Your SlideShare is downloading. ×
Image Quality, lesion detection, and diagnostic efficacy in ...
Image Quality, lesion detection, and diagnostic efficacy in ...
Image Quality, lesion detection, and diagnostic efficacy in ...
Image Quality, lesion detection, and diagnostic efficacy in ...
Image Quality, lesion detection, and diagnostic efficacy in ...
Image Quality, lesion detection, and diagnostic efficacy in ...
Image Quality, lesion detection, and diagnostic efficacy in ...
Image Quality, lesion detection, and diagnostic efficacy in ...
Image Quality, lesion detection, and diagnostic efficacy in ...
Image Quality, lesion detection, and diagnostic efficacy in ...
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

Image Quality, lesion detection, and diagnostic efficacy in ...

613

Published on

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
613
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
18
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. European Journal of Radiology 67 (2008) 487–496 Image Quality, lesion detection, and diagnostic efficacy in digital mammography: Full-field digital mammography versus computed radiography-based mammography using digital storage phosphor plates Gerd Schuellera,∗, Christopher C. Riedla, Reinhold Malleka, Klemens Eibenbergera, Herbert Langenbergera, Elisabeth Kaindla, Christiane Kulinna-Cosentinia, Margaretha Rudasb, Thomas H. Helbicha a Department of Radiology, Medical University of Vienna, Waehringer Guertel 18-20, A-1090 Vienna, Austria b Department of Pathology, Medical University of Vienna, Waehringer Guertel 18-20, A-1090 Vienna, Austria Received 3 July 2007; received in revised form 17 August 2007; accepted 20 August 2007 Abstract Objective: To compare image quality, the lesion detection, and the diagnostic efficacy of full-field digital mammography (FFDM) and computed radiography-based mammography using digital storage phosphor plates (DSPM) in the evaluation of breast lesions. Materials and methods: In this prospective study, 150 patients with suspicious breast lesions underwent FFDM and DSPM. Nine aspects of image quality (brightness, contrast, sharpness, noise, artifacts, and the detection of anatomic structures, i.e., skin, retromamillary space, glandular tissue, and calcifications) were evaluated by five radiologists. In addition, the detection of breast lesions and the diagnostic efficacy, based on the BI-RADS classification, were evaluated with histologic and follow-up correlation. Results: For contrast, sharpness, and the detection of all anatomic structures, FFDM was rated significantly better (p < 0.05). Mass lesions were equally detected, whereas FFDM detected more lesions consisting of calcifications (85 versus 75). DSPM yielded two false-negative results. Both lesions were rated BI-RADS 4 with FFDM, but BI-RADS 2 with DSPM. Both were invasive carcinoma at histology. The sensitivity, specificity, PPV, NPV, and accuracy of FFDM were 1.0, 0.397, 0.636, 1.0, and 0.707, compared to 0.974, 0.397, 0.630, 0.935, and 0.693 of DSPM. Conclusion: Based on image quality parameters, FFDM is, in part, significantly better than DSPM. Furthermore, the detection of breast lesions with calcifications is favorable with FFDM. However, the diagnostic efficacy of FFDM and DSPM was equal. The interpretation of the false-negative results suggests that the perception and characterization of breast lesions is not defined solely by the digital mammography system but is strongly influenced by the radiologist, who is one of the determinants in the interpretation of breast imaging. © 2007 Elsevier Ireland Ltd. All rights reserved. Keywords: Breast; Mammography; Digital mammography; Image quality; Comparative study 1. Introduction Digital technology is replacing film-screen systems in all areas of radiology, and mammography has been the last field to make this transition. Digital mammography offers the poten- tial for improved detection of breast lesions [1]. The results of a large prospective study, which was designed to compare film- ∗ Corresponding author. E-mail address: gerd.schueller@meduniwien.ac.at (G. Schueller). screen mammography with digital mammography, suggested that digital mammography may be better at detecting breast can- cer, particularly for women with heterogeneously or extremely dense breast tissue [2]. These data have stimulated the discus- sion of digital mammography technology among the women’s health and radiology communities, and may encourage addi- tional breast centers to purchase digital mammography systems [3]. Two digital mammography systems, based on different phys- ical concepts, have been introduced in the last few years [4,5]. These are full-field digital mammography (hereafter, FFDM) 0720-048X/$ – see front matter © 2007 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ejrad.2007.08.016
