An Interactive System for Computer-Aided Diagnosis of Breast Masses

Although mammography is the only clinically accepted imaging modality for screening the general population to detect breast cancer, interpreting mammograms is difficult with lower sensitivity and specificity. To provide radiologists “a visual aid” in interpreting mammograms, we developed and tested an interactive system for computer-aided detection and diagnosis (CAD) of mass-like cancers. Using this system, an observer can view CAD-cued mass regions depicted on one image and then query any suspicious regions (either cued or not cued by CAD). CAD scheme automatically segments the suspicious region or accepts manually defined region and computes a set of image features. Using content-based image retrieval (CBIR) algorithm, CAD searches for a set of reference images depicting “abnormalities” similar to the queried region. Based on image retrieval results and a decision algorithm, a classification score is assigned to the queried region. In this study, a reference database with 1,800 malignant mass regions and 1,800 benign and CAD-generated false-positive regions was used. A modified CBIR algorithm with a new function of stretching the attributes in the multi-dimensional space and decision scheme was optimized using a genetic algorithm. Using a leave-one-out testing method to classify suspicious mass regions, we compared the classification performance using two CBIR algorithms with either equally weighted or optimally stretched attributes. Using the modified CBIR algorithm, the area under receiver operating characteristic curve was significantly increased from 0.865 ± 0.006 to 0.897 ± 0.005 (p < 0.001). This study demonstrated the feasibility of developing an interactive CAD system with a large reference database and achieving improved performance.     

Creation and Implementation of Department-Wide Structured Reports: An Analysis of the Impact on Error Rate in Radiology Reports

The purpose of this study was to evaluate and compare textual error rates and subtypes in radiology reports before and after implementation of department-wide structured reports. Randomly selected radiology reports that were generated following the implementation of department-wide structured reports were evaluated for textual errors by two radiologists. For each report, the text was compared to the corresponding audio file. Errors in each report were tabulated and classified. Error rates were compared to results from a prior study performed prior to implementation of structured reports. Calculated error rates included the average number of errors per report, average number of nongrammatical errors per report, the percentage of reports with an error, and the percentage of reports with a nongrammatical error. Identical versions of voice-recognition software were used for both studies. A total of 644 radiology reports were randomly evaluated as part of this study. There was a statistically significant reduction in the percentage of reports with nongrammatical errors (33 to 26 %; p = 0.024). The likelihood of at least one missense omission error (omission errors that changed the meaning of a phrase or sentence) occurring in a report was significantly reduced from 3.5 to 1.2 % (p = 0.0175). A statistically significant reduction in the likelihood of at least one comission error (retained statements from a standardized report that contradict the dictated findings or impression) occurring in a report was also observed (3.9 to 0.8 %; p = 0.0007). Carefully constructed structured reports can help to reduce certain error types in radiology reports.     

Integration of Imaging Signs into RadLex

Imaging signs form an important part of the language of radiology, but are not represented in established lexicons. We sought to incorporate imaging signs into RSNA''s RadLex® ontology of radiology terms. Names of imaging signs and their definitions were culled from books, journal articles, dictionaries, and biomedical web sites. Imaging signs were added into RadLex as subclasses of the term “imaging sign,” which was defined in RadLex as a subclass of “imaging observation.” A total of 743 unique imaging signs were added to RadLex with their 392 synonyms to yield a total of 1,135 new terms. All included definitions and related RadLex terms, including imaging modality, anatomy, and disorder, when appropriate. The information will allow RadLex users to identify imaging signs by modality (e.g., ultrasound signs) and to find all signs related to specific pathophysiology. The addition of imaging signs to RadLex augments its use to index the radiology literature, create and interpret clinical radiology reports, and retrieve relevant cases and images.     

