Reading and decision aids for improved accuracy and standardization of mammographic diagnosis

Structured reporting and quality control in CT colonography

Standardized recommended techniques for performing and reporting CT colonography (CTC) examinations were developed by a consensus of experts. Published reporting guidelines, known as the CT colonography reporting and data system supplemented by recently updated comprehensive recommendations were incorporated into the American College of Radiology (ACR) practice guidelines. The application of continuous quality improvement to the practice of CT was aided by the development of an ACR national data registry (NRDR) for CTC that addressed both process and outcome quality measures. These measures can be used to benchmark an institution's CTC practice as compared to all participants. This article will discuss the best practices for reporting CTC and describe the use of NRDR to foster quality CTC performance.

CT Colonography Reporting and Data System (C-RADS): benchmark values from a clinical screening program

OBJECTIVE

The CT Colonography Reporting and Data System (C-RADS) is a well-recognized standard for reporting findings at CT colonography (CTC). However, few data on benchmark values for clinical performance have been published to date, especially for screening. The purpose of this study was to establish baseline C-RADS values for CTC screening.

SUBJECTS AND METHODS

From 2005 to 2011, 6769 asymptomatic adults (3110 men and 3659 women) 50-79 years old (mean [± SD] age, 56.7 ± 6.1 years) were enrolled for first-time CTC screening at a single center. CTC results were prospectively classified according to C-RADS for both colorectal and extracolonic findings. C-RADS classification rates and outcomes for positive patients were analyzed.

RESULTS

C-RADS classification rates for colorectal evaluation were C0 (0.7%), C1 (85.0%), C2 (8.6%), C3 (5.2%), and C4 (0.6%). Overall, 14.3% of subjects were positive (C2-C4), and positive findings were more frequent among men (17.5%) than women (11.6%; p < 0.0001). Positivity also increased with age, from 13.4% of patients 50-64 years old to 21.8% of patients 65-79 years old (p < 0.0001). Regarding extracolonic evaluation, 86.6% of patients were either negative for extracolonic findings or had unimportant extracolonic findings (E1 or E2). Likely unimportant but indeterminate extracolonic findings where further workup might be indicated (E3) were found in 11.3% of patients, whereas 2.1% had likely important extracolonic findings (E4). Overall, E3 and E4 rates were increased for older (p < 0.0001) and female (p = 0.008) cohorts.

CONCLUSION

C-RADS results from our initial experience with CTC screening may serve as an initial benchmark for program comparison and quality assurance measures.

Digital interactive learning of oral radiographic anatomy

INTRODUCTION

Studies reporting high number of diagnostic errors made from radiographs suggest the need to improve the learning of radiographic interpretation in the dental curriculum. Given studies that show student preference for computer-assisted or digital technologies, the purpose of this study was to develop an interactive digital tool and to determine whether it was more successful than a conventional radiology textbook in assisting dental students with the learning of radiographic anatomy.

MATERIALS AND METHODS

Eighty-eight dental students underwent a learning phase of radiographic anatomy using an interactive digital tool alongside a conventional radiology textbook. The success of the digital tool, when compared to the textbook, was assessed by quantitative means using a radiographic interpretation test and by qualitative means using a structured Likert scale survey, asking students to evaluate their own learning outcomes from the digital tool.

RESULTS

Student evaluations of the digital tool showed that almost all participants (95%) indicated that the tool positively enhanced their learning of radiographic anatomy and interpretation.

DISCUSSION

The success of the digital tool in assisting the learning of radiographic interpretation is discussed in the broader context of learning and teaching curricula, and preference (by students) for the use of this digital form when compared to the conventional literate form of the textbook.

CONCLUSION

Whilst traditional textbooks are still valued in the dental curriculum, it is evident that the preference for computer-assisted learning of oral radiographic anatomy enhances the learning experience by enabling students to interact and better engage with the course material.

The ACR BI-RADS experience: learning from history

The Breast Imaging Reporting and Data System (BI-RADS) initiative, instituted by the ACR, was begun in the late 1980s to address a lack of standardization and uniformity in mammography practice reporting. An important component of the BI-RADS initiative is the lexicon, a dictionary of descriptors of specific imaging features. The BI-RADS lexicon has always been data driven, using descriptors that previously had been shown in the literature to be predictive of benign and malignant disease. Once established, the BI-RADS lexicon provided new opportunities for quality assurance, communication, research, and improved patient care. The history of this lexicon illustrates a series of challenges and instructive successes that provide a valuable guide for other groups that aspire to develop similar lexicons in the future.

Anniversary paper: History and status of CAD and quantitative image analysis: the role of Medical Physics and AAPM

The roles of physicists in medical imaging have expanded over the years, from the study of imaging systems (sources and detectors) and dose to the assessment of image quality and perception, the development of image processing techniques, and the development of image analysis methods to assist in detection and diagnosis. The latter is a natural extension of medical physicists' goals in developing imaging techniques to help physicians acquire diagnostic information and improve clinical decisions. Studies indicate that radiologists do not detect all abnormalities on images that are visible on retrospective review, and they do not always correctly characterize abnormalities that are found. Since the 1950s, the potential use of computers had been considered for analysis of radiographic abnormalities. In the mid-1980s, however, medical physicists and radiologists began major research efforts for computer-aided detection or computer-aided diagnosis (CAD), that is, using the computer output as an aid to radiologists-as opposed to a completely automatic computer interpretation-focusing initially on methods for the detection of lesions on chest radiographs and mammograms. Since then, extensive investigations of computerized image analysis for detection or diagnosis of abnormalities in a variety of 2D and 3D medical images have been conducted. The growth of CAD over the past 20 years has been tremendous-from the early days of time-consuming film digitization and CPU-intensive computations on a limited number of cases to its current status in which developed CAD approaches are evaluated rigorously on large clinically relevant databases. CAD research by medical physicists includes many aspects-collecting relevant normal and pathological cases; developing computer algorithms appropriate for the medical interpretation task including those for segmentation, feature extraction, and classifier design; developing methodology for assessing CAD performance; validating the algorithms using appropriate cases to measure performance and robustness; conducting observer studies with which to evaluate radiologists in the diagnostic task without and with the use of the computer aid; and ultimately assessing performance with a clinical trial. Medical physicists also have an important role in quantitative imaging, by validating the quantitative integrity of scanners and developing imaging techniques, and image analysis tools that extract quantitative data in a more accurate and automated fashion. As imaging systems become more complex and the need for better quantitative information from images grows, the future includes the combined research efforts from physicists working in CAD with those working on quantitative imaging systems to readily yield information on morphology, function, molecular structure, and more-from animal imaging research to clinical patient care. A historical review of CAD and a discussion of challenges for the future are presented here, along with the extension to quantitative image analysis.