  • 2. 488 G. Schueller et al. / European Journal of Radiology 67 (2008) 487–496 and computed radiography-based mammography using digital storage phosphor plates (hereafter, DSPM). In FFDM, amor- phous silicium or selenium flat-panel detectors convert X-rays into electrical signals [6]. In DSPM, the image is recorded on a digital storage phosphor plate that is scanned using a laser reader [7–11]. Although available for more than 5 years, no more than about 10% of accredited mammography facilities use FFDM [12]. Reasons may include the disadvantage that new purchasers of FFDM must reconfigure all their current X-ray equipment [12]. Conversely,implementingDSPMsimplymeansthereplacement of film-screen cassettes by DSPM that can be used in a standard mammography unit [5]. While DSPM has been well-accepted in Europe and Japan, this system has recently been approved by the Food and Drug Administration (FDA), and is anticipated to experience rapid adoption [5]. Along with economic considerations, the image quality and the accuracy for the detection and radiological diagnosis of breast lesions influences the overall impression of a digital mammography system. To compare the image quality, lesion detection, and diagnostic efficacy of FFDM and DSPM, stud- ies can be designed within either a screening population or a patient population that allows a correlation of mammographi- cally suspicious lesions with histologic and follow-up outcome. However, the use of a screening population would result in a relatively low number of cancers, which decreases the power to detect a difference between the modalities and necessitates studying a large population [13,14]. Therefore, alternatively, we chose to design this prospective study to compare the image quality, lesion detection, and the diagnostic efficacy of FFDM and DSPM, based on the classi- fication of the American College of Radiology Breast Imaging Reporting and Data System (hereafter, BI-RADS [15]). Thus, 150 patients with suspicious breast lesions underwent both FFDMandDSPM,andmammographicdiagnosiswascorrelated with histologic and follow-up outcome. 2. Materials and methods 2.1. Patients, inclusion and exclusion criteria This prospective study was approved by the Institutional Review Board of our University. Written informed consent was obtained from all patients. From January to June 2003, 150 patients (mean age, 56 years; range, 23–81 years) with sus- picious lesions (BI-RADS 4) or lesions highly suggestive of malignancy (BI-RADS 5) on prior mammography, for which a second opinion was offered at our institution (a referral center), were included in the study protocol. All of the patients with lesions that were classified as BI-RADS 4 or 5 on second opin- ion were scheduled to undergo core breast biopsy (CBB [16]) or surgical excision biopsy of the breast lesions, according to the patients’ or referring surgeons’ preference. Patients with lesions classified as probably benign (BI-RADS 3) on second opinion were recommended for follow-up at a short interval (6 months). Patients with mammograms classified as negative (BI-RADS 1) or benign (BI-RADS 2) on second opinion were recommended for follow-up at 1 year after study entry [17]. Exclusion crite- ria were pregnancy, previous breast implant surgery, and age younger than 40 years, unless the lesions that required further evaluation were classified as BI-RADS 4 or 5. 2.2. Digital mammography Each of the 150 patients had mammography with both an FFDM and a DSPM system. Mammograms were performed of the breasts with the suspected lesions in the craniocaudal (CC) and mediolateral (ML) views. In total, 300 images were obtained with each mammography system. 2.3. FFDM A Senographe 2000D (GE Medical Systems, Milwaukee, WI., USA) system was used. This system uses an amorphous silicon flat-detector with cesium iodide as a scintillator. Mam- mograms were printed on a digital wet laser printer (Scopix LR 5200P, Agfa Leverkusen, Germany). Technical specifications are listed in Table 1. 2.4. DSPM A Mammomat 3000 Nova (Siemens, Erlangen, Germany) system was used. The digital storage phosphor plate used was type IP HR V (Fuji, Tokyo, Japan), read by the Image Reader FMDPL (Fuji). Mammograms were printed on a digital dry laser printer (DryPix FM-DP L, Fuji). Technical specifications are listed in Table 1. 2.5. Mammography quality control Technical quality controls were performed by a radiolo- gist, a quality control technologist, and a medical physicist, according to the tests recommended by the manufacturers of the digital mammography systems. Daily imaging of the same mammographic accreditation phantom, used for accreditation of screen-film systems, was performed [17,18]. 2.6. Data acquisition 2.6.1. Image Quality Image evaluation was performed on a viewing box with a luminance above 3000 cd/m2 (Rotolux 320 DS, Schulte, Warstein, Germany) in the center, with a difference between the center and corners of the viewing area of less than 15% [13,19]. Each of the images was interpreted independently by five radiologists with an experience in mammography of 3–6 years. The hard copy images were masked to the edges to pre- vent glare [20]. Reading conditions were kept constant, with the ambient room light subdued. To ensure adaptation of the eyes to the dark, each reading session began with five trial images that were similar to the test images, but were not part of the data analysis. To minimize reading order bias, each radiologist saw the images in a different random order. The radiologists
