Index of content:
Volume 22, Issue 2, February 1995
22(1995); http://dx.doi.org/10.1118/1.597462View Description Hide Description
Mammographiccontrast is commonly evaluated by visualizing small objects of varying size or mass divided by projected area. These qualitative contrast determinations are commonly performed by imaging a phantom like the American College of Radiology accreditation phantom at clinical mammographic settings. However, this contrast assessment does not take into account the kVp of the machine. This work describes a quantitative mammographycontrast threshold test tool which examines light object contrast on a uniform background for a contrast range of 0.32% to 1.38% at 25 kVp. For this mammographycontrast threshold test tool, contrast is defined by ΔI/I=log e (ψ0/ψ b ), where ψ0 is the target energy flux, and ψ b is the background energy flux. Contrast threshold is defined as the lowest contrast value for which the objects are visible. Unlike traditional assessments of mammographiccontrast, this measurement of contrast threshold is kVp corrected. The mammographycontrast threshold test tool is constructed out of common plastics and provides a quantitative means of assessing contrast threshold for individual mammographic units and total mammographic systems.
22(1995); http://dx.doi.org/10.1118/1.597463View Description Hide Description
The precision of quantitative and subjective evaluations of phantom image quality has been studied. Twenty‐seven images of the American College of Radiology (ACR) mammography accreditation phantom were acquired under different x‐ray techniques and digitized. Several quantitative image quality measures were obtained from each image by analyzing microcalcification and nodule target objects in the phantom. All images were also scored subjectively by 8 observers, each of whom provided a count of the number of objects seen in each target class (fibrils, microcalcifications, and nodules). An analysis was performed to predict the subjective measurements from the quantitative measurements and to estimate their variabilities. It was found that the subjective measures could be well predicted by the quantitative measures and that the variance of the quantitative measures was significantly smaller than that of the subjective measure, by almost a factor of 10. The implication for the ACR accreditation program for mammography is that a substantial improvement is possible in the image quality evaluation process by performing computerized analysis of the phantom images in addition to subjective analysis.
22(1995); http://dx.doi.org/10.1118/1.597464View Description Hide Description
A possible first stage in the analysis of the mammographic scene is its segmentation into four major components: background (the nonbreast area), pectoral muscle, fibroglandular region (parenchyma), and adipose region. An algorithm has been developed for this task. It is based on the classification of a feature vector constructed from statistical measures of texture calculated at two window sizes. Separate self‐organizing neural networks are trained on sample data taken from each of the four regions. The feature vectors from the entire mammogram are then classified with the trained networks linked via a decision logic. To overcome the variability of texture between mammograms the algorithm uses data from a mammogram to classify itself in a staged approach consisting of several binary decisions. The training regions for each successive stage are determined from geometric information produced by the previous stages. The dataset in the study consisted of thirty (fifteen pairs) digitized normal mammograms of variable radiographic appearance. As a measure of performance, the outlines of the parenchyma were compared to those drawn by a radiologist experienced in reading mammograms. Comparison of the areas and perimeters generated by the human and computer observers gives a relationship with correlation coefficients of 0.74 and 0.59 for each measure, respectively. The overlapping areas of the parenchymas segmented by the observers normalized by the combined area was also calculated for each case. The mean and standard deviation of this measure was 0.69±0.12.
X‐ray imaging with amorphous selenium: Detective quantum efficiency of photoconductive receptors for digital mammography22(1995); http://dx.doi.org/10.1118/1.597599View Description Hide Description
Factors affecting the zero spatial frequency detective quantum efficiency of photoconductor‐based x‐ray detectors operating in the mammographic energy range are modeled for monoenergetic incident x rays. The problem is separated into two sections: the calculation of the x‐ray absorption and the Swank factor. X‐ray absorption in this energy range, for most practical photoconductors, is dominated by the photoelectric effect. The Swank factor has four components: fluorescence escape, stochastic variations in gain, variations of gain due to incomplete coupling of charge from the photoconductive layer to the detector electrode, and the nonlinear discharge arising from the field‐dependent x‐ray gain, an effect that is unique to photoconductors. Calculations are performed for selenium, which is currently the most technologically advanced photoconductor available for digital x‐ray imaging. For thicknesses of selenium exceeding 50 μm and for energies between 12 and 50 keV, the detective quantum efficiency of this photoconductor is found to exceed that of a conventional Gd2O2S‐based mammographic phosphor screen.