Multi-scaled morphological features for the characterization of mammographic masses using statistical classification schemes

OBJECTIVE

A comprehensive signal analysis approach on the mammographic mass boundary morphology is presented in this article. The purpose of this study is to identify efficient sets of simple yet effective shape features, employed in the original and multi-scaled spectral representations of the boundary, for the characterization of the mammographic mass. These new methods of mass boundary representation and processing in more than one domain greatly improve the information content of the base data that is used for pattern classification purposes, introducing comprehensive spectral and multi-scale wavelet versions of the original boundary signals. The evaluation is conducted against morphological and diagnostic characterization of the mass, using statistical methods, fractal dimension analysis and a wide range of classifier architectures.

METHODS AND MATERIALS

This study consists of (a) the investigation of the original radial distance measurements under the complete spectrum of signal analysis, (b) the application of curve feature extractors of morphological characteristics and the evaluation of the discriminative power of each one of them, by means of statistical significance analysis and dataset fractal dimension, and (c) the application of a wide range of classifier architectures on these morphological datasets, in order to conduct a comparative evaluation of the efficiency and effectiveness of all architectures, for mammographic mass characterization. Radial distance signal was exploited using the discrete Fourier transform (DFT) and the discrete wavelet transform (DWT) as additional carrier signals. Seven uniresolution feature functions were applied over these carrier signals and multiple shape descriptors were created. Classification was conducted against mass shape type and clinical diagnosis, using a wide range of linear and non-linear classifiers, including linear discriminant analysis (LDA), least-squares minimum distance (LSMD), k-nearest neighbor (k-NN), radial basis function (RBF) and multi-layered perceptron (MLP) neural networks (NN), and support vector machines (SVM). Fractal analysis was employed as a dataset analysis tool in the feature selection phase. The discriminative power of the features produced by this composite analysis is subsequently analyzed by means of multivariate analysis of variance (MANOVA) and tested against two distinct classification targets, namely (a) the morphological shape type of the mass and (b) the histologically verified clinical diagnosis for each mammogram.

RESULTS

Statistical analysis and classification results have shown that the discrimination value of the features extracted from the DWT components and especially the DFT spectrum, are of great importance. Furthermore, much of the information content of the curve features in the case of DFT and DWT datasets is directly related to the texture and fine-scale details of the corresponding envelope signal of the spectral components. Neural classifiers outperformed all other methods (SVM not used because they are mainly two-class classifiers) with overall success rate of 72.3% for shape type identification, while SVM achieved the overall highest 91.54% for clinical diagnosis. Receiver operating characteristic (ROC) analysis has been employed to present the sensitivity and specificity of the results of this study.

Mammographic masses characterization based on localized texture and dataset fractal analysis using linear, neural and support vector machine classifiers

OBJECTIVE

Localized texture analysis of breast tissue on mammograms is an issue of major importance in mass characterization. However, in contrast to other mammographic diagnostic approaches, it has not been investigated in depth, due to its inherent difficulty and fuzziness. This work aims to the establishment of a quantitative approach of mammographic masses texture classification, based on advanced classifier architectures and supported by fractal analysis of the dataset of the extracted textural features. Additionally, a comparison of the information content of the proposed feature set with that of the qualitative characteristics used in clinical practice by expert radiologists is presented.

METHODS AND MATERIAL

An extensive set of textural feature functions was applied to a set of 130 digitized mammograms, in multiple configurations and scales, constructing compact datasets of textural "signatures" for benign and malignant cases of tumors. These quantitative textural datasets were subsequently studied against a set of a thorough and compact list of qualitative texture descriptions of breast mass tissue, normally considered under a typical clinical assessment, in order to investigate the discriminating value and the statistical correlation between the two sets. Fractal analysis was employed to compare the information content and dimensionality of the textural features datasets with the qualitative information provided through medical diagnosis. A wide range of linear and non-linear classification architectures was employed, including linear discriminant analysis (LDA), least-squares minimum distance (LSMD), K-nearest-neighbors (K-nn), radial basis function (RBF) and multi-layer perceptron (MLP) artificial neural network (ANN), as well as support vector machine (SVM) classifiers. The classification process was used as the means to evaluate the inherent quality and informational content of each of the datasets, as well as the objective performance of each of the classifiers themselves in real classification of mammographic breast tumors against verified diagnosis.

RESULTS

Textural features extracted at larger scales and sampling box sizes proved to be more content-rich than their equivalents at smaller scales and sizes. Fractal analysis on the dimensionality of the textural datasets verified that reduced subsets of optimal feature combinations can describe the original feature space adequately for classification purposes and at least the same detail and quality as the list of qualitative texture descriptions provided by a human expert. Non-linear classifiers, especially SVMs, have been proven superior to any linear equivalent. Breast mass classification of mammograms, based only on textural features, achieved an optimal score of 83.9%, through SVM classifiers.

Development of radiology prediction models using feature analysis

RATIONALE AND OBJECTIVES

This article provides an introduction to prediction models and their application in diagnostic imaging research. Prediction models capitalize on the different degrees of association among variables to make a prediction of a health state, formulate a rule, or quantify individual contributions of various predictor variables. The purpose of this article is to elucidate the rationale, implication, and interpretation of prediction models using imaging features.

