A Radiomics Approach for the Classification of Fibroepithelial Lesions on Breast Ultrasonography.

A radiomics-based classifier to distinguish phyllodes tumor and fibroadenoma on gray-scale breast ultrasonography was developed and validated. A total of 93 radiomics features were extracted from representative transverse plane ultrasound images of 182 fibroepithelial lesions initially diagnosed by core needle biopsy. High-throughput radiomics features were selected using the intra-class correlation coefficient between two radiologist readers and the Least Absolute Shrinkage and Selection Operator regression through 10-fold cross-validation. When applied to the validation set, the radiomics classifier for the differentiation of phyllodes tumors and benign/fibroadenomas achieved an area under the receiver operating characteristic curve of 0.765 (95% confidence interval [CI]: 0.597-0.888) with an accuracy of 0.703 (sensitivity: 0.857; specificity: 0.5). Our radiomics signature-based classifier may help predict phyllodes tumors among fibroepithelial lesions on breast ultrasonography.

Trends in the Diagnosis of Phyllodes Tumors and Fibroadenomas Before and After Release of WHO Classification Standards.

BACKGROUND: In 2012, the World Health Organization (WHO) released diagnostic criteria for grading phyllodes tumors based on histologic features. This study sought to examine the application of the WHO criteria and the changing epidemiology of fibroepithelial tumors. METHODS: A retrospective review of surgically excised fibroepithelial lesions from 2007 to 2017 at a single tertiary care institution was conducted. Data regarding the WHO criteria (tumor border, stromal cellularity, stromal cell atypia, stromal overgrowth, mitotic activity) and traditional descriptors (leaf-like architecture, periductal stromal condensation) were collected. Clinical and pathologic characteristics of cases with diagnoses determined before and after 2012 were compared. RESULTS: During the study period, 305 fibroepithelial tumors were identified. No significant differences were observed in terms of mean age, race/ethnicity, presenting symptoms, or method of diagnosis between cases diagnosed before and after 2012. After 2012, the findings showed statistically significant increases in reporting of WHO and traditional histologic features, a decrease in diagnoses of fibroadenomas (85.9% [116/135] before vs 70.0% [119/170] after 2012), and an increase in benign phyllodes tumors (0% [0/135] before vs 12.9% [22/170] after 2012). Patients with a diagnosis of benign phyllodes tumors were significantly younger than those with a diagnosis of borderline, malignant, or non-graded phyllodes tumors (mean age, 25.7 ± 10.6 vs 52.8 ± 9.9, 40.7 ± 24, 46.3 ± 1.5 years, respectively; p = 0.006). CONCLUSIONS: The expanding use of the 2012 WHO criteria has been accompanied by an increased diagnostic frequency of benign phyllodes tumors and a decrease in fibroadenomas. As fibroepithelial diagnoses become more distinct, evidence-based management recommendations for less virulent phyllodes diagnoses should be developed.

Myxoid fibroadenomas differ from conventional fibroadenomas: a hypothesis-generating study.

AIMS: Breast myxoid fibroadenomas (MFAs) are characterized by a distinctive hypocellular myxoid stroma, and occur sporadically or in the context of Carney complex, an inheritable condition caused by PRKAR1A-inactivating germline mutations. Conventional fibroadenomas (FAs) are underpinned by recurrent MED12 mutations in the stromal components of the lesions. The aim of this study was to investigate the genomic landscape of MFAs and compare it with that of conventional FAs. METHODS AND RESULTS: Eleven MFAs from patients without clinical and/or genetic evidence of Carney complex were retrieved. DNA samples of tumour and matching normal tissue were subjected to massively parallel sequencing using the Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT) assay, an assay targeting 410 cancer genes. Genetic alterations detected by MSK-IMPACT were tested in samples in which the stromal and epithelial components were separately laser capture-microdissected. Sequencing revealed no germline PRKAR1A mutations and non-synonymous mutations in six MFAs. Interestingly, in three of the MFAs in which the stromal and epithelial components were separately microdissected, the mutations were found to be restricted to the epithelial rather than the stromal component. The sole exception was a lesion harbouring a somatic truncating PRKAR1A mutation. Upon histological re-review, this case was reclassified as a breast myxoma, consistent with the spectrum of tumous observed in Carney complex patients. In this case, the PRKAR1A somatic mutation was restricted to the stromal component. CONCLUSION: MFAs lack MED12 mutations, and their stromal components seem not to harbour mutations in the 410 cancer genes tested. Whole-exome and/or whole-genome analyses of MFAs are required to elucidate their genetic drivers.

