Volume 34, Issue 6, June 2007
Index of content:
- Imaging Scientific Session: Room L100F
- Image and Observer Performance‐Modeling and Measurement
34(2007); http://dx.doi.org/10.1118/1.2761628View Description Hide Description
Purpose: To quantify veiling glare and determine the origins of the MTF low frequency drop (lfd) in indirect x‐ray flat panel detectors(FPDs), by measuring the amplitude of lost scintillation light and its characteristic propagation length. Methods and Materials: A series of radio‐opaque lead disks (1 to 35mm radius) were imaged on a GE structured CsI FPD, under different x‐ray spectra (28 to 49 kVp with Rh/Rh anode/filter, with additional 0.3mm Cu filtration). The signal at the center of the disk as a function of disk radius, the disk transfer function (DTF), was used to determine the amplitude and decay length of scintillation light propagated horizontally in the detector.Results: It is found that there are two additive constituents of glare in this detector. They have similar amplitudes of ∼12%, but have respective propagation lengths of ∼2 and 10mm. The shorter‐range contribution diminishes in amplitude and propagation length with increasing mean x‐ray energy, while the other varies oppositely. By using the DTF it was possible to correct for the MTF lfd. Conclusions: The precise cause of the two additive contributions to glare is not yet known. Future investigations with monoenergetic x‐ray beams will allow precise mapping of the energy dependence of the veiling glare. A better understanding of the physics of scintillation light propagation will allow the optimization of transmission properties in indirect detectors.
TH‐C‐L100F‐02: Instrumentation Noise Equivalent Exposure (INEE) and the Effect of Detector Blurring and Image Post‐Process Smoothing: A Simulation Study34(2007); http://dx.doi.org/10.1118/1.2761629View Description Hide Description
Purpose: To investigate the behavior of Instrumentation‐Noise‐Equivalent Exposure (INEE) when there is detector blurring and image post‐process smoothing. Method and Materials: INEE is the exposure at which detectedquantum noise and instrumentation noise are equal, and below which the system becomes instrumentation noise limited. In order to understand the effect of detector blurring and post‐process smoothing on the determination of INEE, Poisson‐distributed random numbers (representing input x‐ray quanta per pixel) were generated that simulate an ideal input x‐ray image pattern. Blurring functions were simulated by Gaussian pointspread functions with different full‐width‐half‐maxima (FWHM), defined in a 15 pixel × 15 pixel kernel. The reference pixel‐size and x‐ray fluence were selected following a set of measurements on a custom Microangiographic detector (43 μm square‐image‐pixels in 1024×1024 pixel matrix). A two‐dimensional discrete convolution of the Poisson‐distributed image with the Gaussian blurring function results in an image with reduced total‐noise, but constant signal. The instrumentation‐noise was simulated by an additive constant‐variance and zero‐mean noise. Addition of instrumentation noise after the blurring process simulates the detector‐blurring case, whereas addition before the blurring process represents the image‐post‐process smoothing. This study assumed that the system has no secondary‐quantum‐sink and follows Poisson statistics throughout the imaging‐chain. INEE was determined by calculating the signal‐to‐noise ratio (SNR) over a range of input exposures, and analyzed as a function of Gaussian‐width and additivenoise levels. Results: The square‐root of INEE is shown to increase linearly with detector‐Gaussian‐blur width, but is shown to be independent of image‐post‐process smoothing. For this particular study, a one‐pixel increase in FWHM of the blurring function resulted in 1.5x higher output SNR. Conclusion: A simple simulation study was presented to demonstrate that the SNR‐based practical measurement of INEE is independent of image‐post‐process smoothing in digital x‐ray imagingsystems.
(Partial support: NIH R01‐NS43924, R01‐EB002873, Toshiba Corp.).