Image feature analysis and computer‐aided diagnosis in mammography: Reduction of false‐positive clustered microcalcifications using local edge‐gradient analysis22(1995); http://dx.doi.org/10.1118/1.597465View Description Hide Description
To improve the performance of a computerized scheme for detection of clustered microcalcifications in digitized mammograms, causes of detected false–positive microcalcification signals were analyzed. The false positives were grouped into four categories, namely, microcalcificationlike noise patterns, artifacts, linear patterns, and others. In an edge‐gradient analysis, local edge‐gradient values at signal‐perimeter pixels of detected microcalcification signals were determined to eliminate false positives that look like subtle microcalcifications or are due to artifacts. In a linear‐pattern analysis, the degree of linearity for linear patterns was determined from local gradient values from a set of linear templates oriented in 16 different directions. Threshold values for the edge‐gradient analysis and the linear‐pattern analysis were determined using a training database of 39 mammograms. It was possible to eliminate 59% and 25%, respectively, of 91 detected false‐positive clusters with loss of only 3% of true‐positive clusters. The combination of the two methods further improved the scheme in eliminating a total of 73% of the false‐positive clusters with loss of 3% of true‐positive clusters. Using these thresholds, the two methods were evaluated on another database of 50 mammograms. 62%, 31%, and 80% of the false‐positive clusters were eliminated with loss of 3% of true‐positive clusters or less, in the edge‐gradient analysis, the linear‐pattern analysis, and the combination of the two methods, respectively. The edge‐gradient analysis and the linear‐pattern analysis can reduce the false‐positive detection rate, while maintaining a high level of the sensitivity.
Effects of undersampling on the proper interpretation of modulation transfer function, noise power spectra, and noise equivalent quanta of digital imaging systems22(1995); http://dx.doi.org/10.1118/1.597600View Description Hide Description
The proper understanding of modulation transfer function(MTF), noise power spectra (NPS), and noise equivalent quanta (NEQ) in digital systems is significantly hampered when the systems are undersampled. Undersampling leads to three significant complications: (1) MTF and NPS do not behave as transfer amplitude and variance, respectively, of a single sinusoid, (2) the response of a digital system to a delta function is not spatially invariant and therefore does not fulfill certain technical requirements of classical analysis, and (3) NEQ loses its common meaning as maximum available SNR2 (signal‐to‐noise) at a particular frequency. These three complications cause the comparisons of MTF and NEQ between undersampled digital systems to depend on the frequency content of the images being evaluated. A tutorial of MTF, NPS, and NEQ concepts for digital systems is presented, along with a complete theoretical treatment of the above‐mentioned complications from undersampling.
A fully automated algorithm for the segmentation of lung fields on digital chest radiographic images22(1995); http://dx.doi.org/10.1118/1.597539View Description Hide Description
A completely automated algorithm is presented which is capable of identifying both the right‐ and left‐lung fields on digitized chest radiographic images. The algorithm is tested on a sample of 802 chest images against lung fields drawn by a human observer. The average accuracies are found to be 0.957±0.003 and 0.960±0.003 for right‐ and left‐lung regions, respectively. To put them into perspective, the results are compared to several other simple segmentation techniques. These include a comparison of two sets of lung fields drawn by the human observer at different times which yielded accuracies of 0.967±0.005 and 0.967±0.004 for right‐ and left‐lung regions, respectively.