  • 3. G. Schueller et al. / European Journal of Radiology 67 (2008) 487–496 489 Table 1 Technical specifications of the digital mammography systems FFDM Senographe 2000D, GE DSPM Mammomat 3000Nova, Siemens + IP HR V, Fuji Panel active area (cm) 19.2 × 23 18 × 24 Pixel size (␮m) 100 50 DQE at 2 lp mm−1 0.6 (30 KV, 14 mAs) 0.3 (28 KV, 14 mAs) MTF at 1 lp mm−1 (%) 90 62 MTF at 2 lp mm−1 (%) 75 33 Spatial resolution (lp mm−1) 5 9 Wet laser printer Scopix LR 5200P, Agfa Dry laser printer DryPix FM-DP L, Fuji Spatial resolution (dots/in.) 630 300 Pixel size (␮m) 40 100 Gray-scale levels 4096 16,384 Resolution (lp mm−1) 12.5 10 Note: FFDM, full-field digital mammography; DSPM, digital storage phosphor mammography; DQE, detective quantum efficiency; MTF, modular transfer function; lp mm−1, line pairs mm−1; mAs, milliampere-second; KV, kilovolt. were free to vary the viewing distance. There were no time con- straints during the reading sessions [20]. The radiologists were blinded to the patients’ inclusion and exclusion criteria, to the image-generating digital mammography system, and to the his- tologic results of the breast lesions. In order to prevent any bias regarding the hard copies of each system, the captions on the films were masked with a black tape. The breast density was assigned according to the classifica- tion of the BI-RADS-Lexicon: category A, almost entirely fat; category B, scattered fibroglandular densities; category C, het- erogeneously dense; and category D, extremely dense [15]. The readers were asked to compare the images on the basis of the following nine aspects: brightness; contrast; sharpness; noise; artifacts (findings were classified as artifacts when there was no correlation of image anomalies to anatomic and/or pathologic structures, and/or when image anomalies were seen on a portion of the film without breast tissue imaged); and detailed detection of anatomic structures, i.e., skin, retromamillary space, glandu- lar tissue, and the detection of calcifications. Each of the nine aspects was categorized using a five-level scale (1, excellent; 2, good; 3, moderate; 4, poor; 5, not acceptable). 2.6.2. Lesion detection and diagnostic efficacy 2.6.2.1. Lesion detection. In case of breast lesions, lesion size (maximum diameter on mammograms) and characteristics, i.e., mass and/or calcifications, architectural distortion, and asym- metricdensity,wereevaluatedindependentlybytworadiologists with an experience in mammography of 6 and 10 years, respec- tively, in an unlabeled, random order. In cases of discrepancy, conclusions were reached by consensus in the presence of a third experienced radiologist. The reading conditions were identical to those described above. 2.6.2.2. Diagnostic efficacy. The diagnostic efficacy was assessed by the two radiologists who interpreted the lesion detection session, based on the classification of the BI- RADS-Lexicon: 1, negative; 2, benign; 3, probably benign; 4, suspicious;and5,highlysuggestiveofmalignancy[15].Accord- ing to a diagnostic, but, in contrast to a screening situation, the BI-RADS categories 1–3 corresponded to benign diagnoses and the BI-RADS categories 4 and 5 corresponded to malignant diagnoses [14]. 2.6.2.3. Diagnostic efficacy for malignant lesions. In particu- lar,theassessmentofBI-RADS4and5lesionsindifferentbreast density categories with FFDM and DSPM was compared. 2.6.3. Histologic and follow-up correlations Of 150 patients, 121 (80.7%) (mean age, 51 years; range, 23–78 years) with lesions classified as category BI-RADS 4 or 5 on either FFDM or DSPM underwent CBB (n = 98), or sur- gical excision biopsy (n = 23). Those with malignant histologic results on CBB underwent surgical excision, as did those with benign results, according to the patients’ or referring surgeons’ preference. The histologic results were compared to the image features in a multidisciplinary review. Of 150 patients, 29 (19.3%) (mean age, 58 years; range, 30–81 years) with lesions classified as category BI-RADS 1, 2 or 3 had follow-up for at least 2 years after study entry. 2.6.3.1. Statistical analysis. Statistical analysis was performed using commercially available software (SPSS for Windows, Realease12, SPSS Inc., Chicago, IL, USA). For the comparison of image quality, the significance of dif- ferences between FFDM and DSPM was determined using the Wilcoxon test [21]. In addition, the mean, the standard deviation (S.D.), and the level of significance of the rating data of the five radiologists was assessed for each of the nine aspect of image quality. For the diagnostic efficacy, the interrater agreement was evaluatedbymeansofthe␬statistic(Cohen’skappatest).Agree- ment was defined as almost perfect, ␬ > 0.8; good, ␬ = 0.8–0.61; moderate, ␬ = 0.60–0.41; fair, ␬ = 0.4–0.21; or poor, ␬ < 0.20 [21]. McNemar’s test was performed to evaluate whether the differences between FFDM and DSPM, with regard to the diagnostic efficacy, were significant [21]. For all tests, the sig- nificance level was set to p < 0.05 [21].