A comparison of reconstruction algorithms for breast tomosynthesis

Three algorithms for breast tomosynthesis reconstruction were compared in this paper, including (1) a back-projection (BP) algorithm (equivalent to the shift-and-add algorithm), (2) a Feldkamp filtered back-projection (FBP) algorithm, and (3) an iterative Maximum Likelihood (ML) algorithm. Our breast tomosynthesis system acquires 11 low-dose projections over a 50 degree angular range using an a-Si (CsI:Tl) flat-panel detector. The detector was stationary during the acquisition. Quality metrics such as signal difference to noise ratio (SDNR) and artifact spread function (ASF) were used for quantitative evaluation of tomosynthesis reconstructions. The results of the quantitative evaluation were in good agreement with the results of the qualitative assessment. In patient imaging, the superimposed breast tissues observed in two-dimensional (2D) mammograms were separated in tomosynthesis reconstructions by all three algorithms. It was shown in both phantom imaging and patient imaging that the BP algorithm provided the best SDNR for low-contrast masses but the conspicuity of the feature details was limited by interplane artifacts; the FBP algorithm provided the highest edge sharpness for microcalcifications but the quality of masses was poor; the information of both the masses and the microcalcifications were well restored with balanced quality by the ML algorithm, superior to the results from the other two algorithms.     

Marker-Controlled Watershed for Lesion Segmentation in Mammograms

Lesion segmentation, which is a critical step in computer-aided diagnosis system, is a challenging task as lesion boundaries are usually obscured, irregular, and low contrast. In this paper, an accurate and robust algorithm for the automatic segmentation of breast lesions in mammograms is proposed. The traditional watershed transformation is applied to the smoothed (by the morphological reconstruction) morphological gradient image to obtain the lesion boundary in the belt between the internal and external markers. To automatically determine the internal and external markers, the rough region of the lesion is identified by a template matching and a thresholding method. Then, the internal marker is determined by performing a distance transform and the external marker by morphological dilation. The proposed algorithm is quantitatively compared to the dynamic programming boundary tracing method and the plane fitting and dynamic programming method on a set of 363 lesions (size range, 5–42 mm in diameter; mean, 15 mm), using the area overlap metric (AOM), Hausdorff distance (HD), and average minimum Euclidean distance (AMED). The mean ± SD of the values of AOM, HD, and AMED for our method were respectively 0.72 ± 0.13, 5.69 ± 2.85 mm, and 1.76 ± 1.04 mm, which is a better performance than two other proposed segmentation methods. The results also confirm the potential of the proposed algorithm to allow reliable segmentation and quantification of breast lesion in mammograms.     

Efficacy of a Checklist-Style Structured Radiology Reporting Template in Reducing Resident Misses on Cervical Spine Computed Tomography Examinations

The increasing use of medical checklists to promote patient safety raises the question of their utility in diagnostic radiology. This study evaluates the efficacy of a checklist-style reporting template in reducing resident misses on cervical spine CT examinations. A checklist-style reporting template for cervical spine CTs was created at our institution and mandated for resident preliminary reports. Ten months after implementation of the template, we performed a retrospective cohort study comparing rates of emergent pathology missed on reports generated with and without the checklist-style reporting template. In 1,832 reports generated without using the checklist-style template, 25 (17.6 %) out of 142 emergent findings were missed. In 1,081 reports generated using the checklist-style template, 13 (11.9 %) out of 109 emergent findings were missed. The decrease in missed pathology was not statistically significant (p = 0.21). However, larger differences were noted in the detection of emergent non-fracture findings, with 17 (28.3 %) out of 60 findings missed on reports without use of the checklist template and 5 (9.3 %) out of 54 findings missed on reports using the checklist template, representing a statistically significant decrease in missed non-fracture findings (p = 0.01). The use of a checklist-style structured reporting template resulted in a statistically significant decrease in missed non-fracture findings on cervical spine CTs. The lack of statistically significant change in missed fractures was expected given that residents’ search patterns naturally include fracture detection. Our findings suggest that the use of checklists in structured reporting may increase diagnostic accuracy.     

