Index of content:
Volume 24, Issue 6, June 1997
24(1997); http://dx.doi.org/10.1118/1.598002View Description Hide Description
A theoretical approach known as quantum accounting diagram (QAD) analysis has been used to calculate the spatial-frequency-dependent detective quantum efficiency (DQE) of two portal imagingsystems: one based on a video camera and another based on an amorphous silicon array. The spatial frequency-dependent DQEs have then been used to determine indices of displayed and perceived image quality. These indices are figures of merit that can be used to optimize the design of linear imagingsystems. We have used this approach to determine which of eight phosphor screen thicknesses (ranging between 67 and ) is optimal for the two designs of portal imagingsystems. The physical characteristics (i.e., detection efficiencies, gains, and MTFs) of each of the eight x-ray detectors have been measured and combined with the physical characteristics of the remaining components to calculate the theoretical DQEs. In turn, the DQEs have been used to calculate theoretical indices of displayed and perceived image quality for two types of objects: a pelvis object and a pointlike object. The maximal indices of displayed and perceived image quality were obtained with screen thicknesses ranging between 358 and depending upon the imagingsystem design and the object being imaged. Importantly, the results showed that there is no single optimal screen thickness. The optimal thickness depended upon imaging task (e.g., detecting large, low-contrast structures, or detecting edges and small structures). Nevertheless, the results showed that there were only modest improvements in the indices of image quality for phosphor screens thicker than
24(1997); http://dx.doi.org/10.1118/1.598009View Description Hide Description
The quality of images generated with radiographic imagingsystems can be degraded if an inadequate number of secondary quanta are used at any stage before production of the final image. A theoretical technique known as a “quantum accounting diagram” (QAD) analysis has been developed recently to predict the detective quantum efficiency (DQE) of an imagingsystem as a function of spatial frequency based on an analysis of the propagation of quanta. It is used to determine the “quantum sink” stage(s) (stages which degrade the DQE of an imagingsystem due to quantum noise caused by a finite number of quanta), and to suggest design improvements to maximize image quality. We have used this QAD analysis to evaluate a video-based portal imagingsystem to determine where changes in design will have the most benefit. The system consists of a thick phosphor layer bonded to a 1 mm thick copper plate which is viewed by a T.V. camera. The imagingsystem has been modeled as ten cascaded stages, including: (i) conversion of x-ray quanta to light quanta; (ii) collection of light by a lens; (iii) detection of light quanta by a T.V. camera; (iv) the various blurring processes involved with each component of the imagingsystem; and, (v) addition of noise from the T.V. camera. The theoretical DQE obtained with the QAD analysis is in excellent agreement with the experimental DQE determined from previously published data. It is shown that the DQE is degraded at low spatial frequencies (<0.25 cycles/mm) by quantum sinks both in the number of detected x rays and the number of detected optical quanta. At higher spatial frequencies, the optical quantum sink becomes the limiting factor in image quality. The secondary quantum sinks can be prevented, up to a spatial frequency of 0.5 cycles/mm, by increasing the overall system gain by a factor of 9 or more, or by improving the modulation transfer function (MTF) of components in the optical chain.
24(1997); http://dx.doi.org/10.1118/1.598003View Description Hide Description
There is currently much research interest in developing, evaluating, and verifying intensity-modulation techniques. Of particular interest is how well the delivery of intensity-modulated profiles can be simulated by planning algorithms, and how accurately these profiles can be delivered given the specification constraints of linear accelerators. In this paper we present a planning and verification study based on delivering radiation in “static-tomotherapy” mode via the NOMOS MIMiC (Multileaf intensity-modulation collimator), which sheds some light on these issues. An inverse-planning algorithm was used to compute intensity-modulated profiles for a 9-coplanar-field plan for a body phantom. The algorithm makes several approximations about the form of the elementary fluence profile through bixels during delivery. Specifically, it is independent of the state of adjacent bixels (i.e., open or closed) and obeys the superposition principle. From the standpoint of comparing the predicted versus the delivered dose, these assumptions were made irrelevant by a final one-step forward dose calculation performed using the optimized intensity profiles. This forward dose calculation took into account the penumbral characteristics of the delivery system by decomposing the intensity profiles into the set of delivery components. Each component was assigned the appropriate penumbral functions thereby ensuring that the calculated dose distribution closely predicted the delivered dose distribution. The nine intensity modulated fields were delivered to a perspex phantom with the same geometry, containing a verification film. In general good agreement was found between the predicted and the measured delivered dose distributions. All the main features of the predicted dose distribution are seen in the delivered. The 90% isodoses were consistently in spatial agreement to within 3 mm. At the 50% isodose level consistent spatial agreement was again found to within 3 mm, the largest deviation being about 5 mm. The close correspondence between the predicted and measured dose distribution demonstrates the potential of the MIMiC delivery system. Our results indicate the level of dose conformation that is achievable in practice and the accuracy of the dose computation algorithm. However, this study only concerned delivery of radiation to a 2 cm thick slice, and the dose distribution was only verified in the central plane of the phantom where the film was placed. We therefore cannot comment as yet on what happens to the dose distribution away from the central film-plane.