Audit of fibroepithelial tumors of the breast in a Nigerian tertiary institution.

BACKGROUND AND OBJECTIVES: Fibroepithelial lesions are the most common lesions of the breast seen in our laboratory consisting of fibroadenomas (FAs) and phyllodes tumors (PT). The aim of the study was to audit all fibroepithelial lesions and to reclassify all confirmed cases of PTs seen in the study period according to standard criteria. METHODOLOGY: Records and slides of fibroepithelial lesions of the breast received at the department between January 2008 and December 2013 were retrieved and reviewed by the authors. RESULTS: Out of the 1242 fibroepithelial lesions of the breast retrieved, all but 19 were FAs. The 19 were initially reported as PT: 11 benign, 2 borderlines, 2 malignant, and 4 not classified; however, only 16 of these 19, PTs (84%) met the WHO criteria on review. The remaining 3 (16%) turned out to be FAs based on the absence of stroma overgrowth and hypercellularity. The PTs were reclassified into benign PT, borderline PT, and malignant PT accounting for 75% (12/16), 18.7% (3/16), and 6.3% (1/16), respectively. All of the PTs previously not classified turned out benign on review. One of the borderline PTs was originally reported as malignant PT. All cases initially diagnosed as FAs did not change on review. CONCLUSION: These results show that FAs are rarely misdiagnosed. The three cases misdiagnosed as phyllodes may have been prevented if standard data sets were in use. Cases simply referred to as PT without further classification, limit the patients' access to appropriate management as accurate classification helps in the overall management and prognostication.

Identification of lesion subtypes in biopsies of ductal carcinoma in situ of the breast using biomarker ratio imaging microscopy.

Although epidemiological studies propose aggressive and non-aggressive forms of ductal carcinoma in situ (DCIS), they cannot be identified with conventional histopathology. We now report a retrospective study of human biopsy samples using biomarker ratio imaging microscopy (BRIM). Using BRIM, micrographs of biomarkers whose expression correlates with breast cancer aggressiveness are divided by micrographs of biomarkers whose expression negatively correlates with aggressiveness to create computed micrographs reflecting aggressiveness. The biomarker pairs CD44/CD24, N-cadherin/E-cadherin, and CD74/CD59 stratified DCIS samples. BRIM identified subpopulations of DCIS lesions with ratiometric properties resembling either benign fibroadenoma or invasive carcinoma samples. Our work confirms the existence of distinct subpopulations of DCIS lesions, which will likely have utility in breast cancer research and clinical practice.

Can problematic fibroepithelial lesions be accurately classified on core needle biopsies?

Fibroepithelial lesions (FEL) of the breast are notoriously difficult to classify on core needle biopsies. The goal of this study was to evaluate interobserver variability and accuracy of subclassifying difficult FELs into fibroadenoma (FA) and phyllodes tumors (PTs). We identified 50 breast core needle biopsies, initially diagnosed generically as FEL, with subsequent excision and final diagnosis of either FA or benign PT. Five surgical pathologists from one institution independently reviewed these in 3 rounds. The pathologists were blinded to the final excisional diagnosis. Two diagnostic categories were allowed: FA and PT. A set of histologic criteria was provided including the presence of subepithelial condensation, stromal heterogeneity, overgrowth, pleomorphism, fragmentation, cellularity, adipose tissue entrapment, and mitotic count and asked to review the slides for the second round. A third round of interpretations was conducted after each criterion was defined. Interobserver agreement for the diagnosis and each criterion was evaluated using the κ level of agreement. Accuracy of ratings to final diagnosis was calculated using Wilcoxon signed-rank test. κ Values for interobserver agreement were fair for the first and second rounds varying from 0.20 to 0.22, respectively. This increased to 0.27 in round 3. When considering each category, the κ value varied from 0.26 to 0.29 for FA and 0.28 to 0.14 for PT. Overall, there was fair agreement between the pathologists in all categories. The rate of correctly diagnosed cases ranged from 40% in the first round, to 48% in the second round, to 67% in round 3. Overall the pathologists performed better in identifying FA than PT. The accuracy of interpretations was significantly different between the first (40%), second (48%), and third rounds (67%).

