| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Breast Imaging |
1 From the Dept of Radiology, Univ of Colorado Health Sciences Ctr, CB E-030, 4200 E Ninth Ave, Denver, CO 80262 (J.M.L., R.E.H., P.K.I., G.A.S., C.C.K.); the Dept of Radiology, Univ of Massachusetts Medical Center, Worcester (C.J.D., L.J.M., A.K.); and AMC Cancer Research Ctr, Denver, Colo (G.R.C.). From the 1998 RSNA scientific assembly. Received Jan 4, 2000; revision requested Jan 25; revision received Jul 17; accepted Aug 2. R.E.H. supported by grant DAMD17-96-C-6104 from the U.S. Army Breast Cancer Research and Materiel Command. Address correspondence to J.M.L. (e-mail: john.lewin@uchsc.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: At two institutions, 4,945 FFDM examinations were performed in women aged 40 years and older presenting for SFM. Two views of each breast were acquired with each modality. SFM and FFDM images were interpreted independently. Findings detected with either SFM or FFDM were evaluated with additional imaging and, if warranted, biopsy.
RESULTS: Patients in the study underwent 152 biopsies, which resulted in the diagnosis of 35 breast cancers. Twenty-two cancers were detected with SFM and 21 with FFDM. Four were interval cancers that became palpable within 1 year of screening and were considered false-negative findings with both modalities. The difference in cancer detection rate was not significant. FFDM had a significantly lower recall rate (11.5%; 568 of 4,945) than SFM (13.8%; 685 of 4,945) (P < .001, McNemar 
2 model; P < .03, generalized estimating equations model). The positive biopsy rate for findings detected with FFDM (30%; 21 of 69) was higher than that for findings detected with SFM (19%; 22 of 114), but this difference was not significant.
CONCLUSION: No difference in cancer detection rate has yet been observed between FFDM and SFM. FFDM has so far led to fewer recalls than SFM.
Index terms: Breast neoplasms, radiography, 00.321, 00.324, 00.327 • Breast radiography, comparative studies, 00.321, 00.324, 00.327 • Breast radiography, technology, 00.119 • Cancer screening, 00.321, 00.324, 00.327 • Radiography, digital, 00.321, 00.324, 00.327
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Consequently, attempts have been made to develop digital image receptors as a substitute for the screen-film image receptors now used in mammography (3–5). Much of the current clinical experience with digital mammography has been derived from detectors with a small field of view that have been used for stereotactic core biopsies. These digital detectors have spatial resolutions of 5–10 line pairs per millimeter, limited primarily by the pitch of the charge-coupled device and by the use of minimizing optics to allow a 2.5 x 2.5-cm charge-coupled device to cover an adequate field of view for stereotactic imaging (typically 5 x 5 cm). The issues that relate to the performance of digital mammography detectors have been addressed in a number of studies (6–10).
Recently, prototypic whole-breast digital imaging systems have been introduced for clinical testing to compare full-field digital mammography (FFDM) to SFM. The varied designs of these digital imaging systems have been described elsewhere (3,4). The majority of clinical testing with these systems has been performed as part of studies conducted to help the manufacturers of FFDM equipment gain approval from the Food and Drug Administration for their devices (9), and the results of the studies have not been published. These studies use a cohort undergoing diagnostic mammography to provide a sufficient number of cancers for testing. Unfortunately, studies based on such populations are likely to suffer from entry bias if entrance into the cohort is predicated on an abnormal screening mammogram. The present study eliminates that form of entry bias by enrolling a screening cohort. To our knowledge, this study is the first to compare FFDM with SFM in a screening cohort. The goal of our study is to test whether FFDM is superior to, inferior to, or equivalent to SFM for screening for breast cancer. The study protocol is designed to minimize bias from preferential verification of findings detected with SFM ("verification bias") by recommending imaging work-up or biopsy on the basis of positive screening results with either FFDM or SFM.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Image Acquisition
In 4,523 cases (91%), the subject underwent FFDM immediately after SFM. In these cases, the mammograms were acquired by the same technologist. In 422 cases (9%), the subject came to an outlying screening center for SFM and underwent FFDM during a separate visit to the radiology department within 3 days after SFM. In these cases, a different technologist acquired the full-field digital mammogram than the one who acquired the screen-film mammogram. Technique factors were recorded at the time of the SFM examination.