MATERIALS AND METHODS

The techniques and challenges of developing, testing, and implementing prediction models are described. Prediction model development methods are similar to data-mining techniques.

RESULTS

Learning objectives are to review prediction rule (model) methods, learn how prediction models may be applied to feature analysis, and understand the challenges of developing, testing, and implementing prediction models.

Significance analysis of qualitative mammographic features, using linear classifiers, neural networks and support vector machines

Advances in modern technologies and computers have enabled digital image processing to become a vital tool in conventional clinical practice, including mammography. However, the core problem of the clinical evaluation of mammographic tumors remains a highly demanding cognitive task. In order for these automated diagnostic systems to perform in levels of sensitivity and specificity similar to that of human experts, it is essential that a robust framework on problem-specific design parameters is formulated. This study is focused on identifying a robust set of clinical features that can be used as the base for designing the input of any computer-aided diagnosis system for automatic mammographic tumor evaluation. A thorough list of clinical features was constructed and the diagnostic value of each feature was verified against current clinical practices by an expert physician. These features were directly or indirectly related to the overall morphological properties of the mammographic tumor or the texture of the fine-scale tissue structures as they appear in the digitized image, while others contained external clinical data of outmost importance, like the patient's age. The entire feature set was used as an annotation list for describing the clinical properties of mammographic tumor cases in a quantitative way, such that subsequent objective analyses were possible. For the purposes of this study, a mammographic image database was created, with complete clinical evaluation descriptions and positive histological verification for each case. All tumors contained in the database were characterized according to the identified clinical features' set and the resulting dataset was used as input for discrimination and diagnostic value analysis for each one of these features. Specifically, several standard methodologies of statistical significance analysis were employed to create feature rankings according to their discriminating power. Moreover, three different classification models, namely linear classifiers, neural networks and support vector machines, were employed to investigate the true efficiency of each one of them, as well as the overall complexity of the diagnostic task of mammographic tumor characterization. Both the statistical and the classification results have proven the explicit correlation of all the selected features with the final diagnosis, qualifying them as an adequate input base for any type of similar automated diagnosis system. The underlying complexity of the diagnostic task has justified the high value of sophisticated pattern recognition architectures.

A study on several machine-learning methods for classification of malignant and benign clustered microcalcifications

In this paper, we investigate several state-of-the-art machine-learning methods for automated classification of clustered microcalcifications (MCs). The classifier is part of a computer-aided diagnosis (CADx) scheme that is aimed to assisting radiologists in making more accurate diagnoses of breast cancer on mammograms. The methods we considered were: support vector machine (SVM), kernel Fisher discriminant (KFD), relevance vector machine (RVM), and committee machines (ensemble averaging and AdaBoost), of which most have been developed recently in statistical learning theory. We formulated differentiation of malignant from benign MCs as a supervised learning problem, and applied these learning methods to develop the classification algorithm. As input, these methods used image features automatically extracted from clustered MCs. We tested these methods using a database of 697 clinical mammograms from 386 cases, which included a wide spectrum of difficult-to-classify cases. We analyzed the distribution of the cases in this database using the multidimensional scaling technique, which reveals that in the feature space the malignant cases are not trivially separable from the benign ones. We used receiver operating characteristic (ROC) analysis to evaluate and to compare classification performance by the different methods. In addition, we also investigated how to combine information from multiple-view mammograms of the same case so that the best decision can be made by a classifier. In our experiments, the kernel-based methods (i.e., SVM, KFD, and RVM) yielded the best performance (Az = 0.85, SVM), significantly outperforming a well-established, clinically-proven CADx approach that is based on neural network (Az = 0.80).

Qualitative diagnosis of calvarial metastasis by neural network and logistic regression

RATIONALE AND OBJECTIVES

To simplify the diagnostic features used by an artificial neural network compared with logistic regression (LR) in the diagnosis of calvarial metastasis with computed tomography and analyze their accuracy.

MATERIALS AND METHODS

Twenty-one of 167 patients with calvarial lesions were found to have metastasis. Clinical and computed tomography data were used for LR and neural network models. Both models were tested with the leave-one-out method. The final results of each model were compared using the area under receiver operating characteristic curve (Az).

RESULTS

The neural network identified metastasis significantly more successfully than LR with an Az of 0.9324 +/- 0.0386 versus 0.9192 +/- 0.0373, P = .01. The most important features selected by the LR and neural network were age and edge definition.

CONCLUSION

Neural networks offer wide possibilities over statistics for the study of calvarial metastases other than their minimum clinical and radiologic features for diagnosis.

International variation in screening mammography interpretations in community-based programs

BACKGROUND

Variations in mammography interpretations may have important clinical and economic implications. To evaluate international variability in mammography interpretation, we analyzed published reports from community-based screening programs from around the world.

METHODS

A total of 32 publications were identified in MEDLINE that fit the study inclusion criteria. Data abstracted from the publications included features of the population screened, examination technique, and clinical outcomes, including the percentage of mammograms judged to be abnormal, positive predictive value of an abnormal mammogram (PPV(A)), positive predictive value of a biopsy performed (PPV(B)), and percentages of breast cancer patients with ductal carcinoma in situ (DCIS) and minimal disease (DCIS and/or tumor size < or =10 mm). North American screening programs were compared with those from other countries using meta-regression analysis. All statistical tests were two-sided.

RESULTS

Wide ranges were noted for the percentage of mammograms judged to be abnormal (1.2%-15.0%), for PPV(A) (3.4%-48.7%), for PPV(B) (5.0%-85.2%), for percentage diagnosed with DCIS (4.3%-68.1%), and for percentage diagnosed with minimal disease (14.0%-80.6%). The percentage of mammograms judged to be abnormal were 2-4 percentage points higher in North American screening programs than they were in programs from other countries, after adjusting for covariates such as percentage of women who were less than 50 years of age and calendar year in which the mammogram was performed. The percentage of mammograms judged to be abnormal had a negative association with PPV(A) and PPV(B) (both P<.001) and a positive association with the frequency of DCIS cases diagnosed (P =.008) and the number of DCIS cases diagnosed per 1000 screens (P =.024); no consistent relationship was observed with the proportion of breast cancer diagnoses reported as having minimal disease or the number of minimal disease cases diagnosed per 1000 screens.

