Evaluation of computer-aided detection of lesions in mammograms obtained with a digital phase-contrast mammography system

A computer-aided detection (CAD) system was evaluated for its ability to detect microcalcifications and masses on images obtained with a digital phase-contrast mammography (PCM) system, a system characterised by the sharp images provided by phase contrast and by the high resolution of 25-μm-pixel mammograms. Fifty abnormal and 50 normal mammograms were collected from about 3,500 mammograms and printed on film for reading on a light box. Seven qualified radiologists participated in an observer study based on receiver operating characteristic (ROC) analysis. The average of the areas under ROC curve (AUC) values for the ROC analysis with and without CAD were 0.927 and 0.897 respectively (P?=?0.015). The AUC values improved from 0.840 to 0.888 for microcalcifications (P?=?0.034) and from 0.947 to 0.962 for masses (P?=?0.025) respectively. The application of CAD to the PCM system is a promising approach for the detection of breast cancer in its early stages.     

Can computer-aided detection with double reading of screening mammograms help decrease the false-negative rate? Initial experience

PURPOSE: To retrospectively evaluate the role of computer-aided detection (CAD) in reducing the rate of false-negative (FN) findings on screening mammograms considered normal at initial double reading. MATERIALS AND METHODS: At the authors' institution, independent prospective double readings in which the second reader is not blinded to results of the first reading are performed routinely for all mammograms. When cancer is diagnosed, prior mammograms also are reviewed with double reading to determine cancer visibility. Findings are categorized as (a) no evidence of cancer on any prior screening mammogram and patient presents more than 1 year after prior screening, (b) no evidence of cancer on any prior screening mammogram and patient presents with symptoms within 1 year after prior screening (year-interval occult false-negative), or (c) cancer visible. The clinical director separately evaluates each case in the same way. In 2000, 519 histologically proved breast cancers were diagnosed, including 132 for which patients sought a second opinion and FN findings were not tracked. Prior screening mammograms were available in 318 of the other 387 cases. Five radiologists in two reading sessions independently reviewed current and prior mammograms to categorize visible cancers as either threshold or actionable FN findings. Visible cancers deemed actionable by at least three of five readers were analyzed with a commercially available CAD system. FN rates were calculated prior to and after CAD analysis. RESULTS: Twenty-seven occult and 71 visible cancers were found (total FN findings, 98). Three of five readers considered 52 (73%) of 71 visible cancers actionable. The CAD system correctly marked 37 (71%) of these 52 on prior screening mammograms (19 [65%] of 29 masses, seven [88%] of eight microcalcifications, seven [78%] of nine architectural distortions, and four [67%] of six masses with microcalcifications). The FN rate was 98 (31%) of 318 before CAD and 61 (19%) of 318 after CAD. CONCLUSION: In this retrospective review of this small subset of cancers, it appears that CAD has the potential to decrease the FN rate at double reading by more than one-third (from 31% to 19%). The CAD system correctly marked 37 (71%) of 52 actionable findings read as negative in previous screening years.     

Detection of masses and microcalcifications of breast cancer on digital mammograms: comparison among hard-copy film, 3-megapixel liquid crystal display (LCD) monitors and 5-megapixel LCD monitors: an observer performance study

The purpose of the study was to compare observer performance in the detection of masses and microcalcifications of breast cancer among hard-copy reading and soft-copy readings using 3-megapixel (3M) and 5-megapixel (5M) liquid crystal display (LCD) monitors. For the microcalcification detection test, we prepared 100 mammograms: 40 surgically verified cancer cases and 60 normal cases. For the mass detection test, we prepared 100 mammograms: 50 cancer cases and 50 normal cases. After six readers assessed both microcalcifications and masses set for each modality, receiver operating characteristic (ROC) analysis was performed. The average A_zs for mass detection using a hard copy and 3M and 5M LCD monitors were 0.923, 0.927 and 0.920, respectively; there were no significant differences. The average A_z for microcalcification detection using hard copy, 3M and 5M LCD monitors was 0.977, 0.954 and 0.972, respectively. There were no significant differences, but the P-values between the hard copy and 3M LCD monitor and that between the 3M and 5M LCD monitor were 0.08 and 0.09, respectively. In conclusion, the observer performances for detecting masses of breast cancers were comparable among the hard copy and two LCD monitors; however, soft-copy reading with a 3M LCD monitor showed slightly lower observer performance for detecting microcalcifications of breast cancers than hard-copy or 5M LCD monitor reading.     