  • 4. 490 G. Schueller et al. / European Journal of Radiology 67 (2008) 487–496 Table 2 Ratings of image quality Image quality FFDM Senographe 2000D, GE DSPM Mammomat 3000Nova, Siemens + IP HR V, Fuji P-value Brightness 1.73/0.29 1.80/0.32 n.s. Contrast 1.54/0.39 2.09/0.35 <0.05 Sharpness 1.45/0.32 1.90/0.38 <0.05 Noise 1.40/0.27 1.49/0.41 n.s. Artifacts 1.32/0.21 1.38/0.30 n.s. Skin 1.50/0.49 2.31/0.38 <0.05 Retromamillary space 1.55/0.42 2.05/0.30 <0.05 Glandular tissue 1.64/0.42 1.99/0.42 <0.05 Calcifications 1.44/0.40 2.08/0.71 <0.05 Each aspect was categorized using a five-level scale ((1) excellent; (2) good; (3) moderate; (4) poor; (5) not acceptable). The mean/standard deviation and the level of significance are indicated. Note: FFDM, full-field digital mammography; DSPM, digital storage phosphor mammography; n.s., not significant. For FFDM and DSPM, sensitivity was calculated as TP/(TP + FN), specificity was calculated as TN/(TN + FP), PPV was calculated as TP/(TP + FP), and NPV was calcu- lated as TN/(TN + FN), and diagnostic accuracy was calculated as (TN + TP)/(TP + TN + FP + FN), with TP = true positive, FN = false negative, TN = true negative, and FP = false positive. 3. Results 3.1. Image quality Of 150 patients, 14 (9.3%) were rated with a breast density of category A, 64 (42.6%) of category B, 71 (47.4%) of category C, and one (0.7%) category D with FFDM, compared to 14 (9.3%), 51 (34%), 84 (56%), and one (0.7%) with DSPM. Of a total of 6750 image quality scores per digital mam- mography unit (150 patients, 9 scores, 5 readers), 4047 scores (59.96%) rated an excellent, 2019 scores (29.91%) a good, 505 scores (7.48%) a moderate, 163 scores (2.41%) a poor, and 16 scores (.24%) rated an unacceptable image quality for FFDM. ForDSPM,3122scores(46.25%)ratedanexcellent,1609scores (23.84%) a good, 1529 scores (22.65%) a moderate, 422 scores (6.25%) a poor, and 68 scores (1.01%) rated an unacceptable image quality (Fig. 1). For the mean, standard deviation, and level of significance, see Table 2. For each comparative aspect, except brightness, noise, and artifacts, FFDM was rated signifi- cantlybetterthanDSPM.DSPMwasnotratedsuperiortoFFDM images in any of the aspects. 3.2. Lesion detection and diagnostic efficacy 3.2.1. Lesion detection A total of 74 mass lesions (mean size, 16 mm; range 3–90 mm) were seen with FFDM (circumscribed, n = 2; indis- tinct, n = 48; spiculated, n = 24), compared to 80 mass lesions (mean size, 18 mm; range 3–90 mm) seen with DSPM (circum- scribed, n = 5; indistinct, n = 49; spiculated, n = 26); see Table 3, and Fig. 2. A total of 85 lesions consisting of calcifications (mean size, 15.1 mm; range 2–60 mm) were seen with FFDM (benign, n = 4; amorphous, n = 3; pleomorphous, n = 65; linear, n = 13), com- pared to 75 calcifications (mean size, 15.5 mm; range 3–50 mm) with DSPM (benign, n = 8; amorphous, n = 1; pleomorphous, n = 59; linear, n = 7); see Table 3, and Fig. 3. Within these 85 lesions, lesions that consisted of mass lesions as well as cal- cifications were seen with FFDM in 35 and with DSPM in 23 patients, respectively. Architectural distortion or asymmetric density was not observed. 3.2.2. Diagnostic efficacy With FFDM, BI-RADS 1 was rated in seven cases, BI-RADS 2 in one case, BI-RADS 3 in 21 cases, BI-RADS 4 in 81 cases, and BI-RADS 5 in 40 cases, compared to five, nine, 17, 76, and 43 cases, respectively, with DSPM; i.e., malignancy was sus- pected in 121 cases with FFDM and in 119 cases with DSPM; see Table 3, and Figs. 2 and 3. Interrater agreement was almost perfect (0.80) for FFDM, and good (0.78) for DSPM. The dif- ference in diagnostic efficacy between FFDM and DSPM was not significant (p = 0.86). Table 3 Lesion detection and diagnostic efficacy, based on the classification of the American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) FFDM Senographe 2000D, GE DSPM Mammomat 3000Nova, Siemens + IP HR V, Fuji Mass circumscribed 2 5 Indistinct 48 49 Spiculated 24 26 Total 74 80 Calcifications benign 4 8 Amorphous 3 1 Pleomorphous 65 59 Linear 13 7 Total 85 75 BI-RADS 1 7 5 BI-RADS 2 1 9 BI-RADS 3 21 17 BI-RADS 4 81 76 BI-RADS 5 40 43 Number of patients, n = 150. Note: FFDM, full-field digital mammography; DSPM, digital storage phosphor mammography.
