Tatyana Ivanovska1, Thomas G Jentschke2, Amro Daboul3, Katrin Hegenscheid4, Henry Völzke5, Florentin Wörgötter2. 1. Department of Computational Neuroscience, Georg-August-University, Friedrich-Hund Platz, 1, 37077, Göttingen, Germany. tiva@phys.uni-goettingen.de. 2. Department of Computational Neuroscience, Georg-August-University, Friedrich-Hund Platz, 1, 37077, Göttingen, Germany. 3. Department of Prosthodontics, Gerodontology and Biomaterials, University Medicine Greifswald, Fleischmannstr. 42-44, 17475, Greifswald, Germany. 4. Unfallkrankenhaus Berlin, Warener Str. 7, 12683, Berlin, Germany. 5. Institute for Community Medicine, University Medicine Greifswald, Walther-Rathenau-Str. 48, 17489, Greifswald, Germany.
Abstract
PURPOSE: The main purpose of this work is to develop, apply, and evaluate an efficient approach for breast density estimation in magnetic resonance imaging data, which contain strong artifacts including intensity inhomogeneities. METHODS: We present a pipeline for breast density estimation, which consists of intensity inhomogeneity correction, breast volume segmentation, nipple extraction, and fibroglandular tissue segmentation. For the segmentation steps, a well-known deep learning architecture is employed. RESULTS: The average Dice coefficient for the breast parenchyma is [Formula: see text], which outperforms the classical state-of-the-art approach by a margin of [Formula: see text]. CONCLUSION: The proposed solution is accurate and highly efficient and has potential to be applied for big epidemiological data with thousands of participants.
PURPOSE: The main purpose of this work is to develop, apply, and evaluate an efficient approach for breast density estimation in magnetic resonance imaging data, which contain strong artifacts including intensity inhomogeneities. METHODS: We present a pipeline for breast density estimation, which consists of intensity inhomogeneity correction, breast volume segmentation, nipple extraction, and fibroglandular tissue segmentation. For the segmentation steps, a well-known deep learning architecture is employed. RESULTS: The average Dice coefficient for the breast parenchyma is [Formula: see text], which outperforms the classical state-of-the-art approach by a margin of [Formula: see text]. CONCLUSION: The proposed solution is accurate and highly efficient and has potential to be applied for big epidemiological data with thousands of participants.
Entities:
Keywords:
Breast density; Deep learning; MRI; Segmentation
Authors: Margaret M Eberl; Chester H Fox; Stephen B Edge; Cathleen A Carter; Martin C Mahoney Journal: J Am Board Fam Med Date: 2006 Mar-Apr Impact factor: 2.657
Authors: U John; B Greiner; E Hensel; J Lüdemann; M Piek; S Sauer; C Adam; G Born; D Alte; E Greiser; U Haertel; H W Hense; J Haerting; S Willich; C Kessler Journal: Soz Praventivmed Date: 2001
Authors: Bas H M van der Velden; Markus H A Janse; Max A A Ragusi; Claudette E Loo; Kenneth G A Gilhuijs Journal: Sci Rep Date: 2020-10-22 Impact factor: 4.379
Authors: Norbert Hosten; Robin Bülow; Henry Völzke; Martin Domin; Carsten Oliver Schmidt; Alexander Teumer; Till Ittermann; Matthias Nauck; Stephan Felix; Marcus Dörr; Marcello Ricardo Paulista Markus; Uwe Völker; Amro Daboul; Christian Schwahn; Birte Holtfreter; Torsten Mundt; Karl-Friedrich Krey; Stefan Kindler; Maria Mksoud; Stefanie Samietz; Reiner Biffar; Wolfgang Hoffmann; Thomas Kocher; Jean-Francois Chenot; Andreas Stahl; Frank Tost; Nele Friedrich; Stephanie Zylla; Anke Hannemann; Martin Lotze; Jens-Peter Kühn