CONCLUSION

North American screening programs appear to interpret a higher percentage of mammograms as abnormal than programs from other countries without evident benefit in the yield of cancers detected per 1000 screens, although an increase in DCIS detection was noted.

Does training in the Breast Imaging Reporting and Data System (BI-RADS) improve biopsy recommendations or feature analysis agreement with experienced breast imagers at mammography?

PURPOSE

To determine whether training in the Breast Imaging Reporting and Data System (BI-RADS) improves observer performance and agreement with the consensus of experienced breast imagers with regard to mammographic feature analysis and final assessment.

MATERIALS AND METHODS

A test set of mammograms was developed, with 54 proven lesions consisting of 28 masses (nine [32%] malignancies) and 26 microcalcifications (10 [38%] malignancies). Three experienced breast imagers reviewed cases independently and by means of consensus. Twenty-three practicing mammogram-interpreting physicians reviewed mammograms before and after a day's lectures on BI-RADS. Observer performance before and after training was measured by means of agreement (kappa) with consensus description and assessments, rate of biopsy of malignant and benign lesions, and areas under receiver operating characteristic (ROC) curves. Performance was also measured for 11 participants 2-3 months after training.

RESULTS

Improved agreement with consensus feature analysis was found for mass margins and/or asymmetries, with a pretraining generalized kappa value of 0.36 and a posttraining generalized kappa value of 0.41. Similar improvement was seen for description of calcification morphology (pretraining kappa value of 0.36 improving to 0.44 after training). No improvement was seen in describing calcification distribution. Final assessments were more consistent after training, with a pretraining kappa value of 0.31, as compared with 0.45 after training. The mean biopsy rate for malignant lesions improved from 73% (range, 53%-89%) before training to 88% (range, 74%-100%) after training, with minimal increase in mean biopsy rate of benign lesions (43% [range, 26%-60%] before to 51% [range, 31%-63%] after training), and no net change in area under the ROC curve, as compared with histopathologic findings. For the subset of participants with delayed follow-up, no significant decline in posttraining results was seen.

CONCLUSION

BI-RADS training resulted in improved agreement with the consensus of experienced breast imagers for feature analysis and final assessment. It is important that trainees showed improved rates of recommending biopsy for malignant lesions. This effect was maintained over 2-3 months.

Use of the American College of Radiology BI-RADS guidelines by community radiologists: concordance of assessments and recommendations assigned to screening mammograms

OBJECTIVE

This study evaluated the use of the American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) by community radiologists by determining the concordance of assessment categories and recommendations assigned to screening mammograms.

MATERIALS AND METHODS

The study comprised the interpretations of 82,620 consecutive screening mammograms by 18 radiologists between January 1, 1995, and December 31, 1998. For all mammograms, assessment categories and recommendations were compared to determine whether they were in accordance with BI-RADS guidelines. Overall patterns of discordance were analyzed, and comparisons of discordant patterns by assessment category, patient age, breast density, and year of examination were made.

RESULTS

The overall discordance between BI-RADS assessments and recommendations was low (3%). The assessment with the highest discordance was "probably benign finding" (category 3), at 53.5%. Mammograms obtained in 1998 were almost half as likely to have assessment-recommendation discordance compared with those obtained in 1995 (2.4% vs 4.5%, respectively; odds ratio = 0.52; p < 0.001). Mammograms of women with dense breast tissue were 30% more likely to have lesions assigned discordant assessments and recommendations compared with those of women with fatty tissue (3.4% vs 2.7%, respectively; odds ratio = 1.3; p < 0.001). No differences in the patterns of discordance were found between mammograms of women younger than 50 years and those of women 50 years old and older (p = 0.10).

CONCLUSION

There has been improvement in the accurate application of BI-RADS since its introduction. However, variation in the pairing of BI-RADS assessments and recommendations persists. Continued efforts to educate radiologists about the use of BI-RADS and to clarify BI-RADS terms would promote maximum consistency in this use of this reporting method.

Multimodality diagnosis of liver tumors: feature analysis with CT, liver-specific and contrast-enhanced MR, and a computer model

RATIONALE AND OBJECTIVES

The purpose of this study was to measure and to clarify the diagnostic contributions of image-based features in differentiating benign from malignant and hepatocyte-containing from non-hepatocyte-containing liver lesions.

MATERIALS AND METHODS

Six experienced abdominal radiologists each read images from 146 cases (including a contrast material-enhanced computed tomographic [CT] scan and contrast-enhanced and unenhanced magnetic resonance [MR] images) following a checklist-questionnaire requiring them to rate quantitatively each of as many as 131 image features and then reported on each of the two differentiations. The diagnostic value of each feature was assessed, and linear discriminant analysis was used to develop statistical prediction rules (SPRs) for merging feature data into computerized "second opinions." For the two differentiations, accuracy (area under the receiver operating characteristic curve [Az]) was then determined for the radiologists' readings by themselves and for each of three SPRs.

RESULTS

Thirty-seven candidate features had diagnostic value for each of the two differentiations (a slightly different feature set for each). Radiologists' performance at both differentiations was excellent (Az = 0.929 [benign vs malignant] and 0.926 [hepatocyte-containing vs non-hepatocyte-containing]). Performance of the SPR that operated on the features from all modalities together was better than that of radiologists (Az = 0.936 [benign vs malignant] and 0.951 [hepatocyte-containing vs non-hepatocyte-containing]), but this difference was of marginal statistical significance (P = .11). Contrast-enhanced MR imaging and contrast-enhanced CT each made significant adjunctive contributions to accuracy compared with unenhanced MR imaging alone.