  • 5. G. Schueller et al. / European Journal of Radiology 67 (2008) 487–496 491 Fig. 1. Comparison of image quality. Right craniocaudal (CC) and mediolateral (ML) mammogram views of a representative patient (49-year-old woman with breast density category C according to the classification of the BI-RADS-Lexicon) are shown; (A) FFDM CC; (B) FFDM ML; (C) DSPM CC; (D) DSPM ML. The image quality aspects contrast, sharpness, and detailed detection of anatomic structures (skin, retromamillary space, and glandular tissue) were rated slightly better for FFDM than for DSPM. The aspects brightness, noise, and artifacts were rated equal. 3.2.3. Diagnostic efficacy for malignant lesions With FFDM, three BI-RADS 4 and 5 mass lesions (mean size, 14.8 mm; range, 3–20 mm) were seen in breast density cat- egory A, 19 (mean size, 21.8 mm; range, 8–90 mm) in breast density category B, 21 (mean size, 18 mm; range, 10–48 mm) in breast density category C, and zero in breast density category D, compared to seven (mean size, 15.8 mm; range, 3–30 mm), 24 (mean size, 20.2 mm; range, 6–90 mm), 24 (mean size, 20.2 mm; range, 8–50 mm), and zero with DSPM; see Table 4. With FFDM, five (mean size, 14.9 mm; range, 4–35 mm) BI- RADS 4 and 5 lesions consisting of calcifications were seen in breast density category A, 17 (mean size, 16.8 mm; range, 3–60 mm) in breast density category B, 27 (mean size, 15.7 mm; range, 2–60 mm) in breast density category C, and zero in breast density category D, compared to three (mean size, 14.4 mm; range, 3–30 mm), 12 (mean size, 16 mm; range, 3–50 mm), 29 (mean size, 17 mm; range, 3–50 mm), and zero with DSPM; see Table 4. BI-RADS 4 and 5 lesions that consisted of mass lesions as well as calcifications were found in breast density category A, B, C, and D in one (mean size, 12.5 mm; range, 5–25 mm), 16 (mean size, 20.5 mm; range, 7–42 mm), 11 (mean size, 19.5 mm; range, 10–40 mm), and one case (size, 20 mm) with FFDM, as compared to one (mean size, 12.5 mm; range, 5–25 mm),
  • 6. 492 G. Schueller et al. / European Journal of Radiology 67 (2008) 487–496 Fig. 2. Lesion detection and diagnostic efficacy: mass lesion. Left mediolateral mammogram views of a representative patient (65-year-old woman with breast density category A according to the classification of the BI-RADS-Lexicon) are shown; (A) FFDM; (B) DSPM. An irregular, spiculated, high-density mass (size, 21 mm) was detected in the upper portion of the breast. The lesion was rated BI-RADS 5 with both FFDM and DSPM. The lesion was found to be invasive cancer at histology. Note: FFDM, full-field digital mammography; DSPM, digital storage phosphor mammography. nine (mean size, 22.0 mm; range, 9–45 mm), nine (mean size, 19.0 mm; range, 10–38 mm), and one (size, 20 mm) case with DSPM; see Table 4. 3.3. Histologic and follow-up correlations The histologic results of CBB and surgical excision biopsy were malignancy in 77 cases (invasive carcinoma, n = 51; ductal carcinoma in situ, n = 26), and benign lesions in 44 cases. Each malignant result of CBB histology was confirmed by surgery. Follow-up in 29 cases yielded no suspicious breast lesions. Table 4 Assessment of the diagnostic efficacy of BI-RADS 4 and 5 lesions, based on the classification of the BI-RADS-Lexicon in breast density categories A through D Breast density category Mass Calcifications Mass +calcifications Total A 3/7 5/3 1/1 9/11 B 19/24 17/12 16/9 52/45 C 21/24 27/29 11/9 59/62 D 0/0 0/0 1/1 1/1 Total 43/55 49/44 29/20 121/119 The numbers of breast lesions diagnosed with FFDM/DSPM are shown. Note: FFDM, full-field digital mammography; DSPM, digital storage phosphor mam- mography. Accordingly, FFDM encountered 77 TP, 29 TN, 44 FP, and 0 FN results; DSPM encountered 75 TP, 29 TN, 44 FP, and 2 FN results. Both false-negative results with DSPM were lesions that consisted of pleomorphous calcifications (size, 3 mm each) in patients with a breast density category of B, which were rated BI-RADS 4 with FFDM. The same calcifications were rated BI- RADS 2 with DSPM. Both lesions were found to be invasive cancer at histology (Fig. 4). The results of sensitivity, specificity, PPV, NVP, and accuracy of FFDM and DSPM are listed in Table 5. Table 5 Number of patients, n = 150 FFDM Senograph 2000D, GE DSPM Mammomat 3000Nova, Siemens + IP HR V, Fuji Sensitivity 1.0 (0.953/–) 0.974 (0.909/0.997) Specificity 0.397 (0.285/0.519) 0.397 (0.285/0.519) PPV 0.636 (0.544/0.722) 0.630 (0.537/0.717) NPV 1.0 (0.881/–) 0.935 (0.786/0.992) Accuracy 0.707 (0.627/0.778) 0.693 (0.613/0.766) Note: FFDM, full-field digital mammography; DSPM, digital storage phosphor mammography; PPV, positive predictive value; NPV, negative predictive value; lower/upper values of confidence intervals in brackets.
  • 7. G. Schueller et al. / European Journal of Radiology 67 (2008) 487–496 493 Fig. 3. Lesion detection and diagnostic efficacy: calcifications. Left mediolateral mammogram views of a representative patient (69-year-old woman with breast density category A according to the classification of the BI-RADS-Lexicon) are shown; (A) FFDM; (B) DSPM. A lesion that consisted of clustered pleomorphous calcifications (size, 20 mm) was detected in the retromamillary space (white circles). The lesion was rated BI-RADS 4 with both FFDM and DSPM. The lesion was found to be invasive cancer at histology. In addition, linear calcifications that are associated with tubular structures were slightly better detectable with FFDM. They were rated as benign vascular calcifications with both systems. Note: FFDM, full-field digital mammography; DSPM, digital storage phosphor mammography. 4. Discussion Radiologists usually compare the image quality of different mammography systems with non-objective aspects. The most important are brightness, contrast, sharpness, and noise [22]. Our data show that FFDM provides a superior image quality to DSPM. For six of nine scores of image quality, including contrast and sharpness, significantly better results for FFDM were noted, while for brightness and noise, both systems were equal (Table 2). In addition, FFDM is superior to DSPM in the Fig. 4. False-negative result with DSPM. Left mediolateral mammogram views of a representative patient (55-year-old woman with breast density category B according to the classification of the BI-RADS-Lexicon) are shown; (A) FFDM; (B) DSPM. A lesion that consisted of pleomorphous calcifications (size, 3 mm) was detected in the central portion of the breast (white circles). It was rated BI-RADS 4 with FFDM. The same lesion was rated BI-RADS 2 with DSPM. The lesion was found to be invasive cancer at histology. Note: FFDM, full-field digital mammography; DSPM, digital storage phosphor mammography.