Computerized analysis of tissue density effect on missed cancer detection in digital mammography.

This paper presents a study of the analysis of breast density in missed cancer cases and the effect of tissue density on cancer detection. A total of 100 missed cancer cases were collected. The breast density tissue was segmented with a statistical-based method. A set of tests was then applied to examine: (1) the differences in density between the mammograms at the detected stage and that at missed stage; (2) the density difference between the cancerous mammograms and their contra-lateral normal mammograms in the missed cancer cases; (3) the effect of breast density on CAD cancer detection. The results demonstrate that breast density is an important factor affecting not only radiologist's reading but also CAD performance. In order to improve early detection of breast cancer, a special effort should be directed to the high dense breast cases in CAD system design.     

Impact of breast density on computer-aided detection for breast cancer

OBJECTIVE: Our aim was to determine whether breast density affects the performance of a computer-aided detection (CAD) system for the detection of breast cancer. MATERIALS AND METHODS: Nine hundred six sequential mammographically detected breast cancers and 147 normal screening mammograms from 18 facilities were classified by mammographic density. BI-RADS 1 and 2 density cases were classified as nondense breasts; BI-RADS 3 and 4 density cases were classified as dense breasts. Cancers were classified as either masses or microcalcifications. All mammograms from the cancer and normal cases were evaluated by the CAD system. The sensitivity and false-positive rates from CAD in dense and nondense breasts were evaluated and compared. RESULTS: Overall, 809 (89%) of 906 cancer cases were detected by CAD; 455/505 (90%) cancers in nondense breasts and 354/401 (88%) cancers in dense breasts were detected. CAD sensitivity was not affected by breast density (p=0.38). Across both breast density categories, 280/296 (95%) microcalcification cases and 529/610 (87%) mass cases were detected. One hundred fourteen (93%) of the 122 microcalcifications in nondense breasts and 166 (95%) of 174 microcalcifications in dense breasts were detected, showing that CAD sensitivity to microcalcifications is not dependent on breast density (p=0.46). Three hundred forty-one (89%) of 383 masses in nondense breasts, and 188 (83%) of 227 masses in dense breasts were detected-that is, CAD sensitivity to masses is affected by breast density (p=0.03). There were more false-positive marks on dense versus nondense mammograms (p=0.04). CONCLUSION: Breast density does not impact overall CAD detection of breast cancer. There is no statistically significant difference in breast cancer detection in dense and nondense breasts. However, the detection of breast cancer manifesting as masses is impacted by breast density. The false-positive rate is lower in nondense versus dense breasts. CAD may be particularly advantageous in patients with dense breasts, in which mammography is most challenging.     

Understanding CAD (computer-aided diagnosis) in mammography

Generalization of breast screening programs requires efficient double reading of mammograms, which allows reduction of false negative interpretations, but it may be difficult to achieve. CAD (Computed Aided Detection) systems are dramatically improving and can now assist in the detection of suspicious mammographic lesions, either suspicious microcalcifications, masses or architectural distortion. Characterization of the lesions is improving as well. CAD mammography might complete or substitute to "human" double reading. The aim of this review is to present the main CAD systems commercially available, review the principles of CAD and discuss the results of CAD mammography. Specifically, the role of CAD within breast screening program, according to the results of recent prospective studies will be discussed.     

The performance of computer-aided detection when analyzing prior mammograms of newly detected breast cancers with special focus on the time interval from initial imaging to detection

Purpose

The clinical role of CAD systems to detect breast cancer, which have not been on cancer containing mammograms not detected by the radiologist was proven retrospectively.

Methods

All patients from 1992 to 2005 with a histologically verified malignant breast lesion and a mammogram at our department, were analyzed in retrospect focussing on the time of detection of the malignant lesion. All prior mammograms were analyzed by CAD (CADx, USA). The resulting CAD printout was matched with the cancer containing images yielding to the radiological diagnosis of breast cancer. CAD performance, sensitivity as well as the association of CAD and radiological features were analyzed.

