摘要:

Clinical efficacy of diagnostic radiology for mammographic examinations is critically dependent on source characteristics, detection efficiency, image resolution, and applied high voltage. In this report we focus on means for evaluation of source‐dependent issues including noninvasive determination of the applied high voltage, and characterization of intrinsic spectral distributions which in turn reflect the effects of added filtration and target and window contamination. It is shown that a particular form of x‐ray curved crystal spectrometry with electronic imaging can serve to determine all relevant parameters within the confines of a standard clinical exposure.

ISSN:0094-2405    

DOI:10.1118/1.597845

出版商:American Association of Physicists in Medicine    

年代:1998

数据来源: WILEY

摘要:

We investigated a new approach to feature selection, and demonstrated its application in the task of differentiating regions of interest (ROIs) on mammograms as either mass or normal tissue. The classifier included a genetic algorithm (GA) for image feature selection, and a linear discriminant classifier or a backpropagation neural network (BPN) for formulation of the classifier outputs. The GA‐based feature selection was guided by higher probabilities of survival for fitter combinations of features, where the fitness measure was the areaAzunder the receiver operating characteristic (ROC) curve. We studied the effect of different GA parameters on classification accuracy, and compared the results to those obtained with stepwise feature selection. The data set used in this study consisted of 168 ROIs containing biopsy‐proven masses and 504 ROIs containing normal tissue. From each ROI, a total of 587 features were extracted, of which 572 were texture features and 15 were morphological features. The GA was trained and tested with several different partitionings of the ROIs into training and testing sets. With the best combination of the GA parameters, the average testAzvalue using a linear discriminant classifier reached 0.90, as compared to 0.89 for stepwise feature selection. TestAzvalues with a BPN classifier and a more limited feature pool were 0.90 with GA‐based feature selection, and 0.89 for stepwise feature selection. The use of a GA in tailoring classifiers with specific design characteristics was also discussed. This study indicates that a GA can provide versatility in the design of linear or nonlinear classifiers without a trade‐off in the effectiveness of the selected features.

ISSN:0094-2405    

DOI:10.1118/1.597829

出版商:American Association of Physicists in Medicine    

年代:1998

数据来源: WILEY

摘要:

This paper presents segmentation and classification results of an automated algorithm for the detection of breast masses on digitized mammograms. Potential mass regions were first identified using density‐weighted contrast enhancement (DWCE) segmentation applied to single‐view mammograms. Once the potential mass regions had been identified, multiresolution texture features extracted from wavelet coefficients were calculated, and linear discriminant analysis (LDA) was used to classify the regions as breast masses or normal tissue. In this article the overall detection results for two independent sets of 84 mammograms used alternately for training and test were evaluated by free‐response receiver operating characteristics (FROC) analysis. The test results indicate that this new algorithm produced approximately 4.4 false positive per image at a true positive detection rate of 90% and 2.3 false positives per image at a true positive rate of 80%.

ISSN:0094-2405    

DOI:10.1118/1.597756

出版商:American Association of Physicists in Medicine    

年代:1998

数据来源: WILEY

摘要:

This study was performed to assess density resolution in quantitative computed tomography (CT) of foam and lung. Density resolution, a measure for the ability to discriminate materials of different density in a CT number histogram, is normally determined by quantum noise. In a cellular solid, variations in mass in the volumes sampled by CT cause an additional degradation of density resolution by the linear partial volume effect. The sample volume, which is directly related to spatial resolution, can be varied by choosing different section thicknesses and reconstruction filters. Several polyethene (PE) foams, as simple models of lung tissue, and five patients were investigated using various sample volumes. For the uniform PE foams, density resolution could be directly determined as the full width at half maximum of CT number histograms. Density resolution for foams with cell sizes of 0.8–1.5 mm was dominated by effects caused by the limited sample size, not by quantum noise. The relative magnitudes of density resolution could roughly be explained with a model for a hypothetic random cellular solid. Since lungs are not of uniform density, analysis of patient data was more complicated. A combined convolution least‐squares fit procedure, together with information obtained in the studies of foam, were used to determine density resolution in lung studies. Density resolution, both for foams and lung, was strongly dependent on sample volume, and was quite poor for thin sections and sharp filters. Consequently, histogram‐shape related parameters are sensitive to the spatial resolution chosen on CT. Thin section densitometry, using a 1‐mm section with a standard or high resolution filter, is not recommended except in determining average density. When using thicker sections, an in‐plane spatial resolution similar to section thickness is advised.