Differentiation of Breast Lesions by Use of HyperSPACE: Hyper-Spectral Analysis for Characterization in Echography.

Early diagnosis represents the cornerstone in breast cancer control. Ultrasound is still a valid tool because of its low invasiveness, reduced costs and reduced risk of harm, but better exploitation of its potential is necessary to extract information on tissue features. The proposed method, HyperSPACE (hyper-spectral analysis for characterization in echography), which processes the ultrasonic radiofrequency signal in an N-dimension spectral hyperspace to define several characteristic parameters of the tissue under investigation, was used with the aim of differentiating two types of breast lesion: infiltrating ductal carcinoma and fibroadenoma. The analyzed data set consisted of 2000 radiofrequency frames related to 200 sections of pathologic breast nodules: 104 infiltrating ductal carcinomas and 96 fibroadenomas. The algorithm was trained on single radiofrequency frames related to 50 sections (26 carcinomas, 24 fibroadenomas) to recognize the two pathologies considered, and all the radiofrequency frames related to the other 150 sections were classified, yielding a sensitivity of 92.2%, specificity of 93%, positive predictive value of 93.2% and negative predictive value of 91%. The results were compared with those of RULES (radiofrequency ultrasonic local estimators), a processing method set developed by our group and used by other researchers in clinical and laboratory environments.

Breast Imaging-Reporting and Data System (BI-RADS) classification in 51 excised palpable pediatric breast masses.

INTRODUCTION: The American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) classification was developed to risk stratify breast lesions and guide surgical management based on imaging. Previous studies validating BI-RADS for US do not include pediatric patients. Most pediatric breast masses present as palpable lesions and frequently undergo ultrasound, which is often accompanied with a BI-RADS classification. Our study aimed to correlate BI-RADS with pathology findings to assess applicability of the classification system to pediatric patients. METHODS: We performed a retrospective review of all patients who underwent excision of a breast mass at a single center from July 2010 to November 2013. We identified all patients who underwent preoperative ultrasound with BI-RADS classification. Demographic data, imaging results, and surgical pathology were analyzed and correlated. RESULTS: A total of 119 palpable masses were excised from 105 pediatric patients during the study period. Of 119 masses, 81 had preoperative ultrasound, and BI-RADS categories were given to 51 masses. Of these 51, all patients were female and the average age was 15.9 years. BI-RADS 4 was given to 25 of 51 masses (49%), and 100% of these lesions had benign pathology, the most common being fibroadenoma. CONCLUSIONS: Treatment algorithm based on BI-RADS classification may not be valid in pediatric patients. In this study, all patients with a BI-RADS 4 lesion had benign pathology. BI-RADS classification may overstate the risk of malignancy or need for biopsy in this population. Further validation of BI-RADS classification with large scale studies is needed in pediatric and adolescent patients.

Gene expression-based classifications of fibroadenomas and phyllodes tumours of the breast.