SFM was performed with a commercially available unit (General Electric DMR unit; GE Medical Systems, Milwaukee, Wis) by using a commercial screen-film system (Kodak Min-R 2000; Eastman Kodak, Rochester, NY). Technique factors (peak kilovoltage, radiation dose, target, and filter) for SFM were automatically selected by the unit (Automated Optimization Parameters feature). The contrast mode was used when the compressed breast thickness was less than 5 cm, and the dose mode was used when the compressed breast thickness was greater than 5 cm. Quality control procedures for SFM were in accordance with the guidelines of the Mammography Quality Standards Act.
FFDM was performed with a prototypic system that used an amorphous silicon area detector bonded to a cesium iodide crystal (GE Medical Systems). The pitch of the detector elements was 100 µm, yielding a limiting spatial resolution of 5 line pairs per millimeter. The active area of the detector was 18 x 23 cm, which yielded an image size of 1,800 x 2,304 pixels. The x-ray tube, support, and generator were identical to those in the commercial SFM system. Each digital detector system underwent extensive acceptance testing (7,8). Ongoing quality control for the FFDM systems included daily phantom imaging and weekly flat-field calibration of the detectors at each site.
Technique factors and breast doses for FFDM were matched to those of SFM by using the same target, filter, and peak kilovoltage as used for SFM. Values for radiation dose were matched as closely as possible between the two techniques, but only discrete mAs stations were available with FFDM. When the mAs value used for SFM was between two allowed choices on the FFDM unit, the lower mAs step was used for FFDM. Both FFDM and SFM units allowed the selection of rhodium/rhodium and molybdenum/rhodium target/filter choices in addition to molybdenum/molybdenum. Identical reciprocating grids were used for both SFM and FFDM. For the study, the mean peak kilovoltage was 26.90 kVp with each modality. The mean mAs value was 128.6 mAs for FFDM and 128.9 mAs for SFM.
Rather than trying to match compression force and compressed breast thickness, the technologist was instructed to obtain the best possible positioning and compression with each modality. Compression force and compressed breast thickness were recorded for each view.
Each examination consisted of the two standard screening views of each breast: craniocaudal and mediolateral oblique views. Both 18 x 24-cm and 24 x 30-cm screen-film image receptors were used for SFM. The image receptor size of the FFDM prototypic unit was 18 x 23 cm. If a subject’s breast was too large to fit on the 18 x 24-cm screen-film image receptor, she was advised after SFM that additional exposures would be needed for her FFDM to obtain coverage of each breast in each of the two standard projections. If she chose to continue in the study, her breast was imaged with FFDM by overlapping as many 18 x 23-cm views as needed to cover each breast. This typically resulted in one or two extra exposures per breast.
Image Interpretation
Screen-film images were interpreted with the routine clinical images of that day in a darkened room on a standard mammography alternator with a luminance of at least 3,000 candelas per square meter. The reader knew at the time of interpretation that the examination was part of the study.
Digital images were interpreted in soft copy on a prototypic Unix-based workstation consisting of a computer (UltraSparc; Sun Microsystems, Palo Alto, Calif) with dual 21-inch-high luminance monitors, each capable of displaying 1,800 x 2,300 pixels (Megascan, North Billerica, Mass). These were driven by 4-megabyte video cards with 8-bit output (Dome, Waltham, Mass). The images were interpreted without postprocessing other than an initial window and level setting derived from the image histogram. A gamma function was used to map the 14-bit image data to the 8-bit display data.
The radiologist had full freedom to adjust window and level and to magnify each image interactively. Magnification of each view to x2 power was typically performed for each reading. For the first approximately 200–400 cases, magnification was accomplished with a moving square showing a portion of the image magnified ("mag glass"). Subsequently, each quadrant of each image was magnified in its entirety and examined.
The digital workstation was located in a darkened room away from mammographic alternators. Comparison mammograms were viewed on a standard light box placed next to the digital workstation. The light box was turned off during the detailed evaluation of digital images to avoid glare.
