Abstract. The objective of this study is to compare image quality and lesion detection for full field digital mammography (FFDM) and filmâscreen mammography ...
The British Journal of Radiology, 78 (2005), 312–315 DOI: 10.1259/bjr/33317317
E
2005 The British Institute of Radiology
Comparison of full-field digital mammography and film–screen mammography: image quality and lesion detection 1
A FISCHMANN, MD, 1K C SIEGMANN, MD, 1A WERSEBE, MD, 1C D CLAUSSEN, MD and ¨ LLER-SCHIMPFLE, MD M MU
1,2 1 2
Department of Diagnostic Radiology, University Hospital Tu¨bingen, Hoppe-Seyler-Str. 3, 72076 Tu¨bingen and Sta¨dtische Kliniken Frankfurt/Main, Central Institute of Radiology, Germany
Abstract. The objective of this study is to compare image quality and lesion detection for full field digital mammography (FFDM) and film–screen mammography (FSM). In 200 women we performed digital mammography of one breast and film–screen mammography of the other breast. Imaging parameters were set automatically. Image quality, visualization of calcifications and masses were rated by three readers independently. Mean glandular dose was calculated for both systems. We found no significant difference in mean glandular dose. Image quality was rated by reader A/B/C as excellent for FFDM in 153/155/167 cases and for FSM in 139/116/114 cases (p,0.03/0.001/0.001). Microcalcifications were detected by FFDM in 103/89/98 and by FSM in 76/76/76 cases (p,0.01/0.06/0.01). Detection of masses did not differ significantly. FFDM provided significantly better visibility of skin and nipple-areola region (p,0.01). FFDM demonstrated improved image quality compared with film–screen mammography. Microcalcification detection was also significantly better with the digital mammography system for two of the three readers.
Digital technology is replacing conventional film–screen systems in all aspects of clinical radiology. Several digital mammography systems based on different physical concepts have been introduced and approved by the Food and Drug Administration in the last few years [1]. The first system introduced was the full field digital mammography (FFDM) system based on amorphous silicon (General Electric Medical Systems, Milwaukee, WI), followed by the charge coupled device (CCD)-based slot-scan-system (Fischer Imaging, Denver, CO) and the FFDM based on amorphous selenium (LoRad, Danbury, CT). Since November 2000 we have used a GE FFDM unit with a soft copy review workstation. The spatial resolution of the unit for high contrast objects is 5 line pairs mm21 (lp mm21). This is below the minimum requirements for film–screen mammography (FSM) in Europe, which require a spatial resolution of at least 10 lp mm21 (.10 lp mm21 acceptable, .13 lp mm21 desirable) [2]. The system has, however, a higher detective quantum efficiency (DQE) and contrast resolution [3, 4]. This allows for better contrast resolution of structures at a spatial resolution of 1–4 lp mm21, where the human eye performs best. Therefore the European guidelines for screening mammography exclude digital mammography from the higher requirements in spatial resolution [2]. A study using FFDM with screening patients showed a better detection of calcifications and better visualization of dense breast parenchyma compared with FSM [5]. However, few prospective clinical studies comparing image quality have been published [6–8]. The aim of this study was to compare image quality and lesion detection of FFDM and FSM in a prospective randomized study on the same woman without double radiation exposure of the breasts, a method already used in other studies [9]. Received 5 May 2004 and in revised form 20 October 2004, accepted 23 November 2004.