34(2007); http://dx.doi.org/10.1118/1.2761630View Description Hide Description
Purpose: To investigate the imaging capabilities of two image‐guidedradiotherapy(IGRT) systems, TomoTherapy Hi‐Art™ and Varian Trilogy™. Methods and Materials: A 35 cm diameter phantom, developed in house for overall assessment of imaging capabilities for pelvic radiotherapy, was scanned on the Hi‐Art™ from TomoTherapy Inc. and on Varian Medical System's Trilogy™ system half fan (boby scan) with half bow tie filter. This phantom consisted line pairs for assessing the X, Y and Z resolution, CT density plugs, a steel ball, air surrounded by acrylic, water ovoid in an acrylic base, low and high contrast plugs, and water holes surrounded by acrylic. Moreover all these objects are immersed in water. We used a farmer chamber for dose measurement for both the systems at the centre (isocenter) of the phantom and near the edge of the phantom. Results: The X and Z resolution is better for Trilogy scans while Y resolution was better for Hi‐Art™ scans. Metal artifacts were more pronounced in Trilogy™ scans as compared to Tomo scans. −6% contrast object was more prominent for Trilogy™ scans but for 3% contrast was more visible on the Tomo scans.. Quantitative analysis of CT density plugs and uniform water showed that the CT number uniformity is far better for TomoTherapy images than for Trilogy™ scans. The dose measured at the center of the phantom is more than 3 times higher for Trilogy™ and the dose at the off axis is more than 5 times higher as compared to the TomoTherapy imagingdose.Conclusion: The Trilogy™ and Hi‐Art™ scan perspicuity is similar. However, the CT numbers are more consistent on the Hi‐Art™ unit and it produces fewer metal artifacts. The dose delivered to the patient is much greater on a Trilogy™ than on the Hi‐Art™ unit.
TH‐C‐L100F‐04: A Fast Method for Measurement of Modulation Transfer Function (MTF) and Detective Quantum Efficiency (DQE) in Presence of Phantom Scatter in Image Guided Radiotherapy34(2007); http://dx.doi.org/10.1118/1.2761631View Description Hide Description
Purpose: To present an improved bar‐pattern method that allows for fast measurement of modulation transfer function(MTF) and detective quantum efficiency (DQE) in conventional no‐phantom scatter and phantom scatter conditions to characterize spatial resolution of both 2D and 3D imagingsystems for image guided radiation therapy(IGRT) with both kilovoltage and megavoltage x‐ray sources. Method and Materials: X‐ray imaging requires MTFmeasurements under phantom scatter free condition for DQE measurements. While slit/edge MTFmeasurements are laborious, the bar‐pattern method allows for a simplified approach for discrete MTFmeasurement although its implementation for megavoltage x‐ray imaging may lead to errors due to inadequate normalization. We introduce a new bar‐pattern method based on based on improved normalization techniques that provide extremely accurate MTFmeasurement. It is well suited for MTF and DQE measurements under phantom scatter conditions, which more closely mimic patient imaging and cannot be obtained using slit/edge methods. Comparisons of the bar‐pattern method are presented with slit/edge MTFs. We present a methodology to image bar‐patterns placed on a multi‐axial rotation jig that allows measurement of spatial resolution of not only 2D detectors(EPIDs) but also volumetric imagingsystems, i.e. CBCT and MVCT. Results: Comparisons of MTFmeasured with custom tungsten bar‐patterns show good statistical agreement with measurements with the slit and edge for two clinical EPIDsystems. Conclusions: The proposed bar‐pattern method provides fast accurate measurement of discrete MTF and DQE values, and is much simpler and faster to use than traditional slit and edge methods. This method also allows for MTF and DQE measurements under scatter conditions that more closely mimic patient imaging conditions. It can be used for EPID and CBCT/MVCT imaging quality assurance within minutes of imaging time and is therefore well suited for clinical measurements.
TH‐C‐L100F‐05: Measurement of the MTF of a Flat Panel Detector in Variable Resolution X‐Ray Detection Mode34(2007); http://dx.doi.org/10.1118/1.2761632View Description Hide Description
Purpose: To quantify the spatial resolution of a flat panel detector(FPD) in the variable resolution x‐ray (VRX) detection mode and to understand the dependence of this resolution on the VRX angle, x‐ray energy, and detector in‐plane orientation. Method and Materials: The detector presampling modulation transfer function(MTF) of a FPD was measured in the VRX detection mode, in which the detector was angulated to match its field of view to the size of an object being imaged and thus to improve the spatial resolution. The edge spread function was obtained by exposing a slightly tilted tungsten edge that was attached directly onto the surface of the detector. The x‐ray beam was generated by a micro‐focal x‐ray tube. The VRX angle, the angle between the central x‐ray and the detector plane, was varied by rotating the detector about the vertical axis. The MTF was measured at various VRX angles and beam energies as well as in two orthogonal orientations of the detector. This MTF was divided by the correction MTF resulting from the edge thickness. Results: In the horizontal orientation, the presampling MTF was measured at 10 VRX angles (90°–10°). There was no significant variation in the MTF at angles above 40°, but conspicuous improvements were observed at angles below 40°. The MTF in the vertical orientation did not vary much at various VRX angles (90°–20°). The MTF at 20° and 45° was measured at several x‐ray tube voltages (40–80 kVp), and no distinct dependence on the voltage was observed. Conclusion: The study indicates that the presampling detectorMTF in the horizontal detector orientation can be improved at VRX angles smaller than 40°. The MTF in the vertical orientation remains relatively constant at angles from 90° to 20°. Little dependence of the MTF on the x‐ray tube voltage was found.