The screen-film mammogram and the digital mammogram were interpreted at the institution at which they were acquired. Interpretation was done independently by board-certified radiologists qualified under the Mammography Quality Standards Act who also interpreted clinical mammograms at that institution. For a given subject, one radiologist interpreted the SFM and another the FFDM. Each reader was blinded to the results of the other and to the images from the other modality. Comparison images, prior reports, and the subject’s history were available to each radiologist for both the SFM and FFDM interpretations. At one institution, residents participated in some of the interpretations in the study but never participated in both the SFM and FFDM readings of a given subject; at the other institution, residents did not participate. Resident participation was not recorded but was more likely to occur with the SFM reading than the FFDM reading. The attending radiologist, of course, gave the final interpretation. A given attending radiologist was required to interpret approximately equal numbers of SFM and FFDM examinations. Both the SFM and the FFDM interpretations were used for clinical management of the patient.
For each finding, the reader was required to give the following information:
Additional Imaging Evaluation of Findings
Additional imaging evaluation of a given finding was performed to establish truth for that finding in a manner that minimized bias toward either modality. Both additional mammographic views and ultrasonographic images could be obtained as indicated. Concordant findings (ie, those detected with both modalities) needing additional mammographic views were imaged with SFM. Discordant findings (ie, those detected with one modality only) needing additional mammographic views were imaged with the modality with which they were detected, with the exception that magnification views were obtained by using SFM only. This exception was because of the inability to easily remove the grid on the FFDM prototype. Thus all calcifications were worked up with SFM.
Review of Discordant Cases
For all findings recalled for additional evaluation on only one of the two modalities, the two radiologists would evaluate the FFDM and SFM images side by side. At this time, the radiologists had the option of dismissing a finding (ie, not working it up) if, after viewing both images, they could determine a benign cause for the finding or believed that there was no reasonable chance that it represented cancer. Dismissed findings were still counted as positive in calculations of recall rate, sensitivity, and other performance measures.
Data Collection
Data were collected on paper forms and then entered into a customized database program (Microsoft Access 2.0; Microsoft, Redmond, Wash). Each subject filled out a history form that detailed their risk factors and demographic data. The technologist reviewed the forms with the subject. The technologists filled out forms that specified the reason for the examination and the technique factors. The radiologists filled out the interpretation forms. Computer data entry was performed by a professional research assistant at each site.
Long-term Surveillance
Each woman in the study is followed for 1 year following her participation in the study to assess for possible development of cancers that were not detected with either modality or were detected at screening but not subjected to biopsy. Surveillance comes for most subjects when they return to the participating institution for subsequent screening mammography. Those who do not return are contacted by telephone or mail to determine if they have undergone subsequent mammography and whether a diagnosis of breast cancer has been made either through mammographic or clinical detection. In addition, for subjects from one site, cases of breast cancer recorded in the state tumor registry will be periodically cross-matched against subjects in the study. Permission to obtain this information was included in the informed consent.
Data Analysis
For calculating sensitivity, a finding was considered to have been called with a given modality if it was assigned a recommendation for immediate work-up, such as additional imaging, obtaining prior images for comparison, or biopsy. These recommendations generally corresponded to a BI-RADS assessment of 0, 4, or 5. A finding detected but given a recommendation of short-interval (6-month) or routine follow-up was considered negative. These recommendations generally corresponded to a BI-RADS assessment of 2 or 3. Each finding was assessed separately to determine truth by imaging work-up and, if warranted, by biopsy. The final assessment of truth for a finding determined benign with additional imaging was modified if a biopsy performed on the finding within 1 year demonstrated cancer.
For purposes of establishing recall rate, an examination was considered positive for recall (a) if immediate work-up was recommended or (b) if comparison studies were recommended and those comparison studies led to additional work-up. For calculating the number of biopsies and the positive biopsy rate, fine-needle aspiration of solid masses, core-needle biopsy, and surgical biopsy were all included. The pathologic classifications of benign, high-risk, and malignant follow the second edition of BI-RADS (11). In this classification, atypical ductal hyperplasia and lobular carcinoma in situ are considered high-risk lesions.