Retrospective Review of the Drop in Observer Detection Performance Over Time in Lesion-enriched Experimental Studies

The vigilance decrement describes a decrease in sensitivity or increase in specificity with time on task. It has been observed in a variety of repetitive visual tasks, but little is known about these patterns in radiologists. We investigated whether there is systematic variation in performance over the course of a radiology reading session. We re-analyzed data from six previous lesion-enriched radiology studies. Studies featured 8–22 participants assessing 27–100 cases (including mammograms, chest CT, chest x-ray, and bone x-ray) in a reading session. Changes in performance and speed as the reading session progressed were analyzed using mixed effects models. Time taken per case decreased 9–23 % as the reading session progressed (p < 0.005 for every study). There was a sensitivity decrease or specificity increase over the course of reading 100 chest x-rays (p = 0.005), 60 bone fracture x-rays (p = 0.03), and 100 chest CT scans (p < 0.0001). This effect was not found in the shorter mammography sessions with 27 or 50 cases. We found evidence supporting the hypothesis that behavior and performance may change over the course of reading an enriched test set. Further research is required to ascertain whether this effect is present in radiological practice.     

Observer Performance Using Virtual Pathology Slides: Impact of LCD Color Reproduction Accuracy

The use of color LCDs in medical imaging is growing as more clinical specialties use digital images as a resource in diagnosis and treatment decisions. Telemedicine applications such as telepathology, teledermatology, and teleophthalmology rely heavily on color images. However, standard methods for calibrating, characterizing, and profiling color displays do not exist, resulting in inconsistent presentation. To address this, we developed a calibration, characterization, and profiling protocol for color-critical medical imaging applications. Physical characterization of displays calibrated with and without the protocol revealed high color reproduction accuracy with the protocol. The present study assessed the impact of this protocol on observer performance. A set of 250 breast biopsy virtual slide regions of interest (half malignant, half benign) were shown to six pathologists, once using the calibration protocol and once using the same display in its “native” off-the-shelf uncalibrated state. Diagnostic accuracy and time to render a decision were measured. In terms of ROC performance, Az (area under the curve) calibrated = 0.8570 and Az uncalibrated = 0.8488. No statistically significant difference (p = 0.4112) was observed. In terms of interpretation speed, mean calibrated = 4.895 s; mean uncalibrated = 6.304 s which is statistically significant (p = 0.0460). Early results suggest a slight advantage diagnostically for a properly calibrated and color-managed display and a significant potential advantage in terms of improved workflow. Future work should be conducted using different types of color images that may be more dependent on accurate color rendering and a wider range of LCDs with varying characteristics.     

Urban-rural differences in a population-based breast cancer screening program in Croatia

Aim

To investigate urban-rural differences in the distribution of risk factors for breast cancer.

Methods

We analyzed the data from the first round of the “Mamma” population based-screening program conducted in Croatia between 2007 and 2009 and self-reported questionnaire results for 924 patients with histologically verified breast cancer. Reproductive and anthropometric characteristics, family history of breast cancer, history of breast disease, and prior breast screening history were compared between participants from the city of Zagreb (n = 270) and participants from 13 counties with more than 50% of rural inhabitants (n = 654).

Results

The screen-detected breast cancer rate was 4.5 per 1000 mammographies in rural counties and 4.6 in the city of Zagreb, while the participation rate was 61% in rural counties and 59% in Zagreb. Women from Zagreb had significantly more characteristics associated with an increased risk of breast cancer (P < 0.001 in all cases): no pregnancies (15% vs 7%), late age of first pregnancy (≥30 years) (10% vs 4%), and the most recent mammogram conducted 2-3 years ago (32% vs 14%). Women from rural counties were more often obese (41% vs 28%) and had early age of first live birth (<20 years) (20% vs 7%, P < 0.001 for both).