CONCLUSION

Many CT- and MR imaging-based features have diagnostic value in differentiating benign from malignant and hepatocyte-containing from non-hepatocyte-containing liver lesions. Radiologists could also benefit from the fully informed SPR's "second opinions."

Biopsy of amorphous breast calcifications: pathologic outcome and yield at stereotactic biopsy

PURPOSE

To assess the pathologic outcome of amorphous breast calcifications and the success of stereotactic biopsy for such lesions.

MATERIALS AND METHODS

From July 1995 through February 2000, biopsy of all clustered amorphous calcifications not clearly stable for at least 5 years or in a diffuse scattered distribution was recommended. Logistic regression analysis was used to stratify the risk of malignancy by patient risk factors, calcification distribution, and stability.

RESULTS

Calcifications were retrieved from 150 biopsies; 30 (20%) proved malignant and included 27 ductal carcinomas in situ and three low-grade invasive and intraductal carcinomas (2-5 mm). Another 30 (20%) yielded high-risk lesions, including 21 atypical ductal hyperplasia, eight atypical lobular hyperplasia, and one lobular carcinoma in situ. In 150 lesions, stereotactic biopsy was performed on 113 and aborted in 10. Calcifications were retrieved from all 113 stereotactic biopsies. Of those with calcification retrieval, there were three histologic underestimates (accuracy, 97%). Stereotactic biopsy spared a surgical procedure in 57 (46%) of 123 patients. Needle localization was required for 23 (15%) of 150 patients due to poor conspicuity. Five (45%) of 11 biopsies performed in women with ipsilateral breast cancer showed malignancy (P = .025). When multiple lesions of amorphous calcifications were present in one breast, sampling of one reliably predicted the outcome of others.

CONCLUSION

We found a substantial rate of ductal carcinoma in situ and high-risk lesions associated with amorphous calcifications. Stereotactic biopsy can be successfully performed for the majority of subtle amorphous calcifications; however, only a minority were spared a surgical procedure.

Effect of training with the american college of radiology breast imaging reporting and data system lexicon on mammographic interpretation skills in developing countries

RATIONALE AND OBJECTIVES

The authors evaluated the effect of training in the American College of Radiology (ACR) Breast Imaging Reporting and Data System (BI-RADS) lexicon on the interpretive skills of radiologists evaluating screening mammograms in Ukraine.

MATERIALS AND METHODS

As part of a program to improve breast cancer detection and treatment in Ukraine, a series of training sessions was given to a group of radiologists across Ukraine to improve their interpretive skills in screening mammography. The training sessions focused on the use of the lexicon and assessment categories developed by the ACR BI-RADS committee. Participants (n = 14) evaluated 30 test screening mammograms before and after the training sessions. The test sets were randomly selected from a larger collection of training sets containing normal, benign, and abnormal mammograms. False-positive, false-negative, true-positive, and true-negative evaluations were determined, and sensitivity, specificity, and positive predictive values were calculated for each participant before and after training.

RESULTS

The mean baseline sensitivity, specificity, and positive predictive values were 50%, 77%, and 43%, respectively. Each of these measures of interpretive skills improved significantly after training in the use of the lexicon, to 87%, 89%, and 78% (P < .0001, P < .01, and P < .0001, respectively).

CONCLUSION

As the use of mammography spreads throughout developing countries, it is essential to address training and educational needs, as well as equipment needs. The ACR BI-RADS lexicon provides a systematic and efficient method for training radiologists to interpret screening mammograms. Educating radiologists on the use of this lexicon proved an effective way to improve their interpretive skills in screening mammography.

The reliability issue of computer-aided breast cancer diagnosis

This paper introduces a number of reliability criteria for computer-aided diagnostic systems for breast cancer. These criteria are then used to analyze some published neural network systems. It is also shown that the property of monotonicity for the data is rather natural in this medical domain, and it has the potential to significantly improve the reliability of breast cancer diagnosis while maintaining a general representation power. A central part of this paper is devoted to the representation/narrow vicinity hypothesis, upon which existing computer-aided diagnostic methods heavily rely. The paper also develops a framework for determining the validity of this hypothesis. The same framework can be used to construct a diagnostic procedure with improved reliability.

False-negative mammograms. Medical, legal, and risk management implications

From a strictly biologic perspective, delay in diagnosis of breast cancer is axiomatic. The number of cell divisions that must occur before detection is possible by either clinical or mammographic methods means that a finite time has occurred in which the outcome for any given case may have already been determined. That early detection and diagnosis of breast cancer lead to improved survival may be intuitive, but clinical trials have been necessary to validate the concept. Delay in diagnosis is unavoidable but the period of delay may be lessened in many cases, prompting earlier intervention and impacting outcomes. Mammography is an important vehicle for such earlier intervention and the issue of the false-negative mammogram is of concern to the radiology community, the lay community, and the courts. Mammographic interpretation has not yet approached a sufficiently standardized benchmark. Detection and diagnosis are dependent on a series of factors that need to be integrated to achieve the dual goals of timely intervention for bonafide purposes and reduction of unnecessary procedures and interventions. Some of the reasons for delay in diagnosis are unavoidable, beginning with the absence of clinical or imaging features of malignancy and extending to limitations of sufficiently specific features to prompt intervention. On the other hand, other reasons are avoidable and attention to many of these causes should lessen the incidence of such delay. Regardless of the reason, those women who feel that their breast cancer should have been diagnosed at an earlier time may consider subjecting their mammographic studies to independent reviews. At such a point, the precise reasons for delay may be better analyzed, all in an attempt to provide an adequate reconciliation of what has come to be known as the false-negative mammogram.

Breast Imaging Reporting and Data System: inter- and intraobserver variability in feature analysis and final assessment

OBJECTIVE

We sought to evaluate the use of the Breast Imaging Reporting and Data System (BI-RADS) standardized mammography lexicon among and within observers and to distinguish variability in feature analysis from variability in lesion management.