  • 8. 494 G. Schueller et al. / European Journal of Radiology 67 (2008) 487–496 detailed detection of anatomic structures of the breast (Table 2), as well as in the detection of breast lesions consisting of calci- fications (Table 3). However, our data demonstrated that FFDM and DSPM were equal in diagnostic efficacy, with compara- ble values of sensitivity, specificity, PPV, NPV, and accuracy (Table 5). Notably, in patients with breast density categories of C and D, no differences in diagnostic efficacy were observed (Table 4). Our data may be of interest, since this is the first study that compares the image quality, lesion detection, and diagnostic efficacy of FFDM and DSPM in a clinical setting. The study designallowedpatientswithsuspiciousbreastlesionstoundergo both FFDM and DSPM, as well as enabling the correlation of mammographic lesion aspects with histologic and follow-up outcome. In our study, markedly more breasts were classified with a breast density category of B with FFDM, and, conversely, as a breast density category of C with DSPM. We suggest that, most likely, this is due to the observed superior detailed detection of fibroglandular tissue in dense breasts with FFDM, leading to a downward categorization of breast density. By revealing that digital mammography has significant ben- efits over film-screen mammography, especially in breasts with density categories of C and D, where the sensitivity of the tradi- tional technology is somewhat limited [23], the study by Pisano et al. has already begun to increase the demand for digital mam- mography among clinicians and patients [2,5]. Notably, in our study, DSPM was proven to be similarly effective as FFDM in the detection of BI-RADS 4 and 5 mass lesions in patients with dense breasts. In our opinion, the fact that, overall, six mass lesions more were suspected with DSPM than with FFDM (Table 3), can, at least in part, be explained by the superior detailed detection of fibroglandular tissue and, hence, a better differentiation between fibroglandular tissue and mass lesions with FFDM. Differences, however, were noted in the detection of cal- cifications. In contrast to previous data that showed an equal detection of calcifications for different digital mammography systems [24,25], in our study, with FFDM, more calcifications were detected than with DSPM (85 versus 75; Table 3). While the number of detected BI-RADS 4 and 5 lesions that consisted of calcifications was comparable in patients with a breast den- sity of C or D with FFDM and DSPM (27 versus 29; Table 4), notably,agreaternumberwasdetectedbyFFDMinpatientswith a breast density category of A or B (22 versus 15; Table 4). This can, in part, be explained by the fact that, with DSPM, a smaller number of breasts were classified with a density category of A or B than with FFDM (65 versus 78). As a consequence, some of the detected lesions that consisted of calcifications in breasts of density category A or B with FFDM were, with DSPM, in fact, detected in breasts of a density category of C or D. Furthermore, DSPM yielded two false-negative results. Both lesions consisted of calcifications in patients with a breast den- sity category of B. Both lesions were displayed by both FFDM and DSPM. Notably, they were rated BI-RADS 4 with FFDM, whereas, with DSPM, they were rated BI-RADS 2 (Fig. 4). Both lesions proved to be invasive carcinoma. In our opinion, these data show that the perception and characterization of breast lesions is not solely dependent on the digital mammography system, but is strongly influenced by the interpretation of the radiologist, rendering the radiologist one of the major determi- nants in breast imaging. This dilemma has also been reported by others [26–28]. However, basically, the superior depiction of fibroglandular tissue details with FFDM, and the differences in the detection of calcifications that favored FFDM in our study, are most likely to be considered an effect of differences in the detective quantum efficiency(DQE)andthemodulartransferfunction(MTF).DQE and MTF are the most well-known methods for the objective assessment of differences of mammography systems [29]. DQE is the measure of the combined effect of the image noise and the contrast of an imaging system, expressed as a function of object detail. The FFDM system has a higher DQE than DSPM (Table 1). This allows for better contrast resolution of structures at a spatial resolution of 1–4 line pairs mm−1 (lp mm−1), where the human eye performs best [30]. MTF is the measure of the ability of an imaging system to render the contrast of an object as a function of object detail. The FFDM system has a much higher MTFthantheDSPMsystem(Table1).This,mostlikely,isdueto the cesium iodide scintillator, which produces higher-resolution images than the DSPM [11]. The assumption that digital mammography would require a spatial resolution similar to that of film-screen mammography (at least 10 lp mm−1 required in Europe [31]) has recently been proven to be a misconception [32]. Equally, in the comparison of the digital mammography systems in our study, the lower spatial resolution of 5 lp mm−1 for FFDM versus 9 lp mm−1 for DSPM, is not arbitrative (Table 1). Considering the superior results of the image quality for FFDM in our study, the better DQE and MTF seem to compensate for the lower spatial