312
Materials and methods This study included 200 consecutive women without visible or palpable breast lesions, who were referred for mammography after individual counselling by a physician. Criteria for exclusion were visible breast asymmetry, previous breast surgery and/or known breast lesions and inflammation. Women below the age of 40 years were excluded to reduce radiation risk. The detector size of the Senographe 2000D is 19 cm623 cm, therefore women with breasts too large to be imaged with one exposure were excluded as well. All participants gave informed consent. The study was approved by the local ethics committee. In one patient breast cancer was detected during the course of the study. The mammograms of this patient were not available during this study owing to further external therapy, therefore the remaining 199 patients formed the study population. All participants received an FFDM of one breast and a FSM of the other, with random allocation of the modalities to the breasts. All images of one patient were taken by one radiographer compressing the breasts with identical force to identical thickness with both modalities. Using this method, comparison between an individual’s breasts could be performed without additional radiation dose. Image parameters were set by both systems automatically in standard mode. FSM was performed using a GE Senographe DMR+ with Kodak MinR 2000 film–screen system (Kodak, Rochester, NY), and developed using a Kodak Xomat M35 developer with RP Xomat chemicals at 33.5 ˚C. Imaging parameters were recorded for each exposure. Entrance skin dose was measured on a phantom for every set of patient parameters. These parameters were used to calculate mean glandular dose according to the method by Sobol and Wu [10]. FFDM was performed using the GE Senographe 2000D. Comparison was made on hardcopies, printed on a Kodak DryView 8610 laser-printer. Printout parameters were set in standard mode. The British Journal of Radiology, April 2005
Comparison of FFDM and FSM Table 1. Classification of parenchymal density Class
Size of a lesion that could be missed
1 2 3 4
,0.5 cm 0.5 to 1 cm 1 to 2 cm .2 cm
To measure breast density, we used a classification modified from the one used in the BI-RADS2-Lexicon by the American College of Radiology [18]. The tissue density is classified into four groups according to the size of lesions that cannot be excluded (Table 1). In this way we provide a measurement for exclusion of opacities especially for inhomogeneous breast parenchyma. Comparison was performed independently by three radiologists with 2, 4 and 10 years experience in mammography. FFDM and FSM were rated separately according to multiple criteria (Table 2), with both modalities rated in each session. To reduce reader bias, the evaluation of digital and conventional images of each patient were separated by at least 1 month. Images from 50 patients were rated twice to estimate intrareadervariability. Significance was tested with the sign-test and McNemar test. Statistical evaluation was performed on the statistic software JMP 4.0 for Windows NT (SAS Institute Inc. Cary, NC).
Results FFDM used significantly more Rh/Rh (rhodium/rhodium) as anode-filter-combination compared with FSM: in craniocaudal view (CC) in 102 cases with FFDM Table 2. Criteria of classification Criteria
Classification
Detection of the pectoral muscle Visibility of the nipple
Visible/not visible
Visibility of the skin
Breast density Sharpness Contrast in parenchymal tissue Contrast in fatty tissue Detection of microcalcifications Diagnostic quality for microcalcifications Detection of masses
Diagnostic quality for masses Overall image quality
Excellent/good/mediocre/not visible with and without bright light Excellent/good/mediocre/not visible with and without bright light Highly dense/dense/mostly fatty/fatty see Table 1 Sharp/slightly unsharp/unsharp Excellent/good/ mediocre/inadequate Excellent/good/mediocre/ inadequate No calcs/probably no calcs/probably calcs/calcs Excellent/good/ mediocre/inadequate No/probably no masses/ probably masses present/masses Excellent/good/ mediocre/inadequate Excellent to inadequate (school marks)
The British Journal of Radiology, April 2005
Table 3. Mean glandular dose calculations in mediolateral oblique (MLO) and craniocaudal (CC)-view
MLO
CC
Parameter
FSM
FFDM
Average dose [mGy] Standard deviation Minimum–maximum Average tube voltage [kV] Standard deviation Minimum–maximum Average tube-current-time product [mAs] Standard deviation Minimum–maximum Average dose [mGy] Standard deviation Minimum–maximum Average tube voltage [kV] Standard deviation Minimum–maximum Average tube current [mAs] Standard deviation Minimum–maximum
1.52 0.58 0.46–3.49 27.24 1.01 24–29 87.42
1.70 0.40 1.00–3.16 28.81 1.43 25–32 73.71
40.63 21–223 1.45 0.55 0.35–3.60 27.0 1.21 22–31 83.31 37.81 21–236
19.53 25–115 1.63 0.38 0.85–3.36 28.81 1.35 25–32 70.74 18.22 27–120
FSM, film–screen mammography; FFDM, full field digital mammography.