Conclusion

Identification of rural-urban differences in mammography use and their causes at the population level can be useful in designing and implementing interventions targeted at the reduction of inequalities and modifiable risk factors.Significant differences in breast cancer frequency have been identified in different socioeconomic groups, ethnic groups, and between urban and rural populations (1,2). Living in rural areas may be associated with lower access to health care and mammography screening (3), as well as with late-stage diagnosis (4). This often means that patients need to travel great distances to receive care (5). Blair et al found that people in rural and urban areas were diagnosed with breast cancer at similar stages of the disease, although those from rural communities lacked basic cancer information because they did not have access to cancer education programs offered in urban areas (6). Robbins et al explained the higher breast cancer incidence in the San Francisco Bay Area than in other regions by known risk factors: parity, age at first full-term pregnancy, breast-feeding, age at menarche, and age at menopause (7). In Croatia, Polašek et al found that in a period without a national cancer screening program access to health care was the strongest cancer screening utilization predictor in adult rural population (8).Risk factors for breast cancer are mostly those related to the reproductive life of women (9,10): menarche, nulliparity or late age at first birth, late menopause, as well as hormonal factors, be they endogenous or exogenous (eg, term use of oral contraceptives or menopausal hormonal replacement). Other risk factors related to hormonal status include obesity and a diet characterized by a high caloric intake, low intake of fruits and vegetables, and lack of physical activity (11). Radiation, in particular during breast development, was also found to be a risk factor (12), while the role of contaminants, such as xenoestrogens and certain pesticides, remains controversial. Four- to 5-fold risk of developing breast cancer was associated with epithelial proliferative lesions, particularly atypical ductal or lobular hyperplasia (11).In Croatia, breast cancer is the leading cancer among women, amounting to 27% of new female cancer cases; moreover, the incidence rate in 2007 was 17% higher than in the previous year (13). In 2007, cancer incidence by county and age-standardized rates per 100 000 women varied considerably: from 273.1 (Šibensko-kninska county) to 437.7 (the city of Zagreb), but the prevalence of breast cancer risk factors remains unknown. A government-funded mammography screening program was established in October 2006 and has since been implemented in 21 counties, including the city of Zagreb (14).Population-based screening for breast cancer is conducted through mammographic examination of all women of a specified age at prescribed time intervals. The implementation of population-based screening requires technical resources and trained personnel for double reading of mammograms, as well as a major media campaign (15).Within a more extensive study of breast cancer risk factors, this study investigated urban-rural differences in reproductive, anthropometric, and family history of breast cancer and personal history of breast disease among women aged 50-69 from 13 rural counties and the city of Zagreb who participated in the first round of population-based mammography screening in Croatia.     

Automated Detection of Radiology Reports that Document Non-routine Communication of Critical or Significant Results

The purpose of this investigation is to develop an automated method to accurately detect radiology reports that indicate non-routine communication of critical or significant results. Such a classification system would be valuable for performance monitoring and accreditation. Using a database of 2.3 million free-text radiology reports, a rule-based query algorithm was developed after analyzing hundreds of radiology reports that indicated communication of critical or significant results to a healthcare provider. This algorithm consisted of words and phrases used by radiologists to indicate such communications combined with specific handcrafted rules. This algorithm was iteratively refined and retested on hundreds of reports until the precision and recall did not significantly change between iterations. The algorithm was then validated on the entire database of 2.3 million reports, excluding those reports used during the testing and refinement process. Human review was used as the reference standard. The accuracy of this algorithm was determined using precision, recall, and F measure. Confidence intervals were calculated using the adjusted Wald method. The developed algorithm for detecting critical result communication has a precision of 97.0% (95% CI, 93.5–98.8%), recall 98.2% (95% CI, 93.4–100%), and F measure of 97.6% (ß = 1). Our query algorithm is accurate for identifying radiology reports that contain non-routine communication of critical or significant results. This algorithm can be applied to a radiology reports database for quality control purposes and help satisfy accreditation requirements.Key words: Critical results reporting, data mining, Joint Commission on Accreditation of Healthcare Organizations (JCAHO), natural language processing, online analytical processing (OLAP), quality assurance, quality control, radiology reporting     

Mammographic Image Denoising and Enhancement Using the Anscombe Transformation, Adaptive Wiener Filtering, and the Modulation Transfer Function

A new restoration methodology is proposed to enhance mammographic images through the improvement of contrast features and the simultaneous suppression of noise. Denoising is performed in the first step using the Anscombe transformation to convert the signal-dependent quantum noise into an approximately signal-independent Gaussian additive noise. In the Anscombe domain, noise is filtered through an adaptive Wiener filter, whose parameters are obtained by considering local image statistics. In the second step, a filter based on the modulation transfer function of the imaging system in the whole radiation field is applied for image enhancement. This methodology can be used as a preprocessing module for computer-aided detection (CAD) systems to improve the performance of breast cancer screening. A preliminary assessment of the restoration algorithm was performed using synthetic images with different levels of quantum noise. Afterward, we evaluated the effect of the preprocessing on the performance of a previously developed CAD system for clustered microcalcification detection in mammographic images. The results from the synthetic images showed an increase of up to 11.5 dB (p = 0.002) in the peak signal-to-noise ratio. Moreover, the mean structural similarity index increased up to 8.3 % (p < 0.001). Regarding CAD performance, the results suggested that the preprocessing increased the detectability of microcalcifications in mammographic images without increasing the false-positive rates. Receiver operating characteristic analysis revealed an average increase of 14.1 % (p = 0.01) in overall CAD performance when restored image sets were used.     