Fibroepithelial tumors (FTs) of the breast are a heterogeneous group of lesions ranging from fibroadenomas (FAD) to phyllodes tumors (PT) (benign, borderline, malignant). Further understanding of their molecular features and classification might be of clinical value. In this study, we analysed the expression of 105 breast cancer-related genes, including the 50 genes of the PAM50 intrinsic subtype predictor and 12 genes of the Claudin-low subtype predictor, in a panel of 75 FTs (34 FADs, 5 juvenile FADs, 20 benign PTs, 5 borderline PTs and 11 malignant PTs) with clinical follow-up. In addition, we compared the expression profiles of FTs with those of 14 normal breast tissues and 49 primary invasive ductal carcinomas (IDCs). Our results revealed that the levels of expression of all breast cancer-related genes can discriminate the various groups of FTs, together with normal breast tissues and IDCs (False Discovery Rate < 5%). Among FTs, the levels expression of proliferation-related genes (e.g. CCNB1 and MKI67) and mesenchymal/epithelial-related (e.g. CLDN3 and EPCAM) genes were found to be most discriminative. As expected, FADs showed the highest and lowest expression of epithelial- and proliferation-related genes, respectively, whereas malignant PTs showed the opposite expression pattern. Interestingly, the overall profile of benign PTs was found more similar to FADs and normal breast tissues than the rest of tumours, including juvenile FADs. Within the dataset of IDCs and normal breast tissues, the vast majority of FADs, juvenile FADs, benign PTs and borderline PTs were identified as Normal-like by intrinsic breast cancer subtyping, whereas 7 (63.6%) and 3 (27.3%) malignant PTs were identified as Claudin-low and Basal-like, respectively. Finally, we observed that the previously described PAM50 risk of relapse prognostic score better predicted outcome in FTs than the morphological classification, even within PTs-only. Our results suggest that classification of FTs using gene expression-based data is feasible and might provide clinically useful biological and prognostic information.

Simple and complex fibroadenomas: are there any distinguishing sonographic features?

OBJECTIVES: Complex fibroadenomas are fibroadenomas harboring 1 or more complex pathologic features: epithelial calcifications, apocrine metaplasia, sclerosing adenosis, and cysts larger than 3 mm. No sonographic features have been clearly defined for the distinction of simple fibroadenomas from complex ones, which are associated with an increased cancer risk. We aimed to evaluate the accuracy of sonographic features for the prediction of complexity in fibroadenomas. METHODS: A total of 252 fibroadenomas were found at consecutive percutaneous needle or excisional surgical biopsy. Sixty-three were excluded because their respective imaging examinations were not recorded on digital support and consequently were not available. According to histologic diagnoses, fibroadenomas (n = 189) were classified into simple (n = 159) and complex (n = 30). The size, number, and sonographic features were assessed, and their respective accuracy rates for prediction of complexity were analyzed. RESULTS: All patients were women. Complex fibroadenomas presented more frequently as solitary nodules (n = 21 [70%]) and were significantly larger than simple fibroadenomas (1.9 versus 1.3 cm; P = .009). Image predictors of complexity were an irregular shape (P< .001), noncircumscribed contours (indistinct, angular, microlobulated, or spiculated; P < .001), a complex echo structure (P < .001), the presence of microcalcifications (P = .002), and posterior acoustic enhancement (P < .001). By logistic multivariate regression, a complex echo structure (odds ratio [OR], 9.5; 95% confidence interval [CI], 2.8-32.3), noncircumscribed contours (OR, 3.7; 95% CI, 1.1-12.8), and posterior acoustic enhancement (OR, 4.0; 95% CI, 1.1-14.6) were independent predictors of complexity. Receiver operating characteristic curve analysis showed that a complex echo structure was the most accurate sonographic finding for identification of complex fibroadenomas (area under the curve, 0.74). CONCLUSIONS: Certain sonographic features are associated with complex fibroadenomas and can help the radiologist decide which ones require biopsy.

Utility of Ki-67, CD10, CD34, p53, CD117, and mast cell content in the differential diagnosis of cellular fibroadenomas and in the classification of phyllodes tumors of the breast.

