Deep learning has been widely used in the field of mammographic image classification owing to its superiority in automatic feature extraction.However,general deep learning models cannot achieve very satisfactory class...Deep learning has been widely used in the field of mammographic image classification owing to its superiority in automatic feature extraction.However,general deep learning models cannot achieve very satisfactory classification results on mammographic images because these models are not specifically designed for mammographic images and do not take the specific traits of these images into account.To exploit the essential discriminant information of mammographic images,we propose a novel classification method based on a convolutional neural network.Specifically,the proposed method designs two branches to extract the discriminative features from mammographic images from the mediolateral oblique and craniocaudal(CC)mammographic views.The features extracted from the two-view mammographic images contain complementary information that enables breast cancer to be more easily distinguished.Moreover,the attention block is introduced to capture the channel-wise information by adjusting the weight of each feature map,which is beneficial to emphasising the important features of mammographic images.Furthermore,we add a penalty term based on the fuzzy cluster algorithm to the cross-entropy function,which improves the generalisation ability of the classification model by maximising the interclass distance and minimising the intraclass distance of the samples.The experimental results on The Digital database for Screening Mammography INbreast and MIAS mammography databases illustrate that the proposed method achieves the best classification performance and is more robust than the compared state-ofthe-art classification methods.展开更多
Objective:Mammographic calcifications are a common feature of breast cancer,but their molecular characteristics and treatment implications in hormone receptor-positive(HR+)/human epidermal growth factor receptor 2-neg...Objective:Mammographic calcifications are a common feature of breast cancer,but their molecular characteristics and treatment implications in hormone receptor-positive(HR+)/human epidermal growth factor receptor 2-negative(HER2−)breast cancer remain unclear.Methods:We retrospectively collected mammography records of an HR+/HER2−breast cancer cohort(n=316)with matched clinicopathological,genomic,transcriptomic,and metabolomic data.On the basis of mammographic images,we grouped tumors by calcification status into calcification-negative tumors,tumors with probably benign calcifications,tumors with calcification of lowmoderate suspicion for maligancy and tumors with calcification of high suspicion for maligancy.We then explored the molecular characteristics associated with each calcification status across multiple dimensions.Results:Among the different statuses,tumors with probably benign calcifications exhibited elevated hormone receptor immunohistochemical staining scores,estrogen receptor(ER)pathway activation,lipid metabolism,and sensitivity to endocrine therapy.Tumors with calcifications of high suspicion for malignancy had relatively larger tumor sizes,elevated lymph node metastasis incidence,Ki-67 staining scores,genomic instability,cell cycle pathway activation,and may benefit from cyclin-dependent kinase 4 and 6(CDK4/6)inhibitors.Conclusions:Our research established links between tumor calcifications and molecular features,thus proposing potential precision treatment strategies for HR+/HER2−breast cancer.展开更多
Machine learning can help differentiating benign and malignant lesions seen on mammographic images. Conventional models require handcrafting features for lesion representation. Due to insufficient medical instances, t...Machine learning can help differentiating benign and malignant lesions seen on mammographic images. Conventional models require handcrafting features for lesion representation. Due to insufficient medical instances, the performance of convolutional neural networks(CNNs) can be further increased. This study makes use of transfer learning for mammographic breast lesion diagnosis and deep neural network(DNN) models pre-trained with large-scale natural images are employed. The diagnosis performance is evaluated with the prediction accuracy(ACC) and the area under the curve(AUC) on average. A histologically verified database is analyzed which contains 406 lesions(230 benign and 176 malignant). Involved models include transferred DNNs(GoogLeNet and AlexNet), shallow CNNs(CNN2 and CNN3) that are fully trained with medical instances and boosted by support vector machine(SVM), and two conventional methods which combine handcrafted features and SVM for lesion diagnosis. Experimental results indicate that GoogLeNet achieves the best performance(ACC=0.81, AUC=0.88), followed by AlexNet(ACC=0.79, AUC=0.83) and CNN3(ACC=0.73, AUC=0.82). Knowledge transfer can improve the mammographic breast cancer diagnosis, while its wide application still requires further verification in medical imaging domain.展开更多
基金Guangdong Basic and Applied Basic Research Foundation,Grant/Award Number:2019A1515110582Shenzhen Key Laboratory of Visual Object Detection and Recognition,Grant/Award Number:ZDSYS20190902093015527National Natural Science Foundation of China,Grant/Award Number:61876051。
文摘Deep learning has been widely used in the field of mammographic image classification owing to its superiority in automatic feature extraction.However,general deep learning models cannot achieve very satisfactory classification results on mammographic images because these models are not specifically designed for mammographic images and do not take the specific traits of these images into account.To exploit the essential discriminant information of mammographic images,we propose a novel classification method based on a convolutional neural network.Specifically,the proposed method designs two branches to extract the discriminative features from mammographic images from the mediolateral oblique and craniocaudal(CC)mammographic views.The features extracted from the two-view mammographic images contain complementary information that enables breast cancer to be more easily distinguished.Moreover,the attention block is introduced to capture the channel-wise information by adjusting the weight of each feature map,which is beneficial to emphasising the important features of mammographic images.Furthermore,we add a penalty term based on the fuzzy cluster algorithm to the cross-entropy function,which improves the generalisation ability of the classification model by maximising the interclass distance and minimising the intraclass distance of the samples.The experimental results on The Digital database for Screening Mammography INbreast and MIAS mammography databases illustrate that the proposed method achieves the best classification performance and is more robust than the compared state-ofthe-art classification methods.