34(2007); http://dx.doi.org/10.1118/1.2761633View Description Hide Description
Purpose: Characterization of a number of commercially available terbium activated gadolinium oxysulfide (Gd2O2S:Tb) phosphors with regard to detection efficiency, spatial resolution, and contrast to noise ratio (CNR) were evaluated for different X‐ray energies. Methods and Materials: 16 samples of Gd2O2S:Tb scintillators from different vendors attached to an a‐Si flat panel detector from Perkin Elmer Optoelectronics (XRD 1621AN). Size of each sample of scintillator screen is 10cm × 10cm and the flat panel detector active area is 41cm × 41cm. Two x‐ray energies of 70kV and 6MV were used for screen evaluation. Kyokko PI200 (Kasei Optonix, Japan) and AST Medex Portal (Applied Scintillation Technologies, UK) screens were selected due their highest CNR for MV imaging. To analyze the image quality of MV cone beam CT and to obtain clinical data, a larger size of Kyokko PI200 (41cm × 41cm) was mounted on a Perkin Elmer 1640AN detector for further evaluation. Results:Detective quantum efficiency (DQE), sensitivity, modulation transfer function (MTF), and noise power spectra (NPS) were evaluated for all screens at 70kV. For 6MV, detection efficiency and spatial frequency were analyzed only. Detection efficiency of AST Medex Portal and Kyokko PI200 are 3.2 and 2.0 times of Lanex fast respectively. CNR and f50 were analyzed for Kyokko PI200 by placing a QC‐3V phantom on the surface of detector. The CNR and f50 were 215 and 0.43 lp/mm, respectively. Low contrast and spatial resolutionanalysis of MV cone beam CT images showed that Kyokko PI200 generates equivalent image quality as Lanex fast with only one half of radiation dose. Conclusion: We characterized 16 commercially available scintillators for MV and kV imaging. Compared to Lanex fast, Kyokko PI200 and AST Medex Portal were determined to be good candidates to optimize MV imaging.Conflict of Interest: Sponsored by Siemens and Perkin Elmer.
TH‐C‐L100F‐07: FROC Curves Without Arbitrary Scoring of Marks: Perceptual Vs. Scored Analysis of CAD Data34(2007); http://dx.doi.org/10.1118/1.2761634View Description Hide Description
A free‐response study generates data in the form of mark‐rating pairs. Any mark closer to a lesion than the acceptance radius (or closeness or overlap criterion) is scored as lesion localization and other marks are scored as non‐lesion localizations. A scored FROC curve is a plot of lesion localization fraction vs. non‐lesion localization fraction. Because of the inherently arbitrary nature of the acceptance radius choice, the FROC curve is also arbitrary. A method of statistically modeling the spatial distribution of the marks with respect to lesion centers is described. This model allows one to obtain statistical (i.e., maximum likelihood) estimates of the numbers of marks that were true lesion localizations as opposed to false localizations. These estimates can be used to plot a perceptual FROC curve that is totally independent of acceptance radius choice. The method was applied to several CAD datasets. It was found that the perceptual and scored curves were in good agreement in most cases, and minor deviations were attributed to some cases of questionable classifications of marks. CAD developers could use this method to evaluate their algorithms in a manner that makes inter‐comparisons more meaningful.