MATERIALS AND METHODS

Five experienced mammographers, not specifically trained in BI-RADS, used the lexicon to describe and assess 103 screening mammograms, including 30 (29%) showing cancer, and a subset of 86 mammograms with diagnostic evaluation, including 23 (27%) showing cancer. A subset of 13 screening mammograms (two with malignant findings, 11 with diagnostic evaluation) were rereviewed by each observer 2 months later. Kappa statistics were calculated as measures of agreement beyond chance.

RESULTS

After diagnostic evaluation, the interobserver kappa values for describing features were as follows: breast density, 0.43; lesion type, 0.75; mass borders, 0.40; special cases, 0.56; mass density, 0.40; mass shape, 0.28; microcalcification morphology, 0.36; and microcalcification distribution, 0.47. Lesion management was highly variable, with a kappa value for final assessment of 0.37. When we grouped assessments recommending immediate additional evaluation and biopsy (BI-RADS categories 0, 4, and 5 combined) versus follow-up (categories 1, 2, and 3 combined), five observers agreed on management for only 47 (55%) of 86 lesions. Intraobserver agreement on management (additional evaluation or biopsy versus follow-up) was seen in 47 (85%) of 55 interpretations, with a kappa value of 0.35-1.0 (mean, 0.60) for final assessment.

CONCLUSION

Inter- and intraobserver variability in mammographic interpretation is substantial for both feature analysis and management. Continued development of methods to improve standardization in mammographic interpretation is needed.

Quantitative characterization of mass lesions on digitized mammograms for computer-assisted diagnosis

RATIONALE AND OBJECTIVES

To investigate features for discriminating benign from malignant mammographic findings by using computer-aided diagnosis (CAD) and to test the accuracy of CAD interpretations of mass lesions.

METHODS

Fifty-five sequential, mammographically detected mass lesions, referred for biopsy, were digitized for computerized reevaluation with a CAD system. Quantitative features that characterize spiculation were automatically extracted by the CAD system. Data generated by 271 known retrospective cases were used to set reference values indicating the range for malignant and benign lesions. After conventional interpretation of the 55 prospective cases, they were evaluated a second time by the radiologist using the extracted features and the reference ranges. In addition, a pattern-recognition scheme based on the extracted features was used to classify the prospective cases. Accuracy of interpretation with and without the CAD system was evaluated using receiver operating characteristic (ROC) curve analysis.

RESULTS

Sensitivity of the CAD diagnosis for the prospective cases improved from 92% to 100%. Specificity improved significantly from 26.7% to 66.7%. This was accompanied by a significant increase in the accuracy of diagnosis from 56.4% to 81.8% and in the positive predictive value from 51.1% to 71.4%. The Az for the CAD ROC curve significantly increased from 0.73 to 0.90. The performance of the classification scheme was slightly lower than that of the radiologists' interpretation with the CAD system.

CONCLUSIONS

Use of the CAD system significantly improved the accuracy of diagnosis. The findings suggest that the classification scheme may improve the radiologist's ability to differentiate benign from malignant mass lesions in the interpretation of mammograms.

The development and evaluation of CADMIUM: a prototype system to assist in the interpretation of mammograms

We have developed CADMIUM, a novel approach for the design of systems to assist in the interpretation of medical images. CADMIUM uses symbolic reasoning to relate information obtained from image processing to the decisions radiologists take. The approach is based on a symbolic decision procedure which has already been used successfully in a variety of nonimaging clinical decision systems. In CADMIUM this decision procedure is extended with models of three generic image interpretation tasks: detection, measurement and classification of image features. The extended procedure is used to construct the lines of reasoning needed in each task and to control the acquisition of information by image processing. CADMIUM has been evaluated as an aid to the differential diagnosis of microcalcifications on mammographic images. Radiographers who had been trained to interpret images performed better when using the advice provided by the system.

Improvement of radiologists' characterization of mammographic masses by using computer-aided diagnosis: an ROC study

PURPOSE

To evaluate the effects of computer-aided diagnosis (CAD) on radiologists' classification of malignant and benign masses seen on mammograms.

MATERIALS AND METHODS

The authors previously developed an automated computer program for estimation of the relative malignancy rating of masses. In the present study, the authors conducted observer performance experiments with receiver operating characteristic (ROC) methodology to evaluate the effects of computer estimates on radiologists' confidence ratings. Six radiologists assessed biopsy-proved masses with and without CAD. Two experiments, one with a single view and the other with two views, were conducted. The classification accuracy was quantified by using the area under the ROC curve, Az.

RESULTS

For the reading of 238 images, the Az value for the computer classifier was 0.92. The radiologists' Az values ranged from 0.79 to 0.92 without CAD and improved to 0.87-0.96 with CAD. For the reading of a subset of 76 paired views, the radiologists' Az values ranged from 0.88 to 0.95 without CAD and improved to 0.93-0.97 with CAD. Improvements in the reading of the two sets of images were statistically significant (P = .022 and .007, respectively). An improved positive predictive value as a function of the false-negative fraction was predicted from the improved ROC curves.

CONCLUSION

CAD may be useful for assisting radiologists in classification of masses and thereby potentially help reduce unnecessary biopsies.

Improving breast cancer diagnosis with computer-aided diagnosis

RATIONALE AND OBJECTIVES

The purpose of this study was to test whether computer-aided diagnosis (CAD) can improve radiologists' performance in breast cancer diagnosis.

MATERIALS AND METHODS

The computer classification scheme used in this study estimates the likelihood of malignancy for clustered microcalcifications based on eight computer-extracted features obtained from standard-view mammograms. One hundred four histologically verified cases of microcalcifications (46 malignant, 58 benign) in a near-consecutive biopsy series were used in this study. Observer performance was measured on 10 radiologists who read the original standard- and magnification-view mammograms. The computer aid provided a percentage estimate of the likelihood of malignancy. Comparison was made between computer-aided performance and unaided (routine clinical) performance by using receiver operating characteristic (ROC) analysis and by comparing biopsy recommendations.