Results

278 mammograms fulfilled the inclusion criteria. 111 cases showed a retrospectively visible lesion (71 masses, 23 single microcalcification clusters, 16 masses with microcalcifications, in one case two microcalcification clusters). 54/87 masses and 34/41 microcalcifications were detected by CAD.Detection rates varied from 9/20 (ACR 1) to 5/7 (ACR 4) (45% vs. 71%). The detection of microcalcifications was not influenced by breast tissue density.

Conclusion

CAD might be useful in an earlier detection of subtle breast cancer cases, which might remain otherwise undetected.     

Improvement in radiologists' detection of clustered microcalcifications on mammograms. The potential of computer-aided diagnosis

Relatively simple, but important, detection tasks in radiology are nearing accessibility to computer-aided diagnostic (CAD) methods. The authors have studied one such task, the detection of clustered microcalcifications on mammograms, to determine whether CAD can improve radiologists' performance under controlled but generally realistic circumstances. The results of their receiver operating characteristic (ROC) study show that CAD, as implemented by their computer code in its present state of development, does significantly improve radiologists' accuracy in detecting clustered microcalcifications under conditions that simulate the rapid interpretation of screening mammograms. The results suggest also that a reduction in the computer's false-positive rate will further improve radiologists' diagnostic accuracy, although the improvement falls short of statistical significance in this study.     

Observer variability in screen-film mammography versus full-field digital mammography with soft-copy reading

Full-field digital mammography (FFDM) with soft-copy reading is more complex than screen-film mammography (SFM) with hard-copy reading. The aim of this study was to compare inter- and intraobserver variability in SFM versus FFDM of paired mammograms from a breast cancer screening program. Six radiologists interpreted mammograms of 232 cases obtained with both techniques, including 46 cancers, 88 benign lesions, and 98 normals. Image interpretation included BI-RADS categories. A case consisted of standard two-view mammograms of one breast. Images were scored in two sessions separated by 5 weeks. Observer variability was substantial for SFM as well as for FFDM, but overall there was no significant difference between the observer variability at SFM and FFDM. Mean kappa values were lower, indicating less agreement, for microcalcifications compared with masses. The lower observer agreement for microcalcifications, and especially the low intraobserver concordance between the two imaging techniques for three readers, was noticeable. The level of observer agreement might be an indicator of radiologist performance and could confound studies designed to separate diagnostic differences between the two imaging techniques. The results of our study confirm the need for proper training for radiologists starting FFDM with soft-copy reading in breast cancer screening. Presented at ECR, Wien 2006.     

CAD in questions/answers Review of the literature

Generalization of breast screening programs requires an efficient double reading of the mammograms, which allows reduction of false-negative rate, but might be difficult to organize. CAD (Computed Assisted Diagnosis) is dramatically improving and is able to detect suspicious mammographic lesions, either suspicious microcalcifications, masses or architectural distorsions. CAD mammography might complete or substitute to "human" double reading. The aim of this review is to describe major CAD systems commercially available, working of CAD and to present principal results of CAD mammography. Specially, place of CAD within breast screening program, according to the results of recent prospective studies will be discussed.     

Digital mammography. ROC studies of the effects of pixel size and unsharp-mask filtering on the detection of subtle microcalcifications

We investigated the spatial resolution requirement and the effect of unsharp-mask filtering on the detectability of subtle microcalcifications in digital mammography. Digital images were obtained by digitizing conventional screen-film mammograms with a 0.1 X 0.1 mm2 pixel size, processed with unsharp masking, and then reconstituted on film with a Fuji image processing/simulation system (Fuji Photo Film Co., Tokyo, Japan). Twenty normal cases and 12 cases with subtle microcalcifications were included. Observer performance experiments were conducted to assess the detectability of subtle microcalcifications in the conventional, the unprocessed digital, and the unsharp-masked mammograms. The observer response data were evaluated using receiver operating characteristic (ROC) and LROC (ROC with localization) analyses. Our results indicate that digital mammograms obtained with 0.1 X 0.1 mm2 pixels provide lower detectability than the conventional screen-film mammograms. The detectability of microcalcifications in the digital mammograms is improved by unsharp-mask filtering; the processed mammograms still provide lower accuracy than the conventional mammograms, however, chiefly because of increased false-positive detection rates for the processed images at each subjective confidence level. Viewing unprocessed digital and unsharp-masked images in pairs resulted in approximately the same detectability as that obtained with the unsharp-masked images alone. However, this result may be influenced by the fact that the same limited viewing time was necessarily divided between the two images.     

