in a fraction of the time required to obtain the observer-assigned indices. Conclusions: ANN-predicted assessability indices performed similar to observer-assigned assessability indices in identifying images with diagnostic calcium scores from the physical dynamic cardiac phantom. The results of this study demonstrate the potential of using computerized methods for identifying images with diagnostic clinical indices in cardiac CT images.

60 APPLIED LIFE SCIENCES↗

TU-G-BRD-08: In-Vivo EPID Dosimetry: Quantifying the Detectability of Four Classes of Errors

Purpose: EPID dosimetry is an emerging method for treatment verification and QA. Given that the in-vivo EPID technique is in clinical use at some centers, we investigate the sensitivity and specificity for detecting different classes of errors. We assess the impact of these errors using dose volume histogram endpoints. Though data exist for EPID dosimetry performed pre-treatment, this is the first study quantifying its effectiveness when used during patient treatment (in-vivo). Methods: We analyzed 17 patients; EPID images of the exit dose were acquired and used to reconstruct the planar dose at isocenter. This dose was compared to the TPS dose using a 3%/3mm gamma criteria. To simulate errors, modifications were made to treatment plans using four possible classes of error: 1) patient misalignment, 2) changes in patient body habitus, 3) machine output changes and 4) MLC misalignments. Each error was applied with varying magnitudes. To assess the detectability of the error, the area under a ROC curve (AUC) was analyzed. The AUC was compared to changes in D99 of the PTV introduced by the simulated error. Results: For systematic changes in the MLC leaves, changes in the machine output and patient habitus, the AUC varied from 0.78–0.97 scaling with the magnitude of the error. The optimal gamma threshold as determined by the ROC curve varied between 84–92%. There was little diagnostic power in detecting random MLC leaf errors and patient shifts (AUC 0.52–0.74). Some errors with weak detectability had large changes in D99. Conclusion: These data demonstrate the ability of EPID-based in-vivo dosimetry in detecting variations in patient habitus and errors related to machine parameters such as systematic MLC misalignments and machine output changes. There was no correlation found between the detectability of the error using the gamma pass rate, ROC analysis and the impact on the dose volume histogram. Funded by grant R18HS022244 from AHRQ.

60 APPLIED LIFE SCIENCES↗

SU-F-T-97: Outlier Identification in Radiation Therapy Knowledge Modeling

Purpose: To investigate the impact of outliers on knowledge modeling in radiation therapy, and develop a systematic workflow for identifying and analyzing geometric and dosimetric outliers using pelvic cases. Methods: Four groups (G1-G4) of pelvic plans were included: G1 (37 prostate cases), G2 (37 prostate plus lymph node cases), and G3 (37 prostate bed cases) are all clinical IMRT cases. G4 are 10 plans outside G1 re-planned with dynamic-arc to simulate dosimetric outliers. The workflow involves 2 steps: 1. identify geometric outliers, assess impact and clean up; 2. identify dosimetric outliers, assess impact and clean up.1. A baseline model was trained with all G1 cases. G2/G3 cases were then individually added to the baseline model as geometric outliers. The impact on the model was assessed by comparing leverage statistic of inliers (G1) and outliers (G2/G3). Receiver-operating-characteristics (ROC) analysis was performed to determine optimal threshold. 2. A separate baseline model was trained with 32 G1 cases. Each G4 case (dosimetric outliers) was then progressively added to perturb this model. DVH predictions were performed using these perturbed models for remaining 5 G1 cases. Normal tissue complication probability (NTCP) calculated from predicted DVH were used to evaluate dosimetric outliers’ impact. Results: The leverage of inliers and outliers was significantly different. The Area-Under-Curve (AUC) for differentiating G2 from G1 was 0.94 (threshold: 0.22) for bladder; and 0.80 (threshold: 0.10) for rectum. For differentiating G3 from G1, the AUC (threshold) was 0.68 (0.09) for bladder, 0.76 (0.08) for rectum. Significant increase in NTCP started from models with 4 dosimetric outliers for bladder (p<0.05), and with only 1 dosimetric outlier for rectum (p<0.05). Conclusion: We established a systematic workflow for identifying and analyzing geometric and dosimetric outliers, and investigated statistical metrics for detecting. Results validated the necessity for outlier detection and clean-up to enhance model quality in clinical practice. Research Grant: Varian master research grant.