Adequate management of phyllodes tumors of the breast (PTB) remains a challenge because of the difficulty in correctly establishing preoperative diagnosis. The aim of this study was to evaluate the usefulness of Ki-67, CD10, CD34, p53, CD117, and of the number of mast cells in the differential diagnosis of benign PTB and cellular fibroadenomas (CFs) as well as in the grading of PTB. Fifty-one primary PTB and 14 CFs were examined by immunohistochemistry.When evaluating CD117 expression, higher epithelial expression was present in CF as well as an increased number of mast cells in benign PTB. Stromal expression of Ki-67, CD10, CD34, and p53 were relevant to PTB grading, of which the first 3 showed significance in the distinction of benign and borderline PTB, as well as between benign and malignant PTB. P53 was relevant only for the discrimination between benign and malign PTB. None of the markers showed significance in distinguishing between borderline and malign PTB.

Computer-aided diagnosis of breast masses using quantified BI-RADS findings.

The information from radiologists was utilized in the proposed computer-aided diagnosis (CAD) for breast tumor classification. The ultrasound (US) database used in this study contained 166 benign and 78 malignant masses. For each mass, six quantitative feature sets were used to describe the radiologists' grading of six Breast Imaging Reporting and Data System (BI-RADS) categories including shape, orientation, margins, lesion boundary, echo pattern, and posterior acoustic features on breast US. The descriptive abilities were between 76% and 82% and the predicted descriptors were then used for tumor classification. Using receiver operating characteristic curve for evaluation, the area under curve (AUC) of the proposed CAD was slightly better than that of a conventional CAD based on the combination of all quantitative features (0.96 vs. 0.93, p=0.18). The partial AUC over 90% sensitivity of the proposed CAD was significantly better than that of the conventional CAD (0.90 vs. 0.76, p<0.05). In conclusion, the computer-aided analysis with qualitative information from radiologists showed a promising result for breast tumor classification.

Myoepithelial and epithelial-myoepithelial, mesenchymal and fibroepithelial breast lesions: updates from the WHO Classification of Tumours of the Breast 2012.

In the 4th edition of the WHO Classification of Tumours of the Breast, myoepithelial lesions are retitled myoepithelial and epithelial-myoepithelial lesions in order to better reflect the dual participation of luminal and myoepithelial compartments in some key entities. Malignant myoepithelioma, described as a section within the chapter on myoepithelial lesions in the 3rd edition, is recognised in the 4th edition as part of metaplastic carcinoma. Adenomyoepithelioma with malignancy is categorised in terms of the cellular component undergoing malignant transformation. The list of antibodies that can be used for identifying myoepithelial cells is updated. Among mesenchymal lesions, new additions are nodular fasciitis and atypical vascular lesions, while the haemangiopericytoma is removed. The 3rd edition stated that pathological prediction of behaviour of phyllodes tumours is difficult in the individual case. In the 4th edition, some progress has been made in prioritisation and weighting of histological parameters that can potentially estimate probability of recurrence. The WHO Working Group advocates leaning towards a diagnosis of fibroadenoma in cases where there is histological uncertainty in distinction from a benign phyllodes tumour, or adopting the neutral term 'benign fibroepithelial neoplasm', as the clinical behaviour of fibroadenoma overlaps with that of benign phyllodes tumour. The 3rd edition terminology of 'periductal stromal sarcoma' is revised to 'periductal stromal tumour', akin to the widespread consensus to avoid the use of the term 'cystosarcoma' in the context of phyllodes tumours.

Progranulin expression in breast cancer with different intrinsic subtypes.