resolution of FFDM compared to DSPM. The interrater agreement for diagnostic efficacy, based on the BI-RADS classification, was almost perfect for FFDM (␬ = 0.80), and good (␬ = 0.78) for DSPM. Recently, Berg et al. [33] reported interrater agreement for the BI-RADS classi- fication to be as low as 0.41–0.44, even subsequent to a short BI-RADS training session. The authors evaluated the diagnostic efficacy of 23 practicing mammogram-interpreting physicians. In contrast, in our study, the mammograms were examined by two experienced radiologists using the BI-RADS system over several years. This may explain the differences in the interrater agreement between our study and the results reported in the literature. Our study suffers from several limitations. First, the per- formance of the lesion detection and diagnostic efficacy may have been biased: although the study was designed to blind the readers to the patients’ inclusion and exclusion criteria and to the histologic results of the breast lesions, the readers might have been more aware of breast lesions, since only the unilat- eral breasts with suspicious lesions were imaged. Therefore, it may be debated whether the number of suspected lesions, as well as the number of false positive lesions (n = 44/150), might have been smaller if the readers would have encountered the mammograms of both breasts. However, additional imaging by
  • 9. G. Schueller et al. / European Journal of Radiology 67 (2008) 487–496 495 both digital systems of the contralateral breast without suspected lesions would have been against ethical principles. Second, the determination of subjective aspects of image quality, as well as lesion detection, was based on visual assess- ment, which is influenced by individual observer characteristics and subjective criteria. We tried to decrease the effect of indi- vidual observer preferences by including five radiologists for the evaluation of the image quality scores. However, we chose to summarize the subjective rating scores for each aspect of the image quality across all five radiologists to minimize the effect of reader variability. ␬ statistics were not applicable because of the asymmetric distribution of the rating scores for the aspects of image quality [20]. Third, although one of the major advantages of digital over film-screen mammography is the possibility of soft-copy read- ing, in this study, both FFDM and DSPM images were printed and compared under the same viewing conditions. With post- processing, soft-copy review may have altered the results of the image quality scores for both systems, as described recently [34,35]. However, at the time of image interpretation, no digi- tal monitor was commercially available that could provide the recommended spatial resolution for soft-copy reading for one of the systems. However, we do not believe that the use of soft- copy reading would have altered the almost identical results of the diagnostic efficacy for both systems, and of course the conclusion drawn for our data. Fourth, FFDM images were printed on a digital wet laser printer, and DSPM images on a digital dry laser printer. How- ever, we do not feel that the difference in the printer devices has influenced our results. Notably, no significant differences were observed in the rating of image artifacts. This is in accord with previous data that observed equal image quality and diagnostic efficacy of wet and dry laser printers in digital mammography [36]. 5. Conclusion In conclusion, the data from this prospective study, which compared the image quality, lesion detection, and diagnostic efficacy of FFDM and DSPM by histologic and follow-up corre- lation of breast lesions, provide better results for FFDM in image quality, and in the detection of lesions that consisted of calci- fications. However, diagnostic efficacy was equal for the two systems. Particularly in patients with breast density categories of C and D, where digital mammography is a ray of hope in the detection of breast lesions, no differences in diagnostic efficacy were observed. Based on the interpretation of the false-negative results, our data suggest that the perception and characterization of breast lesions is not defined solely by the digital mammo- graphic system: it is essentially influenced by the radiologist, who is one of the major determinants in the interpretation of breast imaging. Acknowledgment We are grateful to Mary McAllister, Johns Hopkins Univer- sity Hospital, Baltimore, MD, USA, for manuscript assistance. References [1] Pisano ED, Yaffe MJ, Hemminger BM, et al. Current status of full-field digital mammography. Acad Radiol 2000;7:266–80. [2] Pisano ED, Gatsonis C, Hendrick E, et al. Diagnostic performance of dig- ital versus film mammography for breast-cancer screening. N Engl J Med 2005;353:1773–83. [3] Harris CH. Digital mammography hits the tipping point [Health Imaging & IT Web site]. April 1, 2006. Available at: http://www.healthimaging.com /content/view/4089/68/. Accessed on April 1, 2006. [4] Bick U. Full-field digital mammography. Fortschr R¨ontgenstr 2000;12:957–64. [5] Fischer C. CR digital mammography: an affordable entry [Amer- ican Healthcare Radiology Administartors Web site]. September 1, 2006. Available at: http://www.ahraonline.org/ConfEd/Education /2006SeptemberOctober1/2006SeptemberOctober1.pdf. Accessed