ROC curves were constructed by using the alternative free-response ROC method (12). To use this approach, the value of the highest rated benign finding in each breast was assigned as the false-positive level for that breast. If no finding was called in a given breast, that breast was assigned a false-positive level of 0. Each cancer was assigned the rating given to it; if it was not detected, it was assigned a rating of 0. With alternative free-response ROC analysis, the area under the ROC curve, A1, is analogous to the area under a standard ROC curve, Az, allowing the use of standard ROC analysis techniques (12–15). The 101-point scale was used without binning. Curve areas were integrated by using the trapezoidal rule.
Tests for the significance of the sensitivity, recall rate, positive biopsy rate, and positive predictive value were performed by using the McNemar 2 test and, if found significant with that test, by using a generalized estimating equations model implemented with the PROC GENMOD command in SAS software (SAS Institute, Cary, NC). The latter is a stricter test of significance because it takes into account the variation among readers and adds this random effect to the variance used in testing.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Table 1 summarizes demographic data and risk factors for the women enrolled in the study. The numbers approximate those of the screening population at each institution, with the exception of a relatively smaller number of subjects undergoing their first screening mammography who were enrolled in the study. The biggest difference between the populations at the two institutions was that almost twice as many women at site 1 were receiving hormone replacement therapy.
|
|
|
Table 4 gives the outcome for findings grouped by the modality with which they were called. Sixty-seven (13%) of the 507 FFDM-only findings and 14 (2%) of the 746 SFM-only findings were dismissed at the discrepancy conference. All of the subjects with dismissed findings have been followed for at least 1 year. Three of the FFDM-only findings that had been dismissed became palpable within 1 year. Two were invasive lobular carcinomas. The third was a fibroadenoma. No other dismissed finding has undergone suspicious change or warranted biopsy for any other reason. Most were not present at the follow-up examination. Four other patients refused additional imaging and have unresolved findings. Eighty-three (11%) of the 746 SFM-only findings and 38 (7%) of the 507 FFDM-only findings were biopsied; 31 (16%) of the 195 findings called with both modalities were biopsied.
|
|
Findings Subjected to Biopsy and Cancers Detected
One hundred fifty-two findings were subjected to biopsy in the study, including 23 found to be benign with fine-needle aspiration. Thirty-five cancers were diagnosed. Of these, 26 were invasive, and nine were ductal carcinoma in situ. Nine of the cancers were detected with FFDM only, 10 with SFM only, and 12 with both modalities. None of these subjects were clinically suspected of having cancer at the time of mammography. Four interval cancers were not detected with either modality and were detected clinically within the next 11 months. Two of these were invasive lobular carcinoma, and two were invasive ductal carcinoma. Table 6 gives the outcome of the biopsies grouped by the modality with which the finding was detected at screening.
|
|
Sensitivity, Positive Predictive Value, and ROC Analysis
The sensitivity of FFDM in the detection of screening cancers was 60% (21 of 35). The sensitivity of SFM was 63% (22 of 35). Because not all subjects have been followed for 1 year to obtain complete ascertainment of interval cancers, this represents an upper bound of the sensitivity for each modality. The relative sensitivity of FFDM to SFM was 95% (21 of 22).
The positive predictive value of screening is defined as the fraction of recalled examinations that led to a diagnosis of breast cancer. The positive predictive value of screening with FFDM was 3.7% (21 of 568). The positive predictive value of screening with SFM was 3.2% (22 of 685). The positive biopsy rate is defined as the fraction of biopsies that yielded cancer. The positive biopsy rate for all findings detected with FFDM was 30% (21 of 69); the positive biopsy rate for all findings detected with SFM was 19% (22 of 114).
ROC curves for the two modalities are presented in the Figure. Alternative free-response ROC analysis was performed by considering each breast separately, for a total of 9,716 data points (194 subjects had previously undergone mastectomy). The area under each curve is 0.76.
|
2 model (P < .001) and the generalized estimating equations model (P < .03). The difference in sensitivity between the two modalities lacked significance (P > .5, McNemar 2). The higher positive predictive value of screening for FFDM compared with SFM was found to lack significance (P > .3, McNemar 2), as did the higher positive biopsy rate for findings detected with FFDM as compared with those detected with SFM (P > .08, McNemar 2). |
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
To determine how the advantages and disadvantages of the two modalities contribute to their performance in cancer detection, we are conducting a clinical study. Because it is possible that FFDM may be superior to, equal to, or inferior to SFM, we have designed our study to be able to test for all of these possibilities. To do so necessitates that we use what is essentially a screening population because basing enrollment on a positive clinical (screen-film) mammogram would bias the results toward SFM in terms of sensitivity and toward FFDM in terms of specificity (11). For example, had we selected only the subjects who had positive SFM examinations to receive FFDM, we would have detected only 22 cancers and concluded that SFM was more sensitive than FFDM because it detected all of them, whereas FFDM had detected only 12.