ISSN:0094-2405    

DOI:10.1118/1.597757

出版商:American Association of Physicists in Medicine    

年代:1998

数据来源: WILEY

摘要:

In this paper methods used to measure observer performance are reviewed, and a simple general model for finding and reporting target objects in gray‐scale image backgrounds is presented. That model provides the basis for a combined measurement of detection and localization performance in various image‐interpretation tasks, whether by human observers or by realized computer algorithms. The model assumes that (1) an observer's detection response and first choice of target location both depend on the “maximally suspicious” finding on an image, (2) a correct (first‐choice) localization of the actual target occurs if and only if its location is selected as the most suspicious, and (3) a target's presence does not alter the degree of suspicion engendered by any other (normal) image findings. Formalization of these assumptions relates the ROC curve, which measures the ability todiscriminatebetweenimagescontaining targets and images without targets, to the “Localization Response” (LROC) curve, which measures the conjoint ability todetectandcorrectlylocalizethe actual targets in those images. A maximum‐likelihood statistical procedure, developed for a two‐parameter “binormal” version of this model, concurrently fits both the ROC and LROC curves from an observer's image ratings and target localizations for a set of image interpretations. The model's application is illustrated (and compared to standard ROC analysis) using sets of rating and localization data from radiologists asked to search chest films for pulmonary nodules. This model is then extended to multiple‐report (“free‐response”) interpretations of multiple‐target images, under the stringent requirement that an observer's detection capability and criterion for reporting possible targets both remain stationary across images and across the successive reports made on a given image. That extended model yields formulations and predictions for the so‐called “Free‐Response” (FROC) curve, and for a recently proposed “Alternative FROC” (AFROC) curve. Tests of that model's “stationarity” assumptions are illustrated using radiologists' free‐search interpretations of chest films for pulmonary nodules, and they suggest that human observers may often violate those assumptions when making multiple‐report interpretations of images.

ISSN:0094-2405    

DOI:10.1118/1.597758

出版商:American Association of Physicists in Medicine    

年代:1998

数据来源: WILEY

摘要:

This study compared the relative response of various screen‐film and computed radiography (CR) systems to diagnostic radiation exposure. An analytic model was developed to calculate the total energy deposition within the depth of screen and the readout signal generated from this energy for the x‐ray detection system. The model was used to predict the relative sensitivity of several screen‐film and CR systems to scattered radiation as a function of various parameters, such as x‐ray spectra, phantom thickness, phosphor composition, screen thickness, screen configuration (single front screen, single back screen, screen pair), and readout conditions. In addition, measurements of the scatter degradation factor (SDF) for different screen systems by using the beam stop technique with water phantoms were made to verify the model results. Theoretically calculated values of SDF were in good agreement with experimental data. These results are consistent with the common observation that rare‐earth screens generally produce better image quality than calcium tungstate screens and the CR screen.

ISSN:0094-2405    

DOI:10.1118/1.597830

出版商:American Association of Physicists in Medicine    

年代:1998

数据来源: WILEY

摘要:

A new energy selective detector for dual energy computed radiography has been developed that combines many of the advantages of x‐ray tube voltage switching and single exposure double screen detectors. The new active detector utilizes electro‐optical modulation of the response of the storage phosphor screens to allow dual exposure acquisition with no motion of the screens. Electro‐optical modulation can be done rapidly so the detector can acquire the voltage switched images in a short enough time to minimize patient motion artifacts. Voltage switching produces effective detected energy spectra that result in much lower noise for a given patient dose than the effective spectra of double screen detectors. In this paper the new concept is described, optimal voltages and filter materials are determined by computer simulations, and the active detector performance is compared to other energy selective detectors. The new detector provides over 30 times better signal to noise ratio squared (SNR2) for the same dose and over five times better SNR2for the same x‐ray tube loading than a double screen detector. The effects of incomplete erasure of the x‐ray exposures are determined quantitatively. With achievable erasing fractions, the SNR2is over 20 times better for the same dose and over three times better for constant tube loading than a double screen detector. The active detector is also compared to mechanical screen switching. Mechanical switching provides somewhat better SNR2for the same dose, approximately 1.1 times the active detector SNR2at optimal x‐ray tube voltages. The performance is compared with highly absorbing back screens. If these are used in both the active and passive detectors, both detectors' quality factors increase, but the advantage of the active detector over the passive detector decreases with large back screen thicknesses.