24(1997); http://dx.doi.org/10.1118/1.598004View Description Hide Description
Photon-induced proton and alpha particle production in tissue is estimated for the photon energy range from 3 to 28 MeV, using the authors’ previously established methods. It is shown that charged particle emission exceeds neutron emission for energies greater than 11 MeV by a factor that reaches a maximum of 7.0 at 17 MeV. Due to uncertainties in the source data this maximum value should be regarded as indicative only. Above 17 MeV the neutron yield rises sharply and the ratio of charged particle emission to neutron emission declines to values of 3.0 and 1.7 at 20 and 28 MeV, respectively.
Extension of a numerical algorithm to proton dose calculations. I. Comparisons with Monte Carlo simulations24(1997); http://dx.doi.org/10.1118/1.598010View Description Hide Description
A numerical algorithm originally developed for electron dose calculations [Med. Phys. 21, 1591 (1994)] has been modified for use with proton beams. The algorithm recursively propagates the proton distribution in energy, angle, and space from one level in an absorbing medium to another at slightly greater depth until all protons stop. Vavilov’s theory is used to predict, at any point in the absorber, the broadening of the primary proton energy spectrum. Moliere’s theory is applied to describe the angular distribution, and it is shown that the Gaussian first term of Moliere’s series expansion is of sufficient accuracy for dose calculations. These multiple scattering and energy loss distributions are sampled using equal probability spacing to optimize computational speed while maintaining calculational accuracy. Inelastic nuclear collisions along the proton trajectories are modeled by a simple exponential extinction. Predictions of the algorithm for absolute dose deposition by a 160 MeV initially monoenergetic proton beam are compared with the results of Monte Carlo simulations performed with the PTRAN code. The excellent level of agreement between the results of these two methods of dose calculation ( dose and spatial deviations) demonstrate that dose deposition from proton beams may be computed to high accuracy using this algorithm without the need for extensive empirical measurement as input.
24(1997); http://dx.doi.org/10.1118/1.597984View Description Hide Description
Neutron sources created by 4-, 3.5-, and 3-MeV protons striking a thick beryllium target were studied via the time-of-flight technique. Protons were accelerated by the Peking University 4.5 MV electrostatic accelerator. Two disk-shaped targets with thickness 1.5 and 3 mm were used in the measurements. The time-of-flight spectra were observed at zero degrees with respect to the incident proton beam. The analysis to these time-of-flight spectra is given. The time-of-flight spectra were converted to the energy spectra and compared to a neutron spectrum of reaction with incident energy 2.5 MeV, which was also measured in this work. Restricted by the spectrometer itself, the threshold of the measurements is 400 keV. The results show that by using several MeV protons bombarding a thick beryllium target, reactions other than produce significant contributions to the neutron yield with energy less than 1 MeV.
An automatic six-degree-of-freedom image registration algorithm for image-guided frameless stereotaxic radiosurgery24(1997); http://dx.doi.org/10.1118/1.598005View Description Hide Description
A frameless radiosurgical treatment system has been developed by coupling an orthogonal pair of real-time x-raycameras to a robotically manipulated linear accelerator to guide the therapy beam to treatment sites within a patient’s cranium. The two cameras observe the position and orientation of the patient’s head in the treatment system coordinate frame. An image registration algorithm compares the two real-time radiographs to a corresponding pair of digitally synthesized radiographs derived from a CT study of the patient. The algorithm determines all six degrees of translational and rotational difference between the position of the head in the CT coordinate frame and its position in the treatment room coordinate frame. This allows translation of treatment planning coordinates into treatment room coordinates without rigidly fixing the patient’s head position during either the CT scan or treatment. In this paper the image registration algorithm is described and measurements of the precision and speed with which the process can determine the patient’s position are reported. The tests have demonstrated translational uncertainty of 0.5–1.0 mm per axis and rotational uncertainty of 0.6–1.3 degrees per axis, accomplished in approximately 2 s elapsed time.