Progranulin is a newly discovered 88-kDa glycoprotein originally purified from the highly tumorigenic mouse teratoma-derived cell line PC. We found that high progranulin expression was associated with higher breast carcinoma angiogenesis, reflected by increased vascular endothelial growth factor expression and higher microvessel density. However, no immunohistochemical evidence currently exists to correlate progranulin expression with clinicopathological features in different intrinsic subtypes of breast carcinoma biopsies. The aim of this study was to investigate the progranulin expression profiles in the intrinsic subtypes of breast carcinomas and their relevance to histopathological and clinicopathological features. Tissue blocks containing 264 cases of breast carcinomas from 2006 to 2009 were classified as different intrinsic subtypes. Tissues of four intrinsic subtypes were immunostained for progranulin, vascular endothelial growth factor and CD105. Their relevance to histopathological and clinicopathological features was also analyzed. Twenty tissue samples from breast fibroadenomas were included in this study. Progranulin expression showed no significant differences in different intrinsic subtypes, although an increasing tendency could be found in the triple-negative breast cancer (TNBC) subgroup (χ(2)=5.00, df=3, p=0.17). However, differences were significant when pathologically node metastasis-positive (pN(+)) TNBC were excluded (χ(2)=17.84, df=3, p<0.01). Some clinicopathological parameters, including CK5/6 (χ(2)=0.08, df=3, p=0.78), E-cadherin (χ(2)=0.71, df=3, p=0.40) and P53 (χ(2)=0.05, df=3, p=0.83), displayed no correlation with activity of progranulin in pathologically node metastasis-negative (pN(-)) TNBC. It was noted that the EGFR expression level of the pN(-) TNBC subtype was significantly higher in cases with strong progranulin expression than in cases with weak progranulin expression (χ(2)=11.26, df=1, p<0.01). A significantly higher expression level of progranulin in pN(-) TNBC suggests that progranulin is a promising new target for pN(-) TNBC treatment. Strong expression of progranulin correlates with positive EGFR expression in the pN(-) TNBC subtype. The close relationship between EGFR and progranulin/VEGF/CD105 expression may partly play a role in high angiogenesis levels in the pN(-) TNBC subtype.

Axial-shear strain elastography for breast lesion classification: further results from in vivo data.

The purpose of this work was to investigate the potential of the normalized axial-shear strain area (NASSA) feature, derived from axial-shear strain elastograms (ASSE), for breast lesion classification of fibroadenoma and cancer. This study consisted of previously acquired in vivo digital radiofrequency data of breast lesions. A total of 33 biopsy-proven malignant tumors and 30 fibroadenoma cases were included in the study, which involved three observers blinded to the original BIRADS-ultrasound scores. The observers outlined the lesions on the sonograms. The ASSEs were segmented and color-overlaid on the sonograms, and the NASSA feature from the ASSE was computed semi-automatically. Receiver operating characteristic (ROC) curves were then generated and the area under the curve (AUC) was calculated for each observer performance. A logistic regression classifier was built to compare the improvement in the AUC when using BIRADS scores plus NASSA values as opposed to BIRADS scores alone. BIRADS score ROC had an AUC of 0.89 (95% CI = 0.81 to 0.97). In comparison, the average of the AUC for all the three observers using ASSE feature alone was 0.84. However, the AUC increased to 0.94 (average of 3 observers) when BIRADS score and ASSE feature were combined. The results demonstrate that the NASSA feature derived from ASSE has the potential to improve BIRADS breast lesion classification of fibroadenoma and malignant tumors.

Influence of age on PPV of sonographic BI-RADS categories 3, 4, and 5.

PURPOSE: The purpose of this retrospective study was to calculate the positive predictive value (PPV) of sonographic Breast Imaging Reporting and Data System (BI-RADS) categories 3, 4, and 5 in different age groups to investigate whether age influences the PPV of the BI-RADS category in breast ultrasound. MATERIALS AND METHODS: From our sonography-guided core biopsy database of breasts between 2006 and 2008, we identified 2817 BI-RADS category 3, 4, and 5 lesions with known pathological diagnosis in 2587 women, all of whom underwent the earlier breast assessment via ultrasound with a sonographic BI-RADS lexicon and later sonography-guided core biopsy. All lesions were classified into three age groups (< 45, 45 - 59, and > 59 years). The age-related PPVs of each BI-RADS category among three age groups were calculated on the basis of pathological diagnoses and were compared using a χ(2)-test. RESULTS: The overall PPV of each BI-RADS category was 2.2 % in category 3, 6.5 % in category 4a, 35.2 % in category 4b, 79.6 % in category 4c, and 99.6 % in category 5. The age-related PPVs of category 3 varied significantly among the three age groups (0.9 % versus 3.9 % versus 2.0 % p = 0.048), and notably, the age-related PPV in group 2 was higher than the others. Additionally, there was a significant positive association between the age-related PPVs and increasing age in categories 4a and 4b (4a, p < 0.0001 and 4b, p = 0.0139), but not in categories 4c and 5 (4c, p = 0.1853 and 5, p = 0.2871). CONCLUSION: The incidence of female breast cancer differs not only in different sonographic BI-RADS categories, but also in different age groups. Therefore, more attention should be paid to the special age group that we found for sonographic BI-RADS categories 3, 4a, and 4b.