Comparing Performance of the CADstream and the DynaCAD Breast MRI CAD Systems

Computer-aided diagnosis (CAD) systems are software programs that use algorithms to find patterns associated with breast cancer on breast magnetic resonance imaging (MRI). The most commonly used CAD systems in the USA are CADstream (CS) (Merge Healthcare Inc., Chicago, IL) and DynaCAD for Breast (DC) (Invivo, Gainesville, FL). Our primary objective in this study was to compare the CS and DC breast MRI CAD systems for diagnostic accuracy and postprocessed image quality. Our secondary objective was to compare the evaluation times of radiologists using each system. Three radiologists evaluated 30 biopsy-proven malignant lesions and 29 benign lesions on CS and DC and rated the lesions’ malignancy status using the Breast Imaging Reporting and Data System. Image quality was ranked on a 0–5 scale, and mean reading times were also recorded. CS detected 70 % of the malignant and 32 % of the benign lesions while DC detected 81 % of the malignant lesions and 34 % of the benign lesions. Analysis of the area under the receiver operating characteristic curve revealed that the difference in diagnostic performance was not statistically significant. On image quality scores, CS had significantly higher volume rendering (VR) (p < 0.0001) and motion correction (MC) scores (p < 0.0001). There were no statistically significant differences in the remaining image quality scores. Differences in evaluation times between DC and CS were also not statistically significant. We conclude that both CS and DC perform similarly in aiding detection of breast cancer on MRI. MRI CAD selection will likely be based on other factors, such as user interface and image quality preferences, including MC and VR.     

MRI-Based Breast Volumetry—Evaluation of Three Different Software Solutions

As lipofilling of the female breast is becoming more popular in plastic surgery, the use of MRI to assess breast volume has been employed to control postoperative results. Therefore, we sought to evaluate the accuracy of magnetic resonance imaging (MRI)-based breast volumetry software tools by comparing the measurements of silicone implant augmented breasts with the actual implant volume specified by the manufacturer. MRI-based volume analysis was performed in eight bilaterally augmented patients (46 ± 9 years) with three different software programs (Brainlab© I plan 2.6 neuronavigation software; mass analysis, version 5.3, Medis©; and OsiriX© v.3.0.2. 32-bit). The implant volumes analysed by the BrainLab© software had a mean deviation of 2.2 ± 1.7% (r = 0.99) relative to the implanted prosthesis. OsiriX© software analysis resulted in a mean deviation of 2.8 ± 3.0% (r = 0.99) and the Medis© software had a mean deviation of 3.1 ± 3.0% (r = 0.99). Overall, the volumes of all analysed breast implants correlated very well with the real implant volumes. Processing time was 10 min per breast with each system and 30 s (OsiriX©) to 5 min (BrainLab© and Medis©) per silicone implant. MRI-based volumetry is a powerful tool to calculate both native breast and silicone implant volume in situ. All software solutions performed well and the measurements were close to the actual implant sizes. The use of MRI breast volumetry may be helpful in: (1) planning reconstructive and aesthetic surgery of asymmetric breasts, (2) calculating implant size in patients with missing documentation of a previously implanted device and (3) assessing post-operative results objectively.Key words: MRI, volumetry, mamma, breast, lipofilling, silicone implant, BrainLab, OsiriX, Medis     

Breast Density Analysis Using an Automatic Density Segmentation Algorithm

Breast density is a strong risk factor for breast cancer. In this paper, we present an automated approach for breast density segmentation in mammographic images based on a supervised pixel-based classification and using textural and morphological features. The objective of the paper is not only to show the feasibility of an automatic algorithm for breast density segmentation but also to prove its potential application to the study of breast density evolution in longitudinal studies. The database used here contains three complete screening examinations, acquired 2 years apart, of 130 different patients. The approach was validated by comparing manual expert annotations with automatically obtained estimations. Transversal analysis of the breast density analysis of craniocaudal (CC) and mediolateral oblique (MLO) views of both breasts acquired in the same study showed a correlation coefficient of ρ = 0.96 between the mammographic density percentage for left and right breasts, whereas a comparison of both mammographic views showed a correlation of ρ = 0.95. A longitudinal study of breast density confirmed the trend that dense tissue percentage decreases over time, although we noticed that the decrease in the ratio depends on the initial amount of breast density.     