60 APPLIED LIFE SCIENCES↗

WE-AB-207B-04: A Preliminary Investigation of Indicators for Treatment Outcomes of CT Guided Cervical Cancer Brachytherapy

Purpose: To identify indicators for recurrence in cervical cancer patients treated with high-dose-rate brachytherapy (HDR-BT). Methods: A total of 37 biopsy proven uterine cervical cancer patients of stage IB-IVA treated between 2011 and 2015 were included in this study. All patients were treated with 37.8–52.2Gy of external beam radiotherapy (EBRT) plus 5 × 5.0–6.5Gy of HDR-BT. Patient age, standard deviation of high-risk clinical tumor volume (HR-CTVSD) collected throughout HDR-BT, and D90 (EQD2, α/β=10Gy) of the HR-CTV were investigated as potential indicators for local/distant recurrence using ROC analysis. The optimal cutoff value was identified through the Youden index and was subsequently used to obtain a group assignment for all patients. Another two comparing groups were defined per evidence of post-EBRT tumor shrinkage based on the pre-HDR MR scan. Kaplan-Meier curves were generated for recurrence-free proportions for comparing groups where time was measured from the final HDR treatment date to the date of the final follow-up exam and compared through Cox regression. Results: Patients had a median follow-up of 12 months, where 16% had local pelvic recurrence and 16% had distant recurrence. HR-CTVSD was identified as a statistically significant indicator for recurrence (AUC=0.802, p=0.007) compared to D90 (AUC=0.655, p=0.167) and patient age (AUC=0.683, p=0.103). HR-CTVSD of 6.26cc was chosen as the optimal cutoff value, which was used as the basis for patient group assignment. Patients with HR-CTVSD>6.26cc had a hazard ratio of 3.92 (95% CI-1.37, 11.24; p=0.011) for recurrence compared to HRCTVSD≤ 6.26cc. Patients without evidence of post-EBRT tumor shrinkage compared to those with had a hazard ratio of 4.28 (95% CI-1.14, 16.15; p=0.032) for recurrence. Conclusion: HR-CTVSD was identified as an indicator for recurrence and an optimal value of 6.26cc was established in our study. In addition, patients without evidence of post-EBRT tumor shrinkage demonstrated an increased recurrence rate.

60 APPLIED LIFE SCIENCES↗

WE-FG-206-08: Pulmonary Functional Imaging Biomarkers of NSCLC to Guide and Optimize Functional Lung Avoidance Radiotherapy

Purpose: Functional lung avoidance radiotherapy promises optimized therapy planning by minimizing dose to well-functioning lung and maximizing dose to the rest of the lung. Patients with NSCLC commonly present with co-morbid COPD and heterogeneously distributed ventilation abnormalities stemming from emphysema, airways disease, and tumour burden. We hypothesized that pulmonary functional imaging methods may be used to optimize radiotherapy plans to avoid regions of well-functioning lung and significantly improve outcomes like quality-of-life and survival. To ascertain the utility of functional lung avoidance therapy in clinical practice, we measured COPD phenotypes in NSCLC patients enrolled in a randomized-controlled-clinical-trial prior to curative intent therapy. Methods: Thirty stage IIIA/IIIB NSCLC patients provided written informed consent to a randomized-controlled-clinical-trial ( http://clinicaltrials.gov/ct2/show/NCT02002052 ) comparing outcomes in patients randomized to standard or image-guided radiotherapy. Hyperpolarized noble gas MRI ventilation-defect-percent (VDP) (Kirby et al, Acad Radiol, 2012) as well as CT-emphysema measurements were determined. Patients were stratified based on quantitative imaging evidence of ventilation-defects and emphysema into two subgroups: 1) tumour-specific ventilation defects only (TSD), and, 2) tumour-specific and other ventilation defects with and without emphysema (TSD{sub VE}). Receiver-operating-characteristic (ROC) curves were used to characterize the performance of clinical measures as predictors of the presence of non-tumour specific ventilation defects. Results: Twenty-one out of thirty subjects (70%) had non-tumour specific ventilation defects (TSD{sub VE}) and nine subjects had ONLY tumour-specific defects (TSD). Subjects in the TSD{sub VE} group had significantly greater smoking-history (p=.006) and airflow obstruction (FEV{sub 1}/FVC) (p=.001). ROC analysis demonstrated an 87% classification rate for smoking pack-years, 90% for FEV{sub 1}/FVC, and 56% for tumour RECIST measurements for identifying patients with non-tumour and tumour-specific ventilation abnormalities. Conclusion: 70% of NSCLC patients had ventilation abnormalities stemming from emphysema, airways disease and tumour burden. Smoking-history and airflow obstruction, but not RECIST, identified NSCLC patients with ventilation abnormalities appropriate for functional lung avoidance therapy.

60 APPLIED LIFE SCIENCES↗