Comparison of two methods for accurate measurement of modulation transfer functions of screen‐film systems22(1995); http://dx.doi.org/10.1118/1.597456View Description Hide Description
The modulation transfer function(MTF) of a screen‐film system can be measured by two methods, i.e., a slit method with Fourier transform on the line spread function and a square‐wave response function (SWRF) method. However, it is still uncertain whether MTFs obtained by the two methods are identical. In this study, MTFs of relatively sharp and unsharp screen‐film systems were measured by using the two methods. The slit method provided slightly greater MTF for the relatively sharp system than the SWRF method. However, MTFs of the unsharp system obtained with the two methods were comparable. Generally, the slit method tends to provide reliable results for unsharp systems, whereas the SWRF method is favorable for sharp systems. Accuracy and consistency of these measurements were examined by comparison of experimental and theoretical edge responses derived from the measured MTFs. However, the difference in edge responses obtained by the two methods was relatively small compared with the variation of the measured edge responses, and thus results were considered inconclusive as to whether either of the methods can provide more accurate MTFs. International interlaboratory comparison indicated that the variation in the measured MTFs at six different institutions was relatively large for both methods. However, the MTFs of two screen‐film systems measured by the slit method appear to agree with those by the SWRF method within the variation expected from the interlaboratory comparison.
The spatial resolution performance of a time‐resolved optical imaging system using temporal extrapolation22(1995); http://dx.doi.org/10.1118/1.597457View Description Hide Description
Optical imaging methods are being explored as a potential means of screening for breast cancer. Previous investigations of time‐resolved imaging techniques have suggested that due to the lack of photons with sufficiently small pathlengths, the spatial resolution achievable through a human breast would be unlikely to be better than a centimeter. Experimental results presented here indicate, however, that higher resolution may be achieved by extrapolating the measured temporal distribution of transmitted photons. This is performed using a least‐squares fit between data and an analytic model of photon transport. The spatial resolution of a time‐resolved imagingsystem was evaluated by measuring the edge response produced by an opaque mask embedded in the center of a 51‐mm‐thick, very highly scattering medium. The limiting spatial resolution was improved from about 13 mm to about 5 mm.
22(1995); http://dx.doi.org/10.1118/1.597601View Description Hide Description
22(1995); http://dx.doi.org/10.1118/1.597461View Description Hide Description
The dosimetric characteristics of a multileaf collimator(MLC) retrofitted to a SL25 linear accelerator have been investigated. Central‐axis depth dose, surface dose, penumbra, beam flatness and symmetry, field size factors, beam transmission through leaves and/or diaphragms, and leakage between the leaves were measured. Quantitative measurements of all beam parameters show good agreement with the design specifications of the manufacturer. No changes were observed in flatness, symmetry, penumbra, and penetration for both 6‐ and 25‐MV photon beams when compared to the values for the standard collimator. No significant differences were observed in the penumbra as a function of leaf position. Transmission measurements in areas shielded by either X diaphragms or leaves plus diaphragms are less than 1% of dose within open field. The average leakage between leaves is about 2.5% for 6‐MV and 3.5% for 25‐MV photon beams. The peak value of the leakage at any point between leaves is less than 5%. The dosimetric features of shaped fields using the MLC are comparable to those of alloy shaped fields with the standard SL25 collimator.
22(1995); http://dx.doi.org/10.1118/1.597602View Description Hide Description
Normalized head‐scatter factors were measured with cylindrical beam coaxial miniphantoms and high purity graphite buildup caps for 4‐, 6‐, 10‐, and 24‐MV photon beams at field sizes from 4×4 to 40×40 cm2. The normalized head‐scatter factors determined by the two methods matched well for 4‐ and 6‐MV photon beams. The miniphantom technique produced normalized head‐scatter factors 1.5% and 4.8% lower than the buildup caps for the 10‐ and 24‐MV beams for large field sizes, respectively. At small field sizes, the miniphantom technique produced larger normalized head‐scatter factors than the buildup caps. Measurements made with an electromagnet indicate that a significant portion of the ionization measured in the buildup cap at 24 MV arises from contamination electrons. Measurements made with the miniphantom and magnet found no contamination electron contribution. The miniphantom technique may exclude such contamination electrons, potentially leading to inaccuracies in tissue‐maximum ratios and phantom scatter factors, as well as inaccuracies in monitor unit calculations.