on September 1, 2006. [6] Schulz-Wendtl R, Wenkel E, Schmid A, et al. Experimental investigations of image quality in X-ray mammography with a conventional screen film System (SFS) and a new full-field digital mammography unit (DR) with a-se detector. Fortschr R¨ongenstr 2003;175:766–8. [7] Noel A, Thibault F. Digital detectors for mammography: the technical challenges. Eur Radiol 2004;14:1990–8. [8] Mahesh M. AAPM/RSNA physics tutorial for residents: digital mammog- raphy: an overview. Radiographics 2004;24:1747–60. [9] Pisano ED, Yaffe MJ. Digital mammography. Radiology 2005;234:353– 62. [10] Kheddache S, Thilander-Klang A, Lanhede B, et al. Storage phosphor and film-screen mammography: performance with different mammographic techniques. Eur Radiol 1999;9:591–7. [11] Fischer U, Baum F, Obenauer S, et al. Comparative study in patients with microcalcifications: full-field digital mammography vs. screen-film mammography. Eur Radiol 2002;12:2679–83. [12] Harvey D. CR mammography. Radiol Today 2006;6:12. [13] Bassett LW, Hendrick RE, Bassford TL, et al. Quality determinants of mammography: clinical practice guideline no. 13. AHCPR publication 95- 0632 Rockville, MD: Agency for Health Care Policy and Research, Public Health Service, U.S. Dept of Health and Human Services, 1994. [14] Lewin JM, Hendrick RE, D´orsi CJ, et al. Comparison of full-field digital mammography with screen-film mammography for cancer detec- tion: results of 4945 paired examinations. Radiology 2001;218:873– 80. [15] Breast Imaging Reporting and Data System (BI-RADS) 4th ed., American College of Radiology, Reston VA, 2004. [16] Wallis M, Tarvidon A, Helbich T, et al. Guidelines from the European Society of Breast Imaging for diagnostic interventional breast procedures. Eur Radiol 2007;17:581–8. [17] Pisano ED, Gatsonis CA, Yaffe MJ, et al. American College of Radiol- ogy Imaging Network Digital Mammographic Imaging Screening Trial: objectives and methodology. Radiology 2005;236:404–12. [18] Hendrick RE, Bassett LW, Dodd GD, et al. Mammography quality control: radiologist’s manual, radiologic technologist’s manual, medical physicist’s manual. Reston, VA: American College of Radiology; 1999. [19] American College of Radiology (ACR) Committee on Quality Assurance in Mammography. Mammography Quality Control. Reston, VA, 1999: Amer- ican College of Radiology 1999; 68–70,142,185–196,208,209,287–294. [20] Uffmann M, Prokop M, Kupper W, et al. Soft-copy reading of digital chest radiographs: effect of ambient light and automatic optimization of monitor luminance. Invest Radiol 2005;40:180–5. [21] Davies M, Fleiss JL. Measuring agreement for multinominal data. Biomet- rics 1982;38:1047–51. [22] Schulz-Wendtland R, Hermann KP, Lell M, et al. Phantom study for the detection of simulated lesions in five different digital and one conventional mammography system. Fortschr R¨ontgenstr 2004;176:1127–32. [23] Kolb TM, Lichy J, Newhouse JH. Comparison of the performance of screening mammography, physical examination, and breast US and evalua- tion of factors that influence them: an analysis of 27,825 patient evaluations. Radiology 2002;225:165–75.
  • 10. 496 G. Schueller et al. / European Journal of Radiology 67 (2008) 487–496 [24] Feig SA, Yaffe MJ. Digital mammography. Radiographics 1998; 18:893–901. [25] Cole EB, Pisano ED, Kistner EO, et al. Diagnostic accuracy of digital mam- mography in patients with dense breasts who underwent problem-solving mammography: effects of image processing and lesion type. Radiology 2003;226:153–60. [26] Kopans DB. Double reading. Radiol Clin North Am 2000;38:719– 24. [27] Brown J, Bryan S, Warren R. Mammography screening: an incremental cost effectiveness analysis of double versus single reading of mammograms. BMJ 1996;312:809–12. [28] Nawano S, Murakami K, Moriyama N, et al. Computer-aided diagnosis in full digital mammography. Invest Radiol 1999;34:310–6. [29] Fischer U, Hermann KP, Baum F. Digital mammography: current state and future aspects. Eur Radiol 2006;16:38–44. [30] Fischmann A, Siegmann KC, Wersebe A, et al. Comparison of full-field digital mammography and film–screen mammography: image quality and lesion detection. Br J Radiol 2005;78:312–5. [31] (EUREF) European Protocol for the Quality Control of the Physical and Technical Aspects of Mammography Screening. European Guidelines for Quality Assurance in Mammography Screening, 1999. [32] Obenauer S, Luftner-Nagel S, von Heyden D, et al. Screen film vs. full-field digital mammography: image quality, detectability and characterization of lesions. Eur Radiol 2002;12:1697–702. [33] Berg WA, D’Orsi CJ, Jackson VP, et al. Does training in the Breast Imaging Reporting and Data System (BI-RADS) improve biopsy recommenda- tions or feature analysis agreement with experienced breast imagers at mammography? Radiology 2002;224:871–80. [34] Pisano ED, Cole EB, Major S, et al., The International Digital Mam- mography Development Group. Radiologists’ preferences for digital mammographic display. Radiology 2000;3:820–30. [35] Funke M, Obenauer S, Hermann KP, et al. Soft copy versus hard copy findings in digital mammography. Radiologe 2002;4:265–9. [36] Schueller G, Kaindl E, Langenberger H, et al. Validation of image quality in full-field digital mammography: is the replacement of wet by dry laser printers justified? Eur J Radiol 2007;62:267–72.

×