Single reading with computer-aided detection and double reading of screening mammograms in the United Kingdom National Breast Screening Program

PURPOSE: To retrospectively determine if the use of a computer-aided detection (CAD) system can improve the performance of single reading of screening mammograms to match that of double reading in the United Kingdom. MATERIALS AND METHODS: Local research ethics committee approval was obtained; informed consent was not required. This study included a sample of 10 267 mammograms obtained in women aged 50 years or older who underwent routine screening at one of two breast screening centers in 1996. Mammograms that were double read in 1996 were randomly allocated to be re-read by eight different radiologists using CAD. The cancer detection and recall rates from double reading and single reading with CAD were compared. Statistical significance and confidence intervals were calculated with the McNemar test to account for the matched nature of the data. RESULTS: Single reading with CAD led to a cancer detection rate that was significantly (P = .02) higher than that achieved with double reading: 6.5% more cancers were detected by means of single reading with CAD than by means of double reading. However, the recall rate was higher for single reading with CAD than for double reading (8.6% vs 6.5%, respectively; P < .001). This was equivalent to relative increases of 15% and 32% in the cancer detection and recall rates, respectively. CONCLUSION: Single reading with CAD leads to an improved cancer detection rate and an increased recall rate.     

Computer-assisted detection as a second reader in symptomatic Asian women with palpable breast cancer

Purpose: To determine the potential role of a computer-assisted detection (CAD) algorithm as a second reader for experienced and inexperienced radiologists in mammography reading in Asian women.

Material and Methods: Two-view mammograms performed in 124 consecutive patients who presented with palpable breast cancer masses were retrospectively evaluated by two experienced breast radiologists (7 and 10 years' experience). The original reports of the session radiologists with variable experience of reading mammograms (2 to more than 10 years) were also evaluated. The number of suspicious masses and microcalcification clusters detected in each patient by both groups of radiologists were recorded. The radiologists then re-evaluated the films with the CAD system as a second reader. Any improvement in the detectability of breast pathology by either the experienced radiologists and/or by the session radiologists was then assessed. A total of 127 breasts had biopsy-proven carcinoma; 74 breasts had mastectomy performed. All the imaging results were correlated with tru-cut biopsy or mastectomy histology.

Results: With CAD-aided interpretation, there were altogether 95 visible masses and 77 suspicious microcalcification clusters in 109 mammographically detectable cancers correlated with histology results. There was a 7.4% (7/95) and 10.4% (8/77) increase in the number of masses and microcalcification clusters detected, respectively, by the experienced radiologists after application of CAD, whereas the increase was 13.7% (13/95) and 27.3% (21/77) for detection of masses and microcalcifications by the session radiologists, respectively. In 9 patients, a secondary focus detected by CAD was confirmed by histology. Three patients had contralateral breast tumors, 1 had a satellite invasive tumor while 5 had ductal carcinoma in situ on the same breast. Based on the biopsies and 74 mastectomies, the true-positive and false-positive detection rate of CAD was 92.6% and 31.8% for detection of carcinomas. The true-positive and false-positive detection rates were 100% and 58.8% for microcalcification clusters.

Conclusion: The current generation CAD algorithm helped to improve the detection rate of carcinomas, calcifications and multifocality in Asian breasts.     

计算机辅助检测对检出乳腺X线片中成簇微钙化灶的价值

目的:探讨计算机辅助检测系统(CAD)对检出乳腺X线钼靶摄影片中成簇微钙化灶的临床应用价值。方法:将22例乳腺X线钼靶片上疑有簇状钙化灶患者和13例正常对照者的140张乳腺钼靶X线片,经专业扫描仪数字化处理后,应用CAD软件标记其中的微钙化灶,由6位放射科医师分别单独阅片,再结合CAD阅片,结果采用受试者操作特性(ROC)曲线法进行分析。结果:6位放射科医师结合CAD阅片后,评价效果均优于未结合CAD时,其中3位低年资医师(有1年临床经验)和1位中年资医师(有5年以上临床经验)的两次评价结果有显著性差异(P<0.05)。结论:CAD有助于提高乳腺X线片中成簇微钙化灶的检出率,尤其对缺少诊断经验医师的作用更大。     