Mutual information-based template matching scheme for detection of breast masses: from mammography to digital breast tomosynthesis

Development of a computational decision aid for a new medical imaging modality typically is a long and complicated process. It consists of collecting data in the form of images and annotations, development of image processing and pattern recognition algorithms for analysis of the new images and finally testing of the resulting system. Since new imaging modalities are developed more rapidly than ever before, any effort for decreasing the time and cost of this development process could result in maximizing the benefit of the new imaging modality to patients by making the computer aids quickly available to radiologists that interpret the images. In this paper, we make a step in this direction and investigate the possibility of translating the knowledge about the detection problem from one imaging modality to another. Specifically, we present a computer-aided detection (CAD) system for mammographic masses that uses a mutual information-based template matching scheme with intelligently selected templates. We presented principles of template matching with mutual information for mammography before. In this paper, we present an implementation of those principles in a complete computer-aided detection system. The proposed system, through an automatic optimization process, chooses the most useful templates (mammographic regions of interest) using a large database of previously collected and annotated mammograms. Through this process, the knowledge about the task of detecting masses in mammograms is incorporated in the system. Then, we evaluate whether our system developed for screen-film mammograms can be successfully applied not only to other mammograms but also to digital breast tomosynthesis (DBT) reconstructed slices without adding any DBT cases for training. Our rationale is that since mutual information is known to be a robust inter-modality image similarity measure, it has high potential of transferring knowledge between modalities in the context of the mass detection task. Experimental evaluation of the system on mammograms showed competitive performance compared to other mammography CAD systems recently published in the literature. When the system was applied “as-is” to DBT, its performance was notably worse than that for mammograms. However, with a simple additional preprocessing step, the performance of the system reached levels similar to that obtained for mammograms. In conclusion, the presented CAD system not only performed competitively on screen-film mammograms but it also performed robustly on DBT showing that direct transfer of knowledge across breast imaging modalities for mass detection is in fact possible.     

Evaluating the Referring Physician’s Clinical History and Indication as a Means for Communicating Chronic Conditions That Are Pertinent at the Point of Radiologic Interpretation

The clinical history and indication (CHI) provided with a radiological examination are critical components of a quality interpretation by the radiologist. A patient’s chronic conditions offer the context in which acute symptoms and findings can be interpreted more accurately. Seven pertinent (potentially diagnosis altering) chronic conditions, which are fairly prevalent at our institution, were selected. We analyze if and how in 140 CHIs there was mention of a patient’s previously reported chronic condition and if and how the condition was subsequently described in the radiology report using a four-item scheme (Mention/Specialization, Generalization, Common comorbidity, No mention). In 40.7 % of CHIs, the condition was rated Mention/Specialization. Therefore, we reject our first hypothesis that the CHI is a reliable source for obtaining pertinent chronic conditions (≥90.0 %). Non-oncological conditions were significantly more likely rated No mention in the CHI than oncological conditions (58.7 versus 8.3 %, P < 0.0001). Stat cases were significantly more frequently No mention than non-stat cases (60.0 versus 31.3 %, P = 0.0134). We accept our second hypothesis that the condition’s rating in the CHI is significantly correlated with its rating of the final radiology report (χ² test, P < 0.00001). Our study demonstrates an alarming lack of communication of pertinent medical information to the radiologist, which may negatively impact interpretation quality. Presenting automatically aggregated patient information to the radiologist may be a potential avenue for improving interpretation and adding value of the radiology department to the care chain.

Electronic supplementary material

The online version of this article (doi:10.1007/s10278-014-9751-7) contains supplementary material, which is available to authorized users.