TH‐C‐L100F‐08: Simplification of a Standard Method Used for the Measurement of the Sub‐System MTF in Digital Mammography34(2007); http://dx.doi.org/10.1118/1.2761635View Description Hide Description
Purpose: The determination of the sub‐system resolution using the modulation transfer function(MTF) on one manufacturer's digital mammography machine is a requirement for the annual physics survey. The purpose of this study is to reduce the number of steps in the process. Method and Materials: In performing the required test, it is specified that the MTF values are determined at two frequencies for each of eight MTFs. In the manufacturer's protocol, six numbers are required for the determination of these two values using the built in region‐of‐interest (ROI) tool and the image of a line‐pair phantom using a specified geometry. However, using the same images and same ROI tool, these values can be approximated using only three numbers and a simpler formula. The two methods were compared empirically using the same images. Mathematical analysis was also used to determine the accuracy of the approximation. Results: Under the conditions specified by the manufacturer, MTF values obtained using the proposed approximation are virtually identical to those obtained using the standard technique for the large focal spot. For the small focal spot, the approximation may underestimate the specified formula values by up to a few percent. Conclusion: An approximation to the MTF formula specified for a physics digital mammography QC test reduces by half the required number of recorded values. Because the error, if any, underestimates the values obtained using the specified formula, passing the manufacturer's requirements using the approximation guarantees that the system would pass using the original formula. For the tested system (molybdenum and rhodium, large and small focal spot, parallel and perpendicular to the tube axis), use of the approximation decreases the number of recorded values needed from 48 to 24. The result is a substantial saving in the time required to perform the test.
34(2007); http://dx.doi.org/10.1118/1.2761636View Description Hide Description
Operating characteristics summarize observer performance in diagnostic tasks. Examples are the ROC, the LROC, the FROC, and the AFROC. Models of observer performance differ in their abilities to predict operating characteristics. For example, ROC models can predict the ROC curve, but cannot predict LROC, FROC or AFROC curves. Location specific models are, in historical order, the FROCFIT model, Swensson's LROC model and the search model (SM). In principle any of these models can predict all of the characteristic curves noted above. The purpose of this work was to compare the ability of these models to predict operating characteristics. CAD data consisting of mark‐rating pairs on 450 images were used to determine the various operating characteristics. For example, the highest rating on each image determined the ROC curve; the lesions rated higher than the highest noise determined the ordinate of the LROC, etc. All models agreed with the observed ROC data. Swensson's model agreed with the observed LROC curves but departed mildly from the observed AFROC curve and strongly from the observed FROC curve. The FROCFIT model departed mildly from the AFROC curve and strongly from the FROC curve. The search model provided good fits to all operating characteristics. In conclusion this work provides additional validation of the search model method of analyzing free‐response data.
34(2007); http://dx.doi.org/10.1118/1.2761637View Description Hide Description
Purpose: To develop a model that allows for understanding the physical properties and the effect of readout parameters on objective image quality metrics of a ‘flying spot’ computed radiography(CR)system.Method and Materials: A Monte Carlo simulation method [Fasbender et al., Nucl Instrum Methods Phys Res A 512 (2003): 610–8] that describes the optical transport in a ‘flying spot’ readout was implemented by adapting the code of Jacques [Photochem Photobiol 67 (1998), 23–32]. This program was coupled with a ‘parallel‐cascades’ approach [Yao and Cunningham, Med Phys 28 (2000): 2020–38] to predict systemMTF and DQE. Grain size was assumed to follow a normal distribution in the imaging plate (IP) and granular noise was determined using the works of Nutting and Siedentopf. Additional assumptions include, orthogonal x‐ray incidence, no scatter, no glare, no defects in the IP, a constant discharge fraction, and a linear readout amplifier were assumed for a linear and wide‐sense stationary process. X‐ray spectra representative of radiography and mammography were used in the simulations. Published or estimated parameters that correspond to two common IPs were compiled from multiple sources. Results: For radiography with 0.2‐mm sampling, model results showed good agreement with published literature values for systemMTF (r>0.99) and moderate agreement for DQE (r = 0.75 to 0.99) over four orders of incident exposure (0.03 to 30‐mR). For mammography with 0.1‐mm sampling, model results showed good agreement with published literature values for systemMTF (r>0.99) and for DQE (r = 0.95 to 0.99) over two orders of incident exposure (∼1 to 100‐mR). Conclusion: The proposed model facilitates identifying factors that could improve CR performance. Accurate estimates of IP parameters and model improvements such as incorporation of Lubberts' effect could further improve model accuracy. Supported in part by NIH R01EB004015 and Georgia Cancer Coalition award.