Detection of masses and clustered microcalcifications on data compressed mammograms: an observer performance study

OBJECTIVE: To evaluate observers' ability to detect breast masses and clustered microcalcifications depicted on data compressed mammograms, an observer performance study was performed. MATERIALS AND METHODS: Eight observers assessed 60 mammographic images obtained in six modes, ranging from noncompressed to a maximum data compression level of 101:1. Observers were asked to rate the images on a scale of 0 to 100 for the likelihood of the presence of a mass and also independently for the likelihood of the presence of clustered microcalcifications. In addition, observers were asked to rate their subjective assessment of the quality of each image for the detection of a mass and separately for the detection of microcalcifications. Receiver operating characteristic analyses were performed. RESULTS: The average area under the receiver operating characteristic curve, A(z), for the detection of clustered microcalcifications decreases significantly at the highest data compression level when compared with the noncompressed and two lowest levels of data compression (p < 0.01), and a trend test of the average area under the receiver operating characteristic curve for all observers is statistically significant (p < 0.05). No statistically significant differences among or between any of the data compression level modes for the detection of masses were detected. CONCLUSION: At a high level of mammogram data compression, observer performance was degraded for the detection of clustered microcalcifications. Detection of masses was not affected by the data compression methods and levels used in this study.     

Computer-aided diagnosis scheme for identifying histological classification of clustered microcalcifications by use of follow-up magnification mammograms

RATIONALE AND OBJECTIVES: Our purpose in this study was to investigate the usefulness of follow-up magnification mammograms (i.e., both current and previous magnification mammograms) in a computer-aided diagnosis (CAD) scheme for identifying the histological classification of clustered microcalcifications. MATERIALS AND METHODS: Our database consisted of current and previous magnification mammograms obtained from 93 patients before and after 3-month follow-up: 11 invasive carcinomas, 19 noninvasive carcinomas of the comedo type, 25 noninvasive carcinomas of the noncomedo type, 23 mastopathies, and 15 fibroadenomas. In our CAD scheme, we extracted five objective features of clustered microcalcifications from each of the current and previous magnification mammograms by taking into account image features that experienced radiologists commonly use to identify histological classifications. These features were then merged by a modified Bayes discriminant function for distinguishing among five histological classifications. For the input of the modified Bayes discriminant function, we used five objective features obtained from the previous magnification mammogram (previous features), five objective features obtained from the current magnification mammogram (current features), and the set of the five previous features and the five current features. RESULTS: The classification accuracies with the five current features were higher than those with the five previous features. These classification accuracies were improved substantially by using the set of the five previous features and the five current features. For the set of the five previous features and the five current features, the classification accuracies of our CAD scheme were 81.8% (9 of 11) for invasive carcinoma, 84.2% (16 of 19) for noninvasive carcinoma of the comedo type, 76.0% (19 of 25) for noninvasive carcinoma of the noncomedo type, 73.9% (17 of 23) for mastopathy, and 86.8% (13 of 15) for fibroadenoma. CONCLUSION: Our CAD scheme with use of follow-up magnification mammograms improved classification performance for mammographic clustered microcalcifications.     

Effect of soft-copy display supported by CAD on mammography screening performance

Diagnostic performance and reading speed for conventional mammography film reading is compared to reading digitized mammograms on a dedicated workstation. A series of mammograms judged negative at screening and corresponding priors were collected. Half were diagnosed as cancer at the next screening, or earlier for interval cancers. The others were normal. Original films were read by fifteen experienced screening radiologists. The readers annotated potential abnormalities and estimated their likelihood of malignancy. More than 1 year later, five radiologists reread a subset of 271 cases (88 cancer cases having visible signs in retrospect and 183 normals) on a mammography workstation after film digitization. Markers from a computer-aided detection (CAD) system for microcalcifications were available to the readers. Performance was evaluated by comparison of A_z-scores based on ROC and multiple-Reader multiple-case (MRMC) analysis, and localized receiver operating characteristic (LROC) analysis for the 271 cases. Reading speed was also determined. No significant difference in diagnostic performance was observed between conventional and soft-copy reading. Average A_z-scores were 0.83 and 0.84 respectively. Soft-copy reading was only slightly slower than conventional reading. Using a mammography workstation including CAD for detection of microcalcifications, soft-copy reading is possible without loss of quality or efficiency.     