24(1997); http://dx.doi.org/10.1118/1.598006View Description Hide Description
The availability of digital radiographicimaging on simulators has led to the investigation of a number of new imaging possibilities, including digital linear tomography(tomosynthesis). It has been shown that a single set of projections in this case is sufficient to reconstructimages from multiple planes, including planes tilted with respect to the tube motion. The present work examines the feasibility of tomosynthetic image reconstruction in transverse planes using a CCD-based digital radiotherapy simulator with conventional isocentric rotational geometry. General transformation equations were derived to permit image reconstruction in arbitrary transverse planes. Transverse images of the skull section of the humanoid phantom have been generated using a 360° gantry sweep. Bone, air, and radiographic markers are well resolved, but the image quality is poor due to the suboptimal scanning geometry available on the simulator.
24(1997); http://dx.doi.org/10.1118/1.598000View Description Hide Description
Fixed-separation plane-parallel ionization chambers have been shown to overestimate the dose in the buildup region of normally incident high-energy photon beams. This work shows that these ionization chambers exhibit an even greater over-response in the buildup region of obliquely incident photon beams. This over-response at oblique incidence is greatest at the surface of the phantom and increases with increasing angle of beam incidence. In addition, the magnitude of the over-response depends on field size, beam energy, and chamber construction. This study shows that plane-parallel ionization chambers can over-respond by more than a factor of 2.3 at the phantom surface for obliquely incident high-energy photon fields.
24(1997); http://dx.doi.org/10.1118/1.597985View Description Hide Description
Cumulative radiation damage to siliconsemiconductor diodedetectors can induce dose rate dependent sensitivity, a concern in the pulsed beam of a linac. Two p-Si diodephotondetectors were used in this study, diodes A and B. Both were preirradiated by the supplier to 5 kGy, with diode A receiving an estimated 8 kGy from measurements, and diode B, 25 kGy. At 6 MV, the PDD measured with diode B was lower (by 4.4% at a depth of 25 cm) than diode A. Using SSD to vary the dose per pulse from 0.02 to 0.64 mGy/pulse, diode A was dose rate independent (within 2%), while the sensitivity of diode B changed by 13%. Silicon diode detectors should be checked regularly against ionization chambers in the pulsed beam of a linac, especially older high-resistivity diodes that have accumulated dose from high-energy photon beams.
Effective atomic numbers of composite materials for total and partial interaction processes for photons, electrons, and protons24(1997); http://dx.doi.org/10.1118/1.598001View Description Hide Description
Effective atomic numbers for total and dominant partial interaction processes of photons (1–50 MeV), electrons (1–50 MeV), and protons (1–200 MeV) for the composite materials bone (cortical), muscle (striated), water, polystyrene, Perspex, and Nylon-6 are derived. For photons, the effective atomic number from pair production in the nuclear field is greater than it is from the incoherent scattering. For electrons the effective atomic number from the radiative losses is greater than it is from the collision losses. In both of these cases however, the effective atomic numbers from partial interaction processes remain more or less the same, whereas the number from the total interaction increases with increasing energy. But in the energy regions from 1 to 5 MeV for photons and from 1 to 10 MeV for electrons, the number from the total interaction remains approximately the same for each of these composite materials. For all these materials, in these energy regions the interaction is predominantly with atomic electrons and the contributions from the pair production for photons and radiative losses for electrons are small. In the case of protons the number from total interaction remains more or less the same in the energy region considered. In this energy region collisions with atomic electrons dominate, and the contribution to the total stopping power is mainly from this process only. Hence the derived effective atomic number is basically from the partial process involving the interactions with atomic electrons. Thus, for photons from 1 to 5 MeV for electrons from 1 to 10 MeV and for protons from 1 to 200 MeV, the dosimetric data collected with composite tissue equivalent phantoms, designed on the basis of interaction with atomic electrons for treatment planning, will have less uncertainty.
24(1997); http://dx.doi.org/10.1118/1.598007View Description Hide Description
Direct digital capture systems are relatively new in diagnostic imaging. Full utilization of these devices requires a thorough understanding of the image formation process. The conversion of x-rayphoton energy to a digital pixel value in a commercially available photostimulable phosphor (PSP) imaging system is investigated in this paper. Pixel values measured at 16 different combinations of 4 x-ray beam peak voltages (60, 80, 100, and 120 kVp) and 4 beam qualities are reported. At 60 and 80 kVp exposures were made at (1 mR); at 100 and 120 kVp exposures were made at (2 mR). Analysis of variance was used to determine the statistical significance of the relationship between pixel value and beam quality for a given kVp and exposure. A computer model accounting for x-ray spectral effects that accurately predicts pixel value is presented. Calculated pixel values agree within 5.0% of measured values over the range of beam energies, exposures, and qualities.
