Development and validation of a deep learning model for detection of breast cancers in mammography from multi-institutional datasets.

The objective of this study was to develop and validate a state-of-the-art, deep learning (DL)-based model for detecting breast cancers on mammography. Mammograms in a hospital development dataset, a hospital test dataset, and a clinic test dataset were retrospectively collected from January 2006 through December 2017 in Osaka City University Hospital and Medcity21 Clinic. The hospital development dataset and a publicly available digital database for screening mammography (DDSM) dataset were used to train and to validate the RetinaNet, one type of DL-based model, with five-fold cross-validation. The model's sensitivity and mean false positive indications per image (mFPI) and partial area under the curve (AUC) with 1.0 mFPI for both test datasets were externally assessed with the test datasets. The hospital development dataset, hospital test dataset, clinic test dataset, and DDSM development dataset included a total of 3179 images (1448 malignant images), 491 images (225 malignant images), 2821 images (37 malignant images), and 1457 malignant images, respectively. The proposed model detected all cancers with a 0.45-0.47 mFPI and had partial AUCs of 0.93 in both test datasets. The DL-based model developed for this study was able to detect all breast cancers with a very low mFPI. Our DL-based model achieved the highest performance to date, which might lead to improved diagnosis for breast cancer.

Déclaration de conflit d'intérêts

D.U. reports grants from Wellness Open Living Labs, LLC, during the conduct of the study; N.O. reports grants and personal fees from Bayer, grants and personal fees from Eisai, personal fees from Sanofi, personal fees from Aska, outside the submitted work; T.T. reports personal fees from Taiho Pharmaceutical Co., Ltd., personal fees from Chugai Pharmaceutical Co., Ltd., personal fees from Kyowa Hakko Kirin Co., Ltd., personal fees from Eisai Co., Ltd., personal fees from Pfizer Japan Inc., personal fees from Novartis Pharma K.K., personal fees from AstraZeneca K.K., personal fees from Takeda Pharmaceutical Co., Ltd., outside the submitted work; S.K. reports personal fees from Chugai Pharmaceutical Co., Ltd., personal fees from Eisai Co., Ltd., personal fees from Pfizer Japan Inc., personal fees from Novartis Pharma K.K., personal fees from Asahi Kasei Pharma Co., Ltd., personal fees from Medicon Inc., outside the submitted work; M.M. reports personal fees from Taisho Pharma Co., Ltd., personal fees from MOCHIDA PHARMACEUTICAL CO., LTD, personal fees from TSUMURA & CO., personal fees from Otsuka Pharmaceutical Co., Ltd., personal fees from Mylan EPD G.K., outside the submitted work; A.Y., S.N., T.M., S.H., M.S., T.S., K.K., H.T., K.M., T.H., A.S., and Y.M. have nothing to disclose.