Multicenter study of ultrasound real-time tissue elastography in 779 cases for the assessment of breast lesions: improved diagnostic performance by combining the BI-RADS®-US classification system with sonoelastography.

PURPOSE: Hitachi real-time tissue elastography (HI-RTE) is an ultrasound technique that facilitates the estimation of tissue elasticity. Our study evaluates whether sonoelastography improves the differentiation of benign and malignant breast lesions. MATERIALS AND METHODS: In a multicenter approach sonoelastography of focal breast lesions was carried out in 779 patients with subsequent histological confirmation. We present data from 3 study centers (Berlin, Bielefeld, Homburg/Saar) focusing on the sensitivity (SE), specificity (SP) and the positive (PPV) and negative predictive value (NPV) of sonoelastography. In addition we performed an analysis of the diagnostic performance, expressed by the pretest and posttest probability of disease (POD), in BI-RADS®-US 3 or 4 lesions as these categories can imply both malignant and benign lesions and a more precise prediction would be a preferable aim. RESULTS: Sonoelastography demonstrated an improved SP (89.5 %) and an excellent PPV (86.8 %) compared to B-mode ultrasound (76.1 % and 77.2 %). Especially in dense breasts ACR III-IV, the SP was even higher (92.8 %). In BI-RADS-US 3 lesions, a suspicious elastogram significantly modified the POD from 8.3 % to a posttest POD of 45.5 %. In BI-RADS-US 4 lesions, we found a pretest POD of 56.6 %. The posttest POD changed significantly to 24.2 % with a normal elastogram and to 81.5 % with a suspicious elastogram. CONCLUSION: Our data demonstrates that the complementary use of sonoelastography definitely improves the performance in breast diagnostics. Finally we present a protocol of how sonoelastography can be integrated into our daily practice.

Discrimination of breast tumors in ultrasonic images using an ensemble classifier based on the AdaBoost algorithm with feature selection.

This paper proposes a novel algorithm to estimate a log-compressed K distribution parameter and presents an algorithm to discriminate breast tumors in ultrasonic images. We computed a total of 208 features for discrimination, including those based on a parameter of a log-compressed K-distribution, which quantifies the homogeneity of the echo pattern in the tumor, but is influenced by compression parameters in the ultrasonic device. The proposed algorithm estimates the parameter of the log-compressed K-distribution in a manner free from this influence. To quantify irregularities in tumor shape, pattern-spectrum-based features were newly developed in this paper. The discrimination process uses an ensemble classifier trained by a multiclass AdaBoost learning algorithm (AdaBoost.M2), combined with a sequential feature-selection process. A 10-fold cross-validation test validated the performance, and the results were compared with those of a Mahalanobis distance-based classifier and a multiclass support vector machine. A total of 200 carcinomas, 50 fibroadenomas, and 50 cysts were used in the experiments. This paper demonstrates that the combination of a classifier trained by AdaBoost.M2 and features based on the estimated parameter of a log-compressed K-distribution, as well as those of the pattern spectrum, are useful for the discrimination of tumors.

Automated breast ultrasound: lesion detection and BI-RADS classification--a pilot study.