Malignant lesions on mammography: accuracy of two different computer-aided detection systems

We retrospectively compared the accuracy of two computer-aided detection (CAD) systems for the detection of malignant breast lesions on full-field digital mammograms. Mammograms of 326 patients were analyzed (117 patients with breast cancer, 209 negative cases), and each set of cases was read by two CAD systems (Second Look versus AccuDetect Galileo). True-positive fractions per image and case for soft densities, microcalcifications, and total cancers were assessed. Study results showed better overall performance of AccuDetect Galileo (when compared to Second Look) in detecting masses, microcalcifications, and all cancer types, especially in extremely dense breast parenchyma.     

Screening mammography-detected cancers: sensitivity of a computer-aided detection system applied to full-field digital mammograms

PURPOSE: To retrospectively evaluate the sensitivity of the performance of a computer-aided detection (CAD) system applied to full-field digital mammograms for detection of breast cancers in a screening group, with histologic findings as the reference standard. MATERIALS AND METHODS: This study had institutional review board approval, and patient informed consent was waived. A commercially available CAD system was applied to the digital mammograms of 103 women (mean age, 51 years; range, 35-69 years) with 103 breast cancers detected with screening. Sensitivity values of the CAD system according to mammographic appearance, breast composition, and histologic findings were analyzed. Normal mammograms from 100 women (mean age, 54 years; age range, 35-75 years) with no mammographic and clinical abnormality during 2-year follow-up were used to determine false-positive CAD system marks. Differences between the cancer detection rates in fatty and dense breasts for the CAD system were compared by using the chi(2) test. RESULTS: The CAD system correctly marked 99 (96.1%) of 103 breast cancers. The CAD system marked all 44 breast cancers that manifested as microcalcifications only, all 23 breast cancers that manifested as a mass with microcalcifications, and 32 (89%) of 36 lesions that appeared as a mass only. The sensitivity of the CAD system in the fatty breast group was 95% (59 of 62) and in the dense breast group was 98% (40 of 41) (P = .537). The CAD system correctly marked all 31 lesions of ductal carcinoma in situ (DCIS), all 22 lesions of invasive ductal carcinoma with DCIS, the single invasive lobular carcinoma lesion, and 45 (92%) of 49 lesions of invasive ductal carcinoma. On normal mammograms, the mean number of false-positive marks per patient was 1.80 (range, 0-10 marks; median, 1 mark). CONCLUSION: The CAD system can correctly mark most (96.1%) asymptomatic breast cancers detected with digital mammographic screening, with acceptable false-positive marks (1.80 per patient).     

Tumour detection rate of a new commercially available computer-aided detection system

The aim of this study was to determine the tumour detection rate and false positive rate of a new mammographic computer-aided detection system (CAD) in order to assess its clinical usefulness. The craniocaudal and oblique images of 150 suspicious mammograms from 150 patients that were histologically proven to be malignant were analysed using the Second Look CAD (CADx Medical Systems, Quebec, Canada). Cases were selected randomly using the clinic's internal tumour case sampler. Correct marking of the malignant lesion in at least one view was scored as a true positive. Marks not at the location of the malignant lesion were scored as false positives. In addition, mammograms with histologically proven benign masses ( n=50) and microcalcifications ( n=50), as well as 100 non-suspicious mammograms, were scanned in order to determine the value of false-positive marks per image. The 150 mammograms included 94 lesions that were suspicious due to masses, 26 due to microcalcifications and 30 showed both signs of malignancy. The overall sensitivity was 90.0% (135 of 150). Sensitivity on subsets of the data was 88.7% (110 of 124) for suspicious masses (MA) and 98.2% (55 of 56) for microcalcifications. Eight of 14 false-negative cases were large lesions. The overall false-positive rate was observed as 0.28 and 0.97 marks per image of microcalcifications and masses, respectively. The lowest false-positive rates for microcalcifications and MA were observed in the cancer subgroup, whereas the highest false-positive rates were scored in the benign but mammographically suspicious subgroups, respectively. The new CAD system shows a high tumour detection rate, with approximately 1.3 false positive marks per image. These results suggest that this system might be clinically useful as a second reader of mammograms. The system performance was particularly useful for detecting microcalcifications.