RESULTS

The average ROC curve area (Az) increased from 0.61 without aid to 0.75 with the computer aid (P < .0001). On average, with the computer aid, each observer recommended 6.4 additional biopsies for cases with malignant lesions (P = .0006) and 6.0 fewer biopsies for cases with benign lesions (P = .003). This improvement corresponded to increases in sensitivity (from 73.5% to 87.4%), specificity (from 31.6% to 41.9%), and hypothetical positive biopsy yield (from 46% to 55%).

CONCLUSION

CAD can be used to improve radiologists' performance in breast cancer diagnosis.

Computerized analysis of mammographic microcalcifications in morphological and texture feature spaces

We are developing computerized feature extraction and classification methods to analyze malignant and benign microcalcifications on digitized mammograms. Morphological features that described the size, contrast, and shape of microcalcifications and their variations within a cluster were designed to characterize microcalcifications segmented from the mammographic background. Texture features were derived from the spatial gray-level dependence (SGLD) matrices constructed at multiple distances and directions from tissue regions containing microcalcifications. A genetic algorithm (GA) based feature selection technique was used to select the best feature subset from the multi-dimensional feature spaces. The GA-based method was compared to the commonly used feature selection method based on the stepwise linear discriminant analysis (LDA) procedure. Linear discriminant classifiers using the selected features as input predictor variables were formulated for the classification task. The discriminant scores output from the classifiers were analyzed by receiver operating characteristic (ROC) methodology and the classification accuracy was quantified by the area, Az, under the ROC curve. We analyzed a data set of 145 mammographic microcalcification clusters in this study. It was found that the feature subsets selected by the GA-based method are comparable to or slightly better than those selected by the stepwise LDA method. The texture features (Az = 0.84) were more effective than morphological features (Az = 0.79) in distinguishing malignant and benign microcalcifications. The highest classification accuracy (Az = 0.89) was obtained in the combined texture and morphological feature space. The improvement was statistically significant in comparison to classification in either the morphological (p = 0.002) or the texture (p = 0.04) feature space alone. The classifier using the best feature subset from the combined feature space and an appropriate decision threshold could correctly identify 35% of the benign clusters without missing a malignant cluster. When the average discriminant score from all views of the same cluster was used for classification, the Az value increased to 0.93 and the classifier could identify 50% of the benign clusters at 100% sensitivity for malignancy. Alternatively, if the minimum discriminant score from all views of the same cluster was used, the Az value would be 0.90 and a specificity of 32% would be obtained at 100% sensitivity. The results of this study indicate the potential of using combined morphological and texture features for computer-aided classification of microcalcifications.

Focal calvarial bone lesions. Comparison of logistic regression and neural network models

RATIONALE AND OBJECTIVES

To assess the accuracy of logistic regression (LR) and artificial neural networks (NN) in the diagnosis of calvarial lesions using computed tomography (CT) and to establish the importance of the different features needed for the diagnosis.

METHODS

One hundred sixty-seven patients with calvarial lesions as the only known disease were enrolled. The clinical and CT data were used for LR and NN models. Both models were tested with the leave-one-out method. The final results of each model were compared using the area under receiver operating characteristic curves (Az).

RESULTS

Of the lesions, 73.1%, were benign and 26.9% were malignant. There was no statistically significant difference between LR and NN in differentiating malignancy. In characterizing the histologic diagnoses, NN was statistically superior to LR. Important NN features needed for malignancy classification were age and edge definition, and for the histologic diagnoses matrix, marginal sclerosis, and age.

CONCLUSION

NNs offer wide possibilities over statistics for the study of calvarial lesions other than their superior diagnostic performance.

Voice-activated retrieval of mammography reference images

We undertook this project to integrate context sensitive computer-based educational and decision making aids into the film interpretation and reporting process, and to determine the clinical utility of this method as a guide for further system development. An image database of 347 digital mammography images was assembled and image features were coded. An interface was developed to a computerized speech recognition radiology reporting system which was modified to translate reported findings into database search terms. These observations were used to formulate database search strategies which not only retrieved similar cases from the image database, but also other cases that were related to the index case in different ways. The search results were organized into image sets intended to address common questions that arise during image interpretation. An evaluation of the clinical utility of this method was performed as a guide for further system development. We found that voice dictation of prototypical mammographic cases resulted in automatic retrieval of reference images. The retrieved images were organized into sets matching findings, diagnostic hypotheses, diagnosis, spectrum of findings or diagnoses, closest match to dictated case, or user specified parameters. Two mammographers graded the clinical utility of each form of system output. We concluded that case specific and problem specific image sets may be automatically generated from spoken case dictation. A potentially large number of retrieved images may be divided into subsets which anticipate common clinical problems. This automatic method of context sensitive image retrieval may provide a "continuous" form of education integrated into routine case interpretation.

Computerized characterization of masses on mammograms: the rubber band straightening transform and texture analysis

A new rubber band straightening transform (RBST) is introduced for characterization of mammographic masses as malignant or benign. The RBST transforms a band of pixels surrounding a segmented mass onto the Cartesian plane (the RBST image). The border of a mammographic mass appears approximately as a horizontal line, and possible speculations resemble vertical lines in the RBST image. In this study, the effectiveness of a set of directional textures extracted from the images before the RBST. A database of 168 mammograms containing biopsy-proven malignant and benign breast masses was digitized at a pixel size of 100 microns x 100 microns. Regions of interest (ROIs) containing the biopsied mass were extracted from each mammogram by an experienced radiologist. A clustering algorithm was employed for automated segmentation of each ROI into a mass object and background tissue. Texture features extracted from spatial gray-level dependence matrices and run-length statistics matrices were evaluated for three different regions and representations: (i) the entire ROI; (ii) a band of pixels surrounding the segmented mass object in the ROI; and (iii) the RBST image. Linear discriminant analysis was used for classification, and receiver operating characteristic (ROC) analysis was used to evaluate the classification accuracy. Using the ROC curves as the performance measure, features extracted from the RBST images were found to be significantly more effective than those extracted from the original images. Features extracted from the RBST images yielded an area (Az) of 0.94 under the ROC curve for classification of mammographic masses as malignant and benign.

