Performance of machine learning software to classify breast lesions using BI-RADS radiomic features on ultrasound images.

BACKGROUND: The purpose of this work was to evaluate computable Breast Imaging Reporting and Data System (BI-RADS) radiomic features to classify breast masses on ultrasound B-mode images. METHODS: The database consisted of 206 consecutive lesions (144 benign and 62 malignant) proved by percutaneous biopsy in a prospective study approved by the local ethical committee. A radiologist manually delineated the contour of the lesions on greyscale images. We extracted the main ten radiomic features based on the BI-RADS lexicon and classified the lesions as benign or malignant using a bottom-up approach for five machine learning (ML) methods: multilayer perceptron (MLP), decision tree (DT), linear discriminant analysis (LDA), random forest (RF), and support vector machine (SVM). We performed a 10-fold cross validation for training and testing of all classifiers. Receiver operating characteristic (ROC) analysis was used for providing the area under the curve with 95% confidence intervals (CI). RESULTS: The classifier with the highest AUC at ROC analysis was SVM (AUC = 0.840, 95% CI 0.6667-0.9762), with 71.4% sensitivity (95% CI 0.6479-0.8616) and 76.9% specificity (95% CI 0.6148-0.8228). The best AUC for each method was 0.744 (95% CI 0.677-0.774) for DT, 0.818 (95% CI 0.6667-0.9444) for LDA, 0.811 (95% CI 0.710-0.892) for RF, and 0.806 (95% CI 0.677-0.839) for MLP. Lesion margin and orientation were the optimal features for all the machine learning methods. CONCLUSIONS: ML can aid the distinction between benign and malignant breast lesion on ultrasound images using quantified BI-RADS descriptors. SVM provided the highest ROC-AUC (0.840).

Calcificaciones mamarias: descripción y clasificación según la 5ª edición BI-RADS / Breast calcifications: description and classification according to BI-RADS 5th edition

Abstract. Breast calcifications are frequent findings in mammography. Most of them have a benign origin, such as in the case of the response to inflammatory disease of the ducts or coarse calcifications in benign nodules. Many of these calcifications show a characteristic benign appearance, and they do not need to be magnified or monitored. However, other calcifications can show a grouped pattern, have a suspicious appearance, and transform into an in situ ductal carcinoma or a high risk breast lesion. It is important to know the morphological and distribution patterns of these calcifications in order to make right decisions for each case. In the 5th edition of the BI-RADS atlas, 2013, categories and levels of suspicion for some patterns were modified. The objective of this article is to update descriptors and categories of BI-RADS micro-calcifications, pointing out their most important features and malignancy risk linked to each descriptor.

Resumen. Las calcificaciones mamarias son un hallazgo frecuente en mamografía. La mayoría de ellas tienen un origen benigno, como puede ser la respuesta a patología inflamatoria de los conductos o calcificaciones gruesas en nódulos benignos. Muchas de estas calcificaciones presentan un aspecto benigno característico y no requieren ser magnificadas o controladas. Otras calcificaciones sin embargo pueden presentarse agrupadas, tener un aspecto sospechoso y originarse en un carcinoma ductal in situ o una lesión de alto riesgo. Es relevante conocer los patrones morfológicos y de distribución de estas calcificaciones a fin de tomar la conducta adecuada para cada caso. En la 5.ª edición del atlas BI-RADS, 2013, las categorías y grados de sospecha de algunos patrones fueron modificados. El objetivo del presente artículo es realizar una actualización de los descriptores y las categorías BI-RADS de las microcalcificaciones, señalando sus características más importantes y el riesgo de malignidad asociado a cada descriptor.

The importance of breast elastography added to the BI-RADS® (5th edition) lexicon classification.

OBJECTIVE: the aim of this study was to investigate the addition of elastography to the BI-RADS® lexicon for the classification of breast lesions. METHODS: a total of 955 consecutive patients who were subjected to breast percutaneous biopsy from January 2010 to December 2012 were retrospectively assessed. Overall, 26 patients who did not present with masses on conventional ultrasound were excluded. The patients were classified according to the fifth edition of the breast imaging and reporting data system (BI-RADS®) lexicon, which includes elastographic findings. The BI-RADS®classification is based on the same classification principles that have been suggested by the author, which classify lesions as soft, intermediate, or hard. RESULTS: the addition of elastographic findings to the BI-RADS® lexicon improved the sensitivity (S), specificity (SP), and diagnostic accuracy (DA) of ultrasound in the assessment of breast lesions, which increased from 93.85, 72.07, and 76.64 to 95.90, 80.65, and 91.39%, respectively. CONCLUSION: these findings suggest that the addition of elastography to the BIRADS ® lexicon will improve the SP and DA of ultrasound in the screening of breast lesions.