Unfortunately, the use of a screening population results in a relatively low number of cancers, typically 2–10 per 1,000 women screened, depending on the proportion of first-time screens (16). This low cancer rate decreases the power to detect a difference between the modalities and necessitates studying a large population. Our cancer detection rate per 1,000 women screened was 4.5 for SFM. This value is in the expected range for our cohort, which consisted primarily of repeat screens (16–22). What could not be predicted was the 39% increase in the cancer detection rate, to 6.3 per 1,000, when FFDM was used in addition to SFM. This boost in the cancer detection rate from the addition of FFDM increases the power of the study to above that which would be predicted on the basis of published values for yearly incidence of breast cancer.
The study design maximizes statistical power by having each subject undergo both SFM and FFDM, thereby acting as her own control. This design allows the use of statistical methods for matched paired data (23,24). The power of these methods is largely determined by the number of discordant cases. Thus the statistical power is further increased by the surprisingly large fraction of cancers (19 of 31) detected with only one modality.
A potential source of bias in our study is the higher number of FFDM-only than SFM-only findings dismissed at the discrepancy conference: 67 versus 14, respectively. In designing the study, we allowed a finding to not be worked up if, by consensus, both readers believed that it could be explained as benign by looking at the other modality or was extremely unlikely to persist on additional images. This aspect of the protocol was included to try to reduce recalls without decreasing the rate of cancers detected. Surprisingly, out of only 67 FFDM-only findings dismissed, two (3%) turned out to be lobular carcinomas, a positivity rate higher than that in the 440 FFDM-only findings that were worked up. The potential for bias from this result is that there may be other undetected cancers in the remaining dismissed findings. The probability that there is even one more cancer in the group is low, however, especially because all of these women have been followed for at least 1 year. Note that the dismissed findings were counted as if they had been called back. Thus the two cancers are counted as true-positives for FFDM and the remaining findings as false-positives.
The level of disagreement overall was surprisingly high. Eight hundred twenty-one examinations had discordant interpretations, which represented 17% of all examinations (821 of 4,945) and 79% (821 of 1,037) of the examinations called positive with at least one modality. To the extent that this discordance is due to differences in the modality, statistical power will be increased, but to the extent that it is due to factors independent of the modality, such as positioning and interpretation, statistical power will be decreased. It is well known from double-reading studies (25–27) that there is large reader variability in the interpretation of screening mammograms.
Given the advantage in contrast resolution of FFDM and the advantage in spatial resolution of SFM, one might expect that each modality would excel at detecting different types of cancers. FFDM might be expected to be better for finding densities and masses in dense tissue, while SFM might be better for detecting calcifications. Although SFM did recall a larger number of calcification findings, even more than expected given its overall higher recall rate, the number of cancers manifesting as microcalcifications was the same for both modalities. A higher percentage of FFDM-only calcification findings was positive at biopsy. If these trends persist, then an analysis of the reasons for the discordant readings, given by the readers at the discrepancy analysis, can help to delineate whether the difference is due to FFDM being better for distinguishing benign from malignant calcifications on screening views, perhaps because of the ability to magnify the images on the workstation, or is due to superior detection of subtle calcifications with SFM.
The only other trend in the types of cancers detected was a larger percentage of architectural distortion cancers detected only with SFM. This trend is interesting, given the superior spatial resolution of SFM, because detection of architectural distortion depends on resolving fine lines in breast parenchyma. This task is likely more dependent on high spatial resolution than is the detection of microcalcifications, a task that requires only detecting a high-contrast focus, not on resolving its shape. More subjects are needed to determine whether this trend in detection of architectural distortion represents a true difference between the modalities.