ISSN:0094-2405    

DOI:10.1118/1.597831

出版商:American Association of Physicists in Medicine    

年代:1998

数据来源: WILEY

摘要:

Electron beam dose calculations are often based on pencil beam formulas such as the Fermi–Eyges formula. The Fermi–Eyges formula gives an exact solution of the Fermi equation. The Fermi equation can be derived from a more fundamental mathematical model, the linear Boltzmann equation, in two steps. First, the linear Boltzmann equation is approximated by the Fokker–Planck equation. Second, the Fokker–Planck equation is approximated by the Fermi equation. In this paper, we study these approximations. We use a simplified model problem, but choose parameter values closely resembling those relevant in electron beam therapy. Our main conclusions are: (1) The inaccuracy of the Fokker–Planck approximation is primarily due to neglect of large‐angle scattering. (2) When computing an approximate solution to the Fokker–Planck equation by Monte Carlo simulation of a transport process, one should let the polar scattering angle be deterministic. (3) At shallow depths, the discrepancy between the linear Boltzmann and Fokker–Planck equations is far more important than that between the Fokker–Planck and Fermi equations. The first of these conclusions is certainly not new, but we state and justify it more rigorously than in previous work.

ISSN:0094-2405    

DOI:10.1118/1.597832

出版商:American Association of Physicists in Medicine    

年代:1998

数据来源: WILEY

摘要:

The Fermi pencil beam formula and the higher‐order multiple scattering theory due to Jette are shown to result from a perturbative treatment of the linear Boltzmann equation with Fokker–Planck scattering. Using asymptotic one‐dimensional solutions for the transverse integrated (spherical) fluence as well as its variance, approximate higher‐order pencil beam theories are constructed. These simple and explicit formulae are shown, by comparison with benchmark Monte Carlo results, to be significantly more accurate than the Fermi and Jette equations, particularly at large distances from the beam axis.

ISSN:0094-2405    

DOI:10.1118/1.597759

出版商:American Association of Physicists in Medicine    

年代:1998

数据来源: WILEY

摘要:

The temperature sensitivity of the chemical shift of water (approximately 0.01 ppm/°C) provides a potential method to monitor temperature changesinvivoorinvitrothrough the changes in phase of a gradient‐echo magnetic resonance (MR) image. This relation was studied at 1.5 T in gel materials andinvivoin canine brain and muscle tissue, heated with a radio frequency (rf) annular phased array hyperthermia antenna. The rf fields associated with heating (130 MHz) and imaging (64 MHz) were decoupled using bandpass filters providing isolation in excess of 100 dB, thus allowing simultaneous imaging and rf heating without deterioration of the MR image signal‐to‐noise ratio. In a gel, temperature sensitivity of the MR image phase was observed to be (4.41±0.02) phase degrees/°C forTe=20 ms, which allowed temperature changes of 0.22 °C to be resolved for a 50‐mm3region in less than 10 s of data acquisition.Invivo, forTe=20 ms, the temperature sensitivity was (3.2±0.1) phase degrees/°C for brain tissue, (3.1±0.1) phase degrees/°C for muscle, and (3.0±0.2) phase degrees/°C for a muscle tumor (sarcoma), allowing temperature changes of 0.6 °C to be resolved in a 16‐mm3volume in less than 10 s of data acquisition. We conclude that, while the technique is very sensitive to magnetic field inhomogeneity, stability, and subject motion, it appears to be useful forinvivotemperature change measurement.

ISSN:0094-2405    

DOI:10.1118/1.597760

出版商:American Association of Physicists in Medicine    

年代:1998

数据来源: WILEY