Adequacy of Fine-Needle Aspiration Cytology for Breast Lesions: The SurePath® Liquid-Based Technique versus Conventional Smears.

OBJECTIVE: We aimed to assess the cytological findings from fine-needle aspiration (FNA) of breast lesions with conventional smear (CS) and SurePath® (SP) cytology. STUDY DESIGN: This was a cross-sectional study of women who underwent FNA from January 2012 to June 2013 for breast lesions with benign ultrasonography impressions in Fortaleza, Brazil. Two groups were formed. The first was composed of 102 samples subjected to CS and the second of 65 samples subjected to SP. The number of smears, the cellularity and the diagnostic hypotheses were compared. Fisher's exact tests with 95% confidence intervals were applied. RESULTS: The women ranged in age from 22 to 75 years. SP cytology indicated greater cellularity than CS for the cystic lesions (p < 0.05). In the CS group, 72.5% of the samples required 3-4 slides, but in the SP group, only 9.2% required a second slide. The cellularity of the cystic samples was significantly greater with the SP method (p < 0.01). CONCLUSION: In FNA cytology of the breast, SP is a tool that is comparable to CS, but with the added benefits of better cellularity results for cystic lesions and requiring fewer slides for analysis.

The importance of breast elastography added to the BI-RADS® (5th edition) lexicon classification / A importância da elastografia mamária adicionada à classificação, segundo o léxico BI-RADS® (5ª edição)

SummaryObjective:the aim of this study was to investigate the addition of elastography to the BI-RADS® lexicon for the classification of breast lesions.Methods:a total of 955 consecutive patients who were subjected to breast percutaneous biopsy from January 2010 to December 2012 were retrospectively assessed. Overall, 26 patients who did not present with masses on conventional ultrasound were excluded. The patients were classified according to the fifth edition of the breast imaging and reporting data system (BI-RADS®) lexicon, which includes elastographic findings. The BI-RADS®classification is based on the same classification principles that have been suggested by the author, which classify lesions as soft, intermediate, or hard.Results:the addition of elastographic findings to the BI-RADS® lexicon improved the sensitivity (S), specificity (SP), and diagnostic accuracy (DA) of ultrasound in the assessment of breast lesions, which increased from 93.85, 72.07, and 76.64 to 95.90, 80.65, and 91.39%, respectively.Conclusion:these findings suggest that the addition of elastography to the BIRADS ® lexicon will improve the SP and DA of ultrasound in the screening of breast lesions.

ResumoObjetivo:investigar o impacto da adição da descoberta da elastografia das lesões mamárias à classificação segundo o léxico BI-RADS®.Métodos:estudo retrospectivo com 955 pacientes consecutivas, submetidas à biópsia mamária percutânea no período de janeiro de 2010 a dezembro de 2012. Foram excluídas 26 pacientes que apresentaram lesão não nodular ao ultrassom convencional. As lesões foram classificadas conforme proposta da 5ª edição do léxico BI-RADS®, que inclui os achados de elastografia. A classificação BI- -RADS® é baseada nos mesmos critérios propostos pelo autor, que classifica as lesões como macias, intermediárias e rígidas.Resultados:a adição dos achados da elastografia ao léxico BI-RADS® melhorou a sensibilidade (S), a especificidade (E) e a acurácia diagnóstica (AD) do ultrassom na avaliação das lesões mamárias, de 93.85, 72.07 e 76.64% para 95.90, 80.65 e 91.39%, respectivamente.Conclusão:os achados sugerem que a adição dos achados da elastografia ao léxico BI-RADS® pode melhorar a S, a E e a AD do ultrassom no rastreamento de lesões mamárias.

Processos inflamatórios da mama: caracterização e manejo terapêutico / Inflammatory processes of breast: characterization and therapeutic management

Hiperemia, calor e dor na mama podem relacionar-se a um grande número de causas. Processos inflamatórios da mama geralmente estão relacionados ao ciclo grávido-puerperal e suas modificaçöes, principalmente as formas agudas. Este trabalho tem por objetivo a caracterização dos diferentes processos inflamatórios da mama e seu adequado manejo.

Redness, heat, and breast pain may be related to a number of causes. Inflammatory conditions of the breast are usually related to the pregnancy-puerperal cycle and its modifications, mainly acute forms. This work aims to characterize the different inflammatory breast conditions and its adequate management.

New elastographic classification of breast lesions during and after compression.

