Radiologist-Level Performance by Using Deep Learning for Segmentation of Breast Cancers on MRI Scans.

To develop a deep network architecture that would achieve fully automated radiologist-level segmentation of cancers at breast MRI. In this retrospective study, 38 229 examinations (composed of 64 063 individual breast scans from 14 475 patients) were performed in female patients (age range, 12-94 years; mean age, 52 years ± 10 [standard deviation]) who presented between 2002 and 2014 at a single clinical site. A total of 2555 breast cancers were selected that had been segmented on two-dimensional (2D) images by radiologists, as well as 60 108 benign breasts that served as examples of noncancerous tissue; all these were used for model training. For testing, an additional 250 breast cancers were segmented independently on 2D images by four radiologists. Authors selected among several three-dimensional (3D) deep convolutional neural network architectures, input modalities, and harmonization methods. The outcome measure was the Dice score for 2D segmentation, which was compared between the network and radiologists by using the Wilcoxon signed rank test and the two one-sided test procedure. The highest-performing network on the training set was a 3D U-Net with dynamic contrast-enhanced MRI as input and with intensity normalized for each examination. In the test set, the median Dice score of this network was 0.77 (interquartile range, 0.26). The performance of the network was equivalent to that of the radiologists (two one-sided test procedures with radiologist performance of 0.69-0.84 as equivalence bounds, When trained on a sufficiently large dataset, the developed 3D U-Net performed as well as fellowship-trained radiologists in detailed 2D segmentation of breast cancers at routine clinical MRI.

Informations de copyright

2022 by the Radiological Society of North America, Inc.

Déclaration de conflit d'intérêts

Disclosures of Conflicts of Interest: L.H. Grant from NIBIB and NIMH through the NIH BRAIN Initiative (R01 EB02157); consulting fee from City College of New York. Y.H. Consulting fee from City College of New York. S.L. No relevant relationships. C.R.S. No relevant relationships. R.L.G. No relevant relationships. I.D.N. Grant from Alfonso Martín Escudero Foundation. A.G.V.B. No relevant relationships. N.O. No relevant relationships. E.S.K. No relevant relationships. D.L. No relevant relationships. D.A. Consulting fee AEAD850301RG4. S.E.W. RSNA Research and Education Foundation grant no. RF1905 (content is solely the responsibility of the authors and does not necessarily represent the official views of the RSNA R&E Foundation). M.H. No relevant relationships. D.F.M. No relevant relationships. K.P. Funded in part through the NIH/NCI Cancer Center Support Grant P30 CA008748 and the Breast Cancer Research Foundation; ongoing research grants include Digital Hybrid Breast PET/MRI for Enhanced Diagnosis of Breast Cancer (HYPMED) H2020 - Research and Innovation Framework Programme PHC-11-2015 #667211-2, A Body Scan for Cancer Detection using Quantum Technology (CANCERSCAN) H2020-FETOPEN-2018-2019-2020-01 # 828978, Multiparametric 18F-Fluoroestradiol PET/MRI coupled with Radiomics Analysis and Machine Learning for Prediction and Assessment of Response to Neoadjuvant Endocrine Therapy in Patients with Hormone Receptor+/HER2− Invasive Breast Cancer Jubiläumsfonds of the Austrian National Bank # Nr: 18207, Deciphering breast cancer heterogeneity and tackling the hypoxic tumor microenvironment challenge with PET/MRI, MSI and radiomics The Vienna Science and Technology Fund LS19-046, MSKCC 2020 Molecularly Targeted Intra-Operative Imaging Award 07/2020-06/2021, Breast Cancer Research Foundation 06/2019 - 05/2021 PI Mark Robson Co-I, NIH R01 Breast Cancer Intravoxel-Incoherent-Motion MRI Multisite (BRIMM) 09/01/2020-08/30/2025 PI Eric Sigmund Co-I, NIH R01 subaward: Abbreviated Non-Contrast-Enhanced MRI for Breast Cancer Screening 09/01/2023-08/31/2025 PI Brian Hargreaves; payment for lectures, service on speakers bureaus and for travel/accommodations/meeting expenses from the European Society of Breast Imaging (MRI educational course, annual scientific meeting) and Siemens Healthcare (lectures). K.J. No relevant relationships. A.E.E. No relevant relationships. P.E. No relevant relationships. E.A.M. No relevant relationships. H.A.M. Grant from NIBIB and NIMH through the NIH BRAIN Initiative (R01 EB028157). L.C.P. Grant from NIH (R01CA247910). E.J.S. Grant from National Institutes of Health/National Cancer Institute (P30 CA008748).