compared with 8 times with FSM, in mediolateral-oblique view (MLO) in 110 cases with FFDM compared with 7 cases with FSM. Mean glandular dose was slightly higher with FFDM with 1.63 mGy (0.85–3.36 mGy) in CC-view compared with 1.45 mGy (0.35–3.60 mGy) for FSM. In MLO-view the corresponding values were 1.70 mGy (1.00– 3.16 mGy) for FFDM and 1.52 mGy (0.46–3.49 mGy) with FSM (Table 3). These differences were not significant. FFDM significantly facilitated mammographic positioning. The pectoral muscle was depicted on the CC-view by Reader A/B/C with FFDM in 96/111/110 cases compared with 76/84/88 cases with FSM (p,0.02/0.001/0.01). This difference is demonstrated in Figure 1. The largest difference was in the depiction of the nipple (Figure 1) which was rated as excellent for FFDM even without bright light in 161/141/189 cases, for FSM in 6/7/ 103 cases (p,0.0001/0.0001/0.02). With the use of a bright light FFDM improved to 192/161/192 excellent cases, FSM to 52/55/177 cases (p,0.0001/0.0001/0.01). The depiction of the skin was rated excellent in 195/173/193 cases for FFDM and in 32/10/172 cases for FSM (p,0.0001/0.0001/0.01), which improved with the use of a bright light to 197/168/194 cases for FFMD and 158/104/ 191 cases with FSM (n.s. for reader C). Inter-reader concordance was low for skin and nipple detection with kappa-values ranging from 20.06 to 0.47. Image quality was rated as excellent for FFDM in 153/ 155/167 cases and for FSM in 139/116/114 cases (p,0.03/ 0.001/0.001). Contrast in parenchymal and fatty tissue was rated heterogeneously: while reader A and B found better contrast with FFDM in parenchymal tissue (p,0.05), reader C found a better contrast in fatty tissue (p,0.01). All three readers classified the breast parenchyma on digital mammograms to be less dense than the film–screen images (p,0.01/0.0001/0.01), with good inter-reader and intrareader agreement (kappa-values of 0.47–0.77). Another difference was in the detection of microcalcifications, which were detected with FFDM in 103/89/98 cases (kappa-values from 0.57 to 0.84) and with FSM in 313
A Fischmann, K C Siegmann, A Wersebe et al
Figure 1. Caudocranial-exposures, left film–screen mammography, right full-field digital mammography. Better depiction of m. pectoralis right and nipple.
76/76/76 cases (kappa-values from 0.51 to 0.62). This difference was significant for two readers and showed a strong tendency for the third one (p,0.02/0.07/0.02), while no differences were found in the diagnostic classification of these microcalcifications. No significant differences were found in the detection of masses although all readers were significantly more certain in the exclusion of masses (p,0.02/0.04/0.01).