The only significant result in our study is the lower recall rate of FFDM, which is caused almost entirely by a lower false-positive rate. With the use of the ROC model for evaluating a test, there are two possibilities for such a difference: either FFDM and SFM have different ROC curves, or they are being interpreted at different points along the same curve, with the operating point of SFM being to the right of that of FFDM. Although a difference in the false-positive rate without a difference in sensitivity implies that the two modalities have different ROC curves, because the power to detect a difference in sensitivity is relatively low, we cannot yet exclude a small difference in sensitivity that would explain the difference in false-positive rate as a shift along the same ROC curve.
To help distinguish operating point shifts from differences in ROC curves, ROC data were collected on a 101-point scale separately from the BI-RADS assessments and recommendations. The finding of nearly identical ROC curves for FFDM and SFM supports the belief that the difference in false-positive rate reflects a shift of operating point along the same ROC curve. Such a shift could be caused by a difference in the reading conditions between SFM and FFDM, so these conditions were kept as equivalent as possible.
One difference at one site was the participation of residents, which, anecdotally, was more likely to occur with SFM than with FFDM. The other site had no resident participation. Because resident participation was not tracked, whether the participation of residents had a noticeable effect on recall rates cannot be determined. Both sites had lower recall rates with FFDM, but the effect was larger at the site that allowed resident participation. Because of this and other possible unappreciated differences in the reading environments, whether the lower recall rate with FFDM is a property of the modality that will be observed in clinical practice or is an artifact of the experimental situation cannot be definitely determined from our data.
In conclusion, a prospective study comparing FFDM to SFM in a screening cohort has been conducted at two mammography sites. Results indicate that the two modalities have indistinguishable sensitivities for cancer detection and indistinguishable ROC curves but that FFDM has a significantly lower recall rate than SFM. Statistical power is limited by the inherently small number of cancers in a screening population. For this reason, a second phase of this study, currently underway, includes a third site and will accrue more subjects to increase statistical power. In addition, a larger multi-institutional trial with a similar design is being planned by the American College of Radiology Investigative Network.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
3 Current address: Georgetown University Medical School, Washington, DC.
R.E.H. and G.R.C. are consultants to GE Medical Systems, advising on issues related to FDA approval of full-field digital mammography. J.M.L., L.J.M., and P.K.I. have acted as readers in separate digital mammography research funded by GE Medical Systems as part of its submission to the FDA for approval of full-field digital mammography.
Abbreviations: BI-RADS = Breast Imaging Reporting and Data System, FFDM = full-field digital mammography, ROC = receiver operating characteristic, SFM = screen-film mammography
Author contributions: Guarantors of integrity of entire study, J.M.L., R.E.H.; study concepts, J.M.L., R.E.H., G.A.S.; study design, J.M.L., R.E.H., C.J.D., G.A.S., L.J.M., A.K., G.R.C.; definition of intellectual content, J.M.L., R.E.H., C.J.D.; literature research, J.M.L., R.E.H.; clinical studies, J.M.L., R.E.H., C.C.K., G.A.S., P.K.I., C.J.D., L.J.M., A.K.; data acquisition, all authors; data analysis, J.M.L., R.E.H., G.R.C., P.K.I.; statistical analysis, J.M.L., R.E.H., G.R.C.; manuscript preparation, J.M.L., R.E.H.; manuscript editing, J.M.L.; manuscript review and final version approval, all authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. B. Hruska, S. W. Phillips, D. H. Whaley, D. J. Rhodes, and M. K. O'Connor Molecular Breast Imaging: Use of a Dual-Head Dedicated Gamma Camera to Detect Small Breast Tumors Am. J. Roentgenol., December 1, 2008; 191(6): 1805 - 1815. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. F. Jafri, R. S. Ayyala, A. Ozonoff, J. Jordan-Gray, and P. J. Slanetz Screening Mammography: Does Ethnicity Influence Patient Preferences for Higher Recall Rates Given the Potential for Earlier Detection of Breast Cancer? Radiology, December 1, 2008; 249(3): 785 - 791. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. MINIGH Quality Assurance in Digital Mammography Radiol. Technol., May 1, 2008; 79(5): 433M - 454M. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. N.A. Tosteson, N. K. Stout, D. G. Fryback, S. Acharyya, B. A. Herman, L. G. Hannah, E. D. Pisano, and for the DMIST Investigators Cost-Effectiveness of Digital Mammography Breast Cancer Screening Ann Intern Med, January 1, 2008; 148(1): 1 - 10. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Skaane, S. Hofvind, and A. Skjennald Randomized Trial of Screen-Film versus Full-Field Digital Mammography with Soft-Copy Reading in Population-based Screening Program: Follow-up and Final Results of Oslo II Study Radiology, September 1, 2007; 244(3): 708 - 717. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Jiang, D. L. Miglioretti, C. E. Metz, and R. A. Schmidt Breast Cancer Detection Rate: Designing Imaging Trials to Demonstrate Improvements Radiology, May 1, 2007; 243(2): 360 - 367. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Skaane, A. Kshirsagar, S. Stapleton, K. Young, and R. A. Castellino Effect of Computer-Aided Detection on Independent Double Reading of Paired Screen-Film and Full-Field Digital Screening Mammograms Am. J. Roentgenol., February 1, 2007; 188(2): 377 - 384. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. T. Yang, C.-J. Lai, G. J. Whitman, W. A. Murphy Jr., M. J. Dryden, A. C. Kushwaha, A. A. Sahin, D. Johnston, P. J. Dempsey, and C. C. Shaw Comparison of full-field digital mammography and screen-film mammography for detection and characterization of simulated small masses. Am. J. Roentgenol., December 1, 2006; 187(6): W576 - W581. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. H. Kim, E. D. Pisano, E. B. Cole, M. R. Jiroutek, K. E. Muller, Y. Zheng, C. M. Kuzmiak, and M. A. Koomen Comparison of calcification specificity in digital mammography using soft-copy display versus screen-film mammography. Am. J. Roentgenol., July 1, 2006; 187(1): 47 - 50. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Schueller, E. Kaindl, W. K. Matzek, F. Semturs, C. Schueller-Weidekamm, and T. H. Helbich Image Quality of a Wet Laser Printer Versus a Paper Printer for Full-Field Digital Mammograms Am. J. Roentgenol., January 1, 2006; 186(1): 38 - 43. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. V. Ongeval, H. Bosmans, and A. Van Steen Current challenges of full field digital mammography Radiat Prot Dosimetry, December 1, 2005; 117(1-3): 148 - 153. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. D. Dershaw Film or digital mammographic screening? N. Engl. J. Med., October 27, 2005; 353(17): 1846 - 1847. [Full Text] [PDF]
|
|
|
|
|
|
T. K. Nishino, X. Wu, and R. F. Johnson Jr. Thickness of Molybdenum Filter and Squared Contrast-to-Noise Ratio per Dose for Digital Mammography Am. J. Roentgenol., October 1, 2005; 185(4): 960 - 963. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Skaane, C. Balleyguier, F. Diekmann, S. Diekmann, J.-C. Piguet, K. Young, and L. T. Niklason Breast Lesion Detection and Classification: Comparison of Screen-Film Mammography and Full-Field Digital Mammography with Soft-copy Reading--Observer Performance Study Radiology, October 1, 2005; 237(1): 37 - 44. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Suryanarayanan, A. Karellas, S. Vedantham, S. M. Waldrop, and C. J. D'Orsi Detection of Simulated Lesions on Data-compressed Digital Mammograms Radiology, July 1, 2005; 236(1): 31 - 36. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. G. Elmore, K. Armstrong, C. D. Lehman, and S. W. Fletcher Screening for Breast Cancer JAMA, March 9, 2005; 293(10): 1245 - 1256. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. Gur Technology and Practice Assessment: In Search of a "Desirable" Statement Radiology, March 1, 2005; 234(3): 659 - 660. [Full Text] [PDF]
|
|
|
|
|
|
E. D. Pisano and M. J. Yaffe Digital Mammography Radiology, February 1, 2005; 234(2): 353 - 362. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Mahesh AAPM/RSNA Physics Tutorial for Residents: Digital Mammography: An Overview RadioGraphics, November 1, 2004; 24(6): 1747 - 1760. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Skaane and A. Skjennald Screen-Film Mammography versus Full-Field Digital Mammography with Soft-Copy Reading: Randomized Trial in a Population-based Screening Program--The Oslo II Study Radiology, July 1, 2004; 232(1): 197 - 204. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. T. Yang, G. J. Whitman, M. M. Johns |