PURPOSE: Proposal for the classification of breast masses through ultrasound elastography in order to differentiate benign and malignant lesions with histological correlation. MATERIALS AND METHODS: 188 patients enrolled for percutaneous biopsy of 228 breast lesions. Elastography was performed and interpreted according to criteria created by the authors, with scores varying from 1 to 4 based on elasticity of images obtained upon release of compression. These results were compared with the histological results; elasticity scores of 1 and 2 were considered benign, a score of 3 as probably benign, and 4 as suspicious for malignancy. Positive predictive value, specificity, and diagnostic accuracy have been calculated. The results were evaluated using Fisher's exact test and the analysis of the receiver operating characteristic (ROC) curve to determine the association with the histological results, and diagnostic accuracy of the proposed classification. RESULTS: The positive predictive value, specificity, and diagnostic accuracy of the scores were 76.5%, 95.9%, and 94.7%, respectively. Of 228 lesions tested, 26 tests yielded true positive results; 8 yielded false positive results; 190 true negative results; and 4 false negative results. There was association with the histological results by the Fisher method (P < 0.05) and an excellent area below the ROC curve of 0.954 (confidence range of 95%, 0.925-0.982). CONCLUSION: The classification by elastography proposed by the authors can be used as an important tool combined with ultrasonographic studies for differentiating benign and malignant lesions of the breast.

Avaliação de microcalcificações mamárias de acordo com as classificações do Breast Imaging Reporting and Data System (BI-RADS TM) e de Le Gal / Evaluation of breast microcalcifications according to Breast Imaging Reporting and Data System (BI-RADS TM) and Le Gal's classifications

OBJETIVO: avaliar a acurácia da mamografia para o diagnóstico de microcalcificações mamárias suspeitas, com as classificações do Breast Imaging Reporting and Data System (BI-RADS TM) e Le Gal em comparação com o resultado histopatológico utilizado como padrão-ouro. MÉTODOS: foram selecionados dos arquivos dos blocos cirúrgicos, 130 casos operados com mamografias contendo somente microcalcificações mamárias, inicialmente classificadas como suspeitas sem lesões detectáveis ao exame clínico. Estas foram reclassificadas por dois examinadores, utilizando as classificações de Le Gal e BI-RADS TM, obtendo-se diagnóstico de consenso. As biópsias foram revistas por dois patologistas e foi obtido diagnóstico de consenso. A leitura das mamografias e a revisão das lâminas foram feitas em duplo-cego. As análises estatísticas utilizadas neste estudo foram o teste do chi2, o modelo Fleiss quadrático para VPP e o programa Epi-Info 6.0. RESULTADOS: a correlação entre a análise histopatológica e mamográfica, usando BI-RADS TM e Le Gal, mostrou a mesma sensibilidade de 96,4 por cento, especificidade de 55,9 e 30,3 por cento, valor preditivo positivo (VPP) de 37,5 por cento e 27,5 por cento e acurácia de 64,6 e 44,6 por cento, respectivamente. Quando discriminamos por categorias de BI-RADS TM, obtivemos VPPs: categoria 2, 0 por cento; categoria 3, 1,8 por cento; categoria 4, 31,6 por cento e categoria 5, 60 por cento. Os VPPs pela classificação de Le Gal foram: categoria 2, 3,1 por cento; categoria 3, 18,1 por cento; categoria 4, 26,4 por cento; categoria 5, 66,7 por cento e não classificável, 5,2 por cento. CONCLUSÕES: observou-se uma maior precisão com a classificação de BI-RADS TM, porém não se conseguiu reduzir a ambigüidade na avaliação das microcalcificações mamárias.

PURPOSE: the aim of this study is to evaluate the accuracy of mammography in the diagnosis of suspicious breast microcalcifications, using BI-RADS TM and Le Gal's classifications. METHODS: one hundred and thirty cases were selected with mammograms contain only microcalcifications of file and initially classified as suspicious (categories 4 and 5) without lesions clinical detectable and reclassified by two examiners, getting a consensus diagnosis. The biopsies were reviewed by two pathologists getting also a consensus diagnosis. Both, mammogram and histopathologic analysis were double blinded reviewed. Qui-square test, Fleiss-square statistic and EPI-INFO 6.0 were used in this study. RESULTS: the correlation between histopathological and mammographic analysis using BI-RADS TM and Le Gal classification showed the same sensitivity of 96.4 percent, specificity of 55.9 and 30.3 percent, positive predictive value (PPV) of 37.5 and 27.5 percent, and accuracy of 64.6 and 44.6 percent respectively. The PPV by BI-RADS TM categories was: category 2, 0 percent; category 3, 1.8 percent; category 4, 30.8 percent; and category 5, 60 percent. The PPV by Le Gal classification was: category 2, 3.1 percent; category 3, 18.1 percent; category 4, 26.4 percent;category 5, 66.7 percent, and non classified 5.2 percent. CONCLUSIONS: the results were better for the classification of BI-RADS™, but it did not get to reduce the ambiguity in assessment of breast microcalcifications.