Discussion This study compared both image quality and detection on FFDM and FSM using three experienced readers. The FFDM images were printed and compared under the same viewing conditions as the FSM. The readers were more experienced in FSM, however, all of the readers had read at least 100 digital mammograms prior to reading this study. Soft copy review may have allowed more manipulation of the images and possibly higher image quality scores for the digital images [11, 12]. Despite this and the readers being more used to FSM for diagnostic mammography, FFDM was at least equal to FSM in all examined parameters. Breast composition, size, compression force and thickness of each patient were identical with both modalities. In contrast to early studies on phantoms that showed possible dose reduction of up to 50% [13] we found a tendency for a higher mean glandular dose with FFDM that was not significant. While the dose used by FFDM was in the range reported in other studies, the mean glandular dose 314
used by the FSM was exceptionally low compared with data published by other workgroups [14]. In agreement with studies on phantoms [13, 15] and on patients [5, 6], we found a better detection of microcalcifications which was statistically significant for two out of three readers, considering the fact that two different breasts were examined. Even with FSM, which provides a spatial resolution of more than 15 lp mm21, the minimal size of detectable calcification is 130 mm [16]. In a phantom study FFDM proved to be superior to FSM especially in the detection of small calcifications in the 125–140 mm size range [17]. This indicates that the predominant limiting factor is the signal-to-noise ratio of the microcalcifications compared with the surrounding tissue rather than the theoretical maximal resolution for high contrast objects. For small calcifications, the contrast decreases rapidly with the size of the calcifications. In a clinical setting, the higher contrast resolution of FFDM seems to be more important than the detail resolution of FSM. Concerns that the lower spatial resolution could limit the characterization of calcifications have not been proven in clinical studies. Fischer et al [5] have shown that the FFDM system has a higher reliability in characterizing calcifications than FSM. Because of the ongoing discussion about soft-copy reading, we decided to perform this study with laser print-outs of the FFDM images, realising that this would limit the possibilities of digital mammography, especially concerning contrast resolution. Nevertheless we found no parameter which was rated to be superior in FSM. In contrast to phantom studies [13] we found no improvement in the detection of masses with FFDM. The reason for this seems to be that the detection appears to be limited by inhomogeneous parenchymal background and irregular structures rather than by small contrast differences. The advantages in contrast resolution in parenchymal and fatty tissue, however, made the diagnostic classification or exclusion of masses significantly easier with FFDM. This was also seen in the classification of parenchymal density. With FFDM the readers rated the breast parenchyma to be less dense and they were able to exclude smaller lesions because of the higher contrast resolution. A clear advantage of FFDM was the better depiction of skin and nipple without the use of a bright light. This can reduce reading time (Figure 2). The inter-reader variability for these parameters, especially for reader C, can be explained by a different threshold for excellent images. While A and B rated most of the images to be either good or mediocre, with few excellent images, reader C distributed the images between excellent and good, involuntarily reducing the original four-step-scale to three steps. Another advantage is the better depiction of the pectoral muscle in CC-view (Figure 2), which is an important factor for perfect positioning according to the image quality assurance criteria [2]. As FFDM does not require a film-cassette and cassette-holder, the detector reaches approximately 2 mm closer to the chest wall. The pectoral muscle, as well as parenchymal tissue or carcinomas situated close to the chest wall can be better visualized this way. The major limitation of this study is that owing to the differences between digital and conventional images it is not possible to blind the readers to the modalities. To reduce possible reader bias, we decided to read the The British Journal of Radiology, April 2005
Comparison of FFDM and FSM
References
Figure 2. Caudocranial-exposures, left full-field digital mammography (FFDM), right film–screen mammography (FSM). Clearly recognizably better depiction of skin and nipple-areola region FFDM. Breast parenchyma was rated to be less dense with FFDM.
different studies of each patient in separate sessions. Another limitation is the method to compare two different techniques by comparing differences between one individuals breasts. This could bias the results, as there might have been differences in the number of calcifications in the breasts. To reduce this bias, breasts were randomized to the modalities and possible differences were considered in the calculations. A further limitation is the subjective character of the parameters examined. As the radiologist is one of the major limiting factors in breast imaging, the non-objective parameters play an important role in the diagnostic process. For the acceptance of a new imaging modality, it is important that the radiologists ‘‘like’’ the pictures. Therefore we consider the measurement of subjective factors to be an important part in the evaluation of FFDM. Overall the full field digital mammography system (FFDM) was equal or superior to film–screen mammography (FSM) for all of the image quality and detection parameters studied. The FFDM was also judged superior in the following image quality comparisons: (1) overall image quality; (2) skin and nipple depiction; and (3) pectoralis muscle depiction. The FFDM was superior to FSM in detecting microcalcifications for two out of the three readers. No difference was found in the detection of masses. The full field digital mammography system provided high quality diagnostic mammograms that were equal or superior to film–screen mammograms.