Automated computerized classification of malignant and benign masses on digitized mammograms

RATIONALE AND OBJECTIVES

To develop a method for differentiating malignant from benign masses in which a computer automatically extracts lesion features and merges them into an estimated likelihood of malignancy.

MATERIALS AND METHODS

Ninety-five mammograms depicting masses in 65 patients were digitized. Various features related to the margin and density of each mass were extracted automatically from the neighborhoods of the computer-identified mass regions. Selected features were merged into an estimated likelihood of malignancy by, using three different automated classifiers. The performance of the three classifiers in distinguishing between benign and malignant masses was evaluated by receiver operating characteristic analysis and compared with the performance of an experienced mammographer and that of five less experienced mammographers.

RESULTS

Our computer classification scheme yielded an area under the receiver operating characteristic curve (Az) value of 0.94, which was similar to that for an experienced mammographer (Az = 0.91) and was statistically significantly higher than the average performance of the radiologists with less mammographic experience (Az = 0.81) (P = .013). With the database used, the computer scheme achieved, at 100% sensitivity, a positive predictive value of 83%, which was 12% higher than that for the performance of the experienced mammographer and 21% higher than that for the average performance of the less experienced mammographers (P < .0001).

CONCLUSION

Automated computerized classification schemes may be useful in helping radiologists distinguish between benign and malignant masses and thus reducing the number of unnecessary biopsies.

Computerized classification of malignant and benign microcalcifications on mammograms: texture analysis using an artificial neural network

We investigated the feasibility of using texture features extracted from mammograms to predict whether the presence of microcalcifications is associated with malignant or benign pathology. Eighty-six mammograms from 54 cases (26 benign and 28 malignant) were used as case samples. All lesions had been recommended for surgical biopsy by specialists in breast imaging. A region of interest (ROI) containing the microcalcifications was first corrected for the low-frequency background density variation. Spatial grey level dependence (SGLD) matrices at ten different pixel distances in both the axial and diagonal directions were constructed from the background-corrected ROI. Thirteen texture measures were extracted from each SGLD matrix. Using a stepwise feature selection technique, which maximized the separation of the two class distributions, subsets of texture features were selected from the multi-dimensional feature space. A backpropagation artificial neural network (ANN) classifier was trained and tested with a leave-one-case-out method to recognize the malignant or benign microcalcification clusters. The performance of the ANN was analysed with receiver operating characteristic (ROC) methodology. It was found that a subset of six texture features provided the highest classification accuracy among the feature sets studied. The ANN classifier achieved an area under the ROC curve of 0.88. By setting an appropriate decision threshold, 11 of the 28 benign cases were correctly identified (39% specificity) without missing any malignant cases (100% sensitivity) for patients who had undergone biopsy. This preliminary result indicates that computerized texture analysis can extract mammographic information that is not apparent by visual inspection. The computer-extracted texture information may be used to assist in mammographic interpretation, with the potential to reduce biopsies of benign cases and improve the positive predictive value of mammography.

Construction of a Bayesian network for mammographic diagnosis of breast cancer

Bayesian networks use the techniques of probability theory to reason under uncertainty, and have become an important formalism for medical decision support systems. We describe the development and validation of a Bayesian network (MammoNet) to assist in mammographic diagnosis of breast cancer. MammoNet integrates five patient-history features, two physical findings, and 15 mammographic features extracted by experienced radiologists to determine the probability of malignancy. We outline the methods and issues in the system's design, implementation, and evaluation. Bayesian networks provide a potentially useful tool for mammographic decision support.

Invited review: computer aids for decision-making in diagnostic radiology--a literature review

This review looks at a variety of different ways in which computers can be used to assist in the interpretation of radiological images and in radiological decision-making. The issues involved in the design of computerized decision aids are introduced and four criteria proposed for evaluating such aids: need, practicality, veracity and relevance. These criteria are used to assess research into decision aids based on: image databases, numerical methods, expert systems, image processing and image understanding systems. Possible directions for research leading to aids of practical value are discussed in the conclusion.

A personal-computer semiautomated report-coding system for diagnostic mammography

A computerized report-coding system for diagnostic mammography using a personal computer was developed. Four years' experience with the radiologist-oriented system is presented. The data input time for report generation is in most cases less than 30 seconds by radiologists familiar with keyboard entry. Suspicious and malignant findings are dictated in the conventional way. About 80% of the mammographic examinations at the university Breast Imaging Center were suitable for standardized reporting. Radiologist-generated reports using a personal computer might be an alternative to transcriptionist-oriented systems.

Combining evidence from multiple imaging modalities: a feature-analysis method

This study was designed to develop methods to improve radiologists' ability to detect and diagnose breast cancer. We evaluated the ability of a feature-analysis method to help radiologists merge judgements constructively from two rather disparate breast imaging tests. To accomplish these goals, we developed a list of perceptual features and quantitated the importance of each in the diagnosis of patients having both diaphanography (Test 1) and mammography (Test 2). Then, two decision aids were developed: One was a checklist of the critical diagnostic visual features from both tests that also assisted readers in rating these features numerically. The second was a computer-based classifier that assisted readers in merging the assessments of the two tests into one overall diagnostic probability. The value of these aids was assessed by comparing radiologists' accuracy in reading a set of proven cases in their standard fashion with their accuracy when reading in an enhanced mode, utilizing the checklist and computer classifier. When Test 1 was read adjunctively with Test 2, use of the decision aids led to a significant improvement in accuracy (p = .013) over the unenhanced, combined readings. For Test 1 alone, the aids led to a significant improvement over its low level of unenhanced reading (p = .046). For Test 2 alone, the enhancements provided little gain in accuracy over an already high level of performance on the full case set (p = .081), although significant gains were realized on the most difficult ones. We conclude that methods to aid standardization and merging of feature-based judgements can improve radiologists performance on complex diagnostic tasks.