24(1997); http://dx.doi.org/10.1118/1.598008View Description Hide Description
A frequency domain light transport model to simulate the transillumination of a scattering object with radio frequency intensity modulated light is presented. The model is based on the diffusion approximation to the radiative transfer equation and uses a finite-element model to allow for complex geometries and an inhomogeneous distribution of absorption and scattering. It calculates the complex photon density within the object and the complex exitance on the boundary of the object. The model is validated against an analytic Green’s function model for a circular geometry in the homogeneous case, and its accuracy is investigated for a range of mesh resolutions, optical parameters, and modulation frequencies.
False-positive reduction technique for detection of masses on digital mammograms: Global and local multiresolution texture analysis24(1997); http://dx.doi.org/10.1118/1.598011View Description Hide Description
We investigated the application of multiresolution global and local texture features to reduce false-positive detection in a computerized mass detection program. One hundred and sixty-eight digitized mammograms were randomly and equally divided into training and test groups. From these mammograms, two datasets were formed. The first dataset (manual) contained four regions of interest (ROIs) selected manually from each of the mammograms. One of the four ROIs contained a biopsy-proven mass and the other three contained normal parenchyma, including dense, mixed dense/fatty, and fatty tissues. The second dataset (hybrid) contained the manually extracted mass ROIs, along with normal tissue ROIs extracted by an automated Density-Weighted Contrast Enhancement (DWCE) algorithm as false-positive detections. A wavelet transform was used to decompose an ROI into several scales. Global texture features were derived from the low-pass coefficients in the wavelet transformed images. Local texture features were calculated from the suspicious object and the peripheral subregions. Linear discriminant models using effective features selected from the global, local, or combined feature spaces were established to maximize the separation between masses and normal tissue. Receiver Operating Characteristic (ROC) analysis was conducted to evaluate the classifier performance. The classification accuracy using global features were comparable to that using local features. With both global and local features, the average area, under the test ROC curve, reached 0.92 for the manual dataset and 0.96 for the hybrid dataset, demonstrating statistically significant improvement over those obtained with global or local features alone. The results indicated the effectiveness of the combined global and local features in the classification of masses and normal tissue for false-positive reduction.
Computerized analysis of interstitial disease in chest radiographs: Improvement of geometric-pattern feature analysis24(1997); http://dx.doi.org/10.1118/1.598012View Description Hide Description
We have been developing automated computerized schemes to assist radiologists in interpreting chest radiographs for interstitial disease based on texture analysis and geometric-pattern feature analysis. In this study, we attempted to improve the performance of the geometric-pattern feature analysis, because the current classification performance with geometric-pattern feature analysis is considerably lower than that of texture analysis. In order to improve the performance in distinguishing between normal lungs and abnormal lungs with interstitial disease, we attempted to remove rib edges in regions of interest (ROIs) by using an edge detection technique, and also to reduce false positives by using feature analysis techniques. In addition, the effects of many parameters on classification performance were investigated to identify proper threshold levels, and subsequently the specificity of the geometric-pattern feature analysis was improved from 69.5% to 86.1% at a sensitivity of 95.0%. Using a combined rule-based method with texture analysis and geometric-pattern feature analysis plus the artificial neural network (ANN) method for classification, a high specificity of 96.1% was obtained at a sensitivity of 95.0%.
Theoretical analysis of error propagation in triple-energy absorptiometry: Application to measurement of lead in bone in vivo24(1997); http://dx.doi.org/10.1118/1.597986View Description Hide Description
We propose a three component tissue decomposition for quantifying lead in bone from a mixture of bone and musclein vivo using a triple-energy absorptiometric method. The theoreticaloptimization of this method, by relating signal uncertainty to radiation dose, requires an expression of the signal variance. The error propagation was therefore theoretically modeled for a counting detector, assuming noise dominance by quantum statistics and neglecting covariance between energy levels. A final expression for the lead signal variance at each energy level was obtained via a Jacobian matrix. The Jacobian was maximized by choosing the first energy as low as permissible by dose constraints below the lead edge. A second optimum was achieved when the upper energy was just above and the middle energy was just below the lead edge. While the signal-to-noise ratio (SNR) had similar behavior to that of the Jacobian as a function of middle and upper energies, the SNR was almost constant as a function of lower energy in the 40–60 keV range. Hence, dose could be reduced without SNR loss. A simulated clinical measurement on an adult tibia using a 50 mCi source and a 10 min acquisition time resulted in a standard deviation of 4 μg Pb/g bone mass. This approach can be applied to other systems containing three components, provided there is a edge within the counting energy range.