Acknowledgments This study was supported by grant No. 2-0-0 by the AKF of the University of Tu¨bingen.
The British Journal of Radiology, April 2005
1. Bick U. Full-field digital mammography. Fortschr Ro¨ntgenstr 2000;12:957–64. 2. (EUREF) European Protocol for the Quality Control of the Physical and Technical Aspects of Mammography Screening. ‘‘European Guidelines for Quality Assurance in Mammography Screening’’ Appendix 5: Digital mammography, 1999. 3. Muller S. Full-field digital mammography designed as a complete system. Eur J Radiol 1999;1:25–34. 4. Vedantham S, Karellas A, Suryanarayanan S, Albagli D, Han S, Tkaczyk EJ, et al. Full breast digital mammography with an amorphous silicon-based flat panel detector: physical characteristics of a clinical prototype. Med Phys 2000;3: 558–67. 5. Fischer U, Baum F, Obenauer S, Luftner-Nagel S, Von Heyden D, Vosshenrich R, et al. Comparative study in patients with microcalcifications: full-field digital mammography vs screen-film mammography. Eur Radiol 2002;11: 2679–83. 6. Obenauer S, Luftner-Nagel S, von Heyden D, Munzel U, Baum F, Grabbe E. Screen film vs full-field digital mammography: image quality, detectability and characterization of lesions. Eur Radiol 2002;7:1697–702. 7. Lewin JM, Hendrick RE, D’Orsi CJ, Isaacs PK, Moss LJ, Karellas A, et al. Comparison of full-field digital mammography with screen-film mammography for cancer detection: results of 4,945 paired examinations. Radiology 2001;3: 873–80. 8. Lewin JM, D’Orsi CJ, Hendrick RE, Moss LJ, Isaacs PK, Karellas A, et al. Clinical comparison of full-field digital mammography and screen-film mammography for detection of breast cancer. AJR Am J Roentgenol 2002;3:671–7. 9. Thilander-Klang AC, Ackerholm PH, Berlin IC, Bjurstam NG, Mattsson SL, Mansson LG, et al. Influence of anodefilter combinations on image quality and radiation dose in 965 women undergoing mammography. Radiology 1997;2:348–54. 10. Sobol WT, Wu X. Parametrization of mammography normalized average glandular dose tables. Med Phys 1997;4:547–54. 11. Pisano ED, Cole EB, Major S, Zong S, Hemminger BM, Muller KE, et al. Radiologists’ preferences for digital mammographic display. The International Digital Mammography Development Group. Radiology 2000;3:820–30. 12. Funke M, Obenauer S, Hermann KP, Fischer U, Grabbe E. Soft copy versus hard copy findings in digital mammography. Radiologe 2002;4:265–9. 13. Obenauer S, Hermann KP, Schorn C, Fischer U, Grabbe E. Full-field digital mammography: dose-dependent detectability of breast lesions and microcalcinosis. Fortschr Ro¨ntgenstr 2000;12:1052–6. 14. Hermann KP, Obenauer S, Marten K, Kehbel S, Fischer U, Grabbe E. Average glandular dose with amorphous silicon full-field digital mammography - clinical results. Fortschr Ro¨ntgenstr 2002;6:696–9. 15. Obenauer S, Hermann KP, Schorn C, Funke M, Fischer U, Grabbe E. [Full-field digital mammography: a phantom study for detection of microcalcification]. Fortschr Ro¨ntgenstr 2000;7:646–50. 16. Karssemeijer N, Frieling JT, Hendriks JH. Spatial resolution in digital mammography. Invest Radiol 1993;5:413–9. 17. Rong XJ, Shaw CC, Johnston DA, Lemacks MR, Liu X, Whitman GJ, et al. Microcalcification detectability for four mammographic detectors: flat-panel, CCD, CR, and screen/ film). Med Phys 2002;9:2052–61. 18. Breast Imaging Reporting and Data System (BI-RADS) 4th Edition, American College of Radiology, Reston VA, 2004.
315