[Evaluation of breast microcalcifications according to Breast Imaging Reporting and Data System (BI-RADS) and Le Gal's classifications]. / Avaliação de microcalcificações mamárias de acordo com as classificações do Breast Imaging Reporting and Data System (BI-RADS) e de Le Gal.

PURPOSE: the aim of this study is to evaluate the accuracy of mammography in the diagnosis of suspicious breast microcalcifications, using BI-RADS and Le Gal's classifications. METHODS: one hundred and thirty cases were selected with mammograms contain only microcalcifications of file and initially classified as suspicious (categories 4 and 5) without lesions clinical detectable and reclassified by two examiners, getting a consensus diagnosis. The biopsies were reviewed by two pathologists getting also a consensus diagnosis. Both, mammogram and histopathologic analysis were double blinded reviewed. Chi-square test, Fleiss-square statistic and EPI-INFO 6.0 were used in this study. RESULTS: the correlation between histopathological and mammographic analysis using BI-RADS and Le Gal classification showed the same sensitivity of 96.4%, specificity of 55.9 and 30.3%, positive predictive value (PPV) of 37.5 and 27.5%, and accuracy of 64.6 and 44.6% respectively. The PPV by BI-RADS categories was: category 2, 0%; category 3, 1.8%; category 4, 30.8%; and category 5, 60%. The PPV by Le Gal classification was: category 2, 3.1%; category 3, 18.1%; category 4, 26.4%;category 5, 66.7%, and non classified 5.2%. CONCLUSIONS: the results were better for the classification of BI-RADS, but it did not get to reduce the ambiguity in assessment of breast microcalcifications.

Galactocele: clasificación / Galactocele: classification

Los galactoceles se hacen semisólidos a medida que evolucionan; por lo tanto; la clínica, la imagenología y la citopatología sufren algunos cambios. Por estas razones los hemos clasificados en tres estadios. Estadio I, o agudo, de contenido líquido y aspecto uniforme. Estadio II, o intermedio, con contenido líquido y semisólido, de aspecto mixto. Estadio III, o crónico, de contenido semisólido y aspecto grumoso. La ecotomografía mamaria revela imagen anecogénica en el estadio I, mixta en el II e hipoecogénica en el III. La mamografía en los estadios I y II ocacionan dudas diagnósticas por lo que requieren complemento ecotomográfico, no así en el estadio III, donde las imagenes son propias: presencia de cápsula, fondo grumoso como copos de nieve y a veces calcificaciones gruesas. La citopatología también muestra variaciones en los diferentes estadios. Los galactoceles estudiados tenían como antecedentes embarazo y lactancia

Beyond fibrocystic disease: the evolving concept of pre-malignant breast cancer

There is a broad spectrum of benign and pre-malignant breast disease that may be diagnosed on biopsy. Some of these conditions place affected women at significantly increased risk for breast cancer, particularly the atypical hyperplasias. All health care workers should be aware of recent developments in this important aspect of breast disease. The classification, criteria for diagnosis, associated risks for cancer development and various controversies related to these benign and borderline breast diseases are discussed.(AU)

Beyond fibrocystic disease. The evolving concept of pre-malignant breast disease.

There is a broad spectrum of benign and pre-malignant breast disease that may be diagnosed on biopsy. Some of these conditions place affected women at significantly increased risk for breast cancer, particularly the atypical hyperplasias. All health care workers should be aware of recent developments in this important aspect of breast disease. The classification, criteria for diagnosis, associated risks for cancer development and various controversies related to these benign and borderline breast diseases are discussed.

Malformaciones o anomalías en el desarrollo de la mama / Malformations or abnormalities in breast development

Las malformaciones mamarias se clasifican en congénitas o adquiridas; por defecto o por exceso y en anomalías de número, forma, tamaño y situación. En el capítulo se anlizan los diversos tipos de malformaciones tales como atelia y politelia; amastia, polimastia, macro y micromastia. Se mencionan también algunas alteraciones funcionales como la hipertrofia juvenil y la del embarazo. Se hace hincapié en el pezón invertido o umbilicado por sus repercusiones en la lactancia materna