基金supported by grants from the National Key Research and Development Project of China(Grant No.2020YFA0112304)the National Natural Science Foundation of China(Grant Nos.81922048,82072922,91959207,and 92159301)+3 种基金the Program of Shanghai Academic/Technology Research Leader(Grant No.20XD1421100)the Shanghai Key Laboratory of Breast Cancer(Grant No.12DZ2260100)the Clinical Research Plan of SHDC(Grant Nos.SHDC2020CR4002 and SHDC2020CR5005)the SHDC Municipal Project for Developing Emerging and Frontier Technology in Shanghai Hospitals(Grant No.SHDC12021103).
文摘Objective:Mammographic calcifications are a common feature of breast cancer,but their molecular characteristics and treatment implications in hormone receptor-positive(HR+)/human epidermal growth factor receptor 2-negative(HER2−)breast cancer remain unclear.Methods:We retrospectively collected mammography records of an HR+/HER2−breast cancer cohort(n=316)with matched clinicopathological,genomic,transcriptomic,and metabolomic data.On the basis of mammographic images,we grouped tumors by calcification status into calcification-negative tumors,tumors with probably benign calcifications,tumors with calcification of lowmoderate suspicion for maligancy and tumors with calcification of high suspicion for maligancy.We then explored the molecular characteristics associated with each calcification status across multiple dimensions.Results:Among the different statuses,tumors with probably benign calcifications exhibited elevated hormone receptor immunohistochemical staining scores,estrogen receptor(ER)pathway activation,lipid metabolism,and sensitivity to endocrine therapy.Tumors with calcifications of high suspicion for malignancy had relatively larger tumor sizes,elevated lymph node metastasis incidence,Ki-67 staining scores,genomic instability,cell cycle pathway activation,and may benefit from cyclin-dependent kinase 4 and 6(CDK4/6)inhibitors.Conclusions:Our research established links between tumor calcifications and molecular features,thus proposing potential precision treatment strategies for HR+/HER2−breast cancer.
基金supported in part by the National Key Research and Development Program of China (Grant No. 2016YFC0105102)the Leading Talent of Special Support Project in Guangdong (Grant No. 2016TX03R139)+4 种基金the Shenzhen Key Technical Research Project (Grant No. JSGG20160229203812944)the Science Foundation of Guangdong (Grant Nos. 2017B020229002, 2015B020233011 & 2014A030312006) the National Natural Science Foundation of China (Grant No. 61871374)the Beijing Center for Mathematics and Information Interdisciplinary Sciencesthe Major Scientific Research Project for Universities of Guangdong Province (Grant No. 2016KTSCX167)
文摘Machine learning can help differentiating benign and malignant lesions seen on mammographic images. Conventional models require handcrafting features for lesion representation. Due to insufficient medical instances, the performance of convolutional neural networks(CNNs) can be further increased. This study makes use of transfer learning for mammographic breast lesion diagnosis and deep neural network(DNN) models pre-trained with large-scale natural images are employed. The diagnosis performance is evaluated with the prediction accuracy(ACC) and the area under the curve(AUC) on average. A histologically verified database is analyzed which contains 406 lesions(230 benign and 176 malignant). Involved models include transferred DNNs(GoogLeNet and AlexNet), shallow CNNs(CNN2 and CNN3) that are fully trained with medical instances and boosted by support vector machine(SVM), and two conventional methods which combine handcrafted features and SVM for lesion diagnosis. Experimental results indicate that GoogLeNet achieves the best performance(ACC=0.81, AUC=0.88), followed by AlexNet(ACC=0.79, AUC=0.83) and CNN3(ACC=0.73, AUC=0.82). Knowledge transfer can improve the mammographic breast cancer diagnosis, while its wide application still requires further verification in medical imaging domain.