PURPOSE: Evaluation of an automated breast ultrasound system (ABUS) regarding the detection and classification of breast lesions according to BI-RADS. MATERIALS AND METHODS: Women were selected for the study who had unclear findings in breast diagnosis performed elsewhere (palpation, sonography or mammography) and who were referred for further work-up. All patients received a hand-held ultrasonography (HHUS) with a 13 MHz transducer, clinical examination and mammography of both breasts. Additionally, the affected breast received the ABUS (SomoVuTM, U-Systems, Inc., San Jose, CA, USA; EC Representative: Siemens, Erlangen, Germany) which was performed with an 8 MHz transducer. Five radiologists independently evaluated the ABUS images regarding lesion detectability. All detected lesions were classified according to BI-RADS assessment. The examiners had no knowledge of the patients' clinical examination or of the result of the mammography or the HHUS. Results of the ABUS were compared to HHUS. RESULTS: 35 women were included in the study. 25 BI-RADS 4 or 5 lesions had further histological (n = 23) or cytological (n = 2) work-up which revealed 13 malignant and 12 benign findings. The size of all lesions ranged from 6 to 32 mm (median 14 mm). With the ABUS all examiners detected 29 to 30 lesions while HHUS revealed 30 lesions. One suspicious area in HHUS was not reported by any of the five examiners with the ABUS. Histology of this area revealed mastopathic disease. No benign lesion was classified as BI-RADS 5 with the ABUS or HHUS. All breast cancers were found with the ABUS by all examiners and correctly classified as BI-RADS 4 or 5. There was good agreement regarding BI-RADS classification of HHUS and ABUS for the five different examiners with Kappa values between 0.83 and 0.87. CONCLUSION: These preliminary results show that the ABUS allows detection of solid and cystic lesions and their BI-RADS classification with a high reliability in a selected patient group.

Computer aided classification system for breast ultrasound based on Breast Imaging Reporting and Data System (BI-RADS).

Clinically, the ultrasound findings are evaluated by its sonographic characteristics and then assigned to assessment categories according to the definitions of Breast Imaging Reporting and Data System (BI-RADS) developed by the American College of Radiology. In this study, a computer-aided classification (CAC) system was proposed to classify the masses into assessment categories 3, 4 and 5, which simulated the clinical diagnosis of radiologists. Compared with current computer-aided diagnosis systems, the proposed CAC system classifies the indeterminate cases into BI-RADS category 4 for further diagnosis. Six hundred twenty-six cases were collected from three ultrasound systems and confirmed by pathology and retrospectively classified into categories 3, 4 and 5 by radiologists. The multinomial logistic regression model was trained as the CAC system for predicting the assessment category from the computerized BI-RADS features and from a set of machine-dependent factors. By using the machine-dependent factors to indicate the adopted ultrasound systems, the same regression model could be applied for the cases acquired from different ultrasound systems. A basic CAC system was trained by using the classification result of radiologists. A weighted CAC system, to improve the capacity of the basic CAC system in differentiating benign from malignant lesions, was trained by adding the pathologic result. Between the radiologists and the basic CAC system, a substantial agreement was indicated by Cohen's kappa statistic and the differences in either the performance indices or the A(Z) of receiver operating characteristic (ROC) analysis were not statistically significant. For the weighted CAC system, the performance indices accuracy, sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were 73.00% (457 of 626), 98.17% (215 of 219), 59.46% (242 of 407), 56.58% (215 of 380) and 98.37% (242 of 246), respectively; the A(Z) was 0.94; and the correlation with the radiologists was also substantial agreement. The indices accuracy and specificity of weighted CAC system, compared with those of the radiologists, were improved by 5.91% and 8.85%, respectively and the indices of sensitivity and NPV, compared with those of a conventional CAD system, were improved by 10.5% and 5.21%, respectively; all improvements were statistically significant. To classify the mass into BI-RADS assessment categories by the CAC system is feasible. Moreover, the proposed CAC system is flexible because it can be used to diagnose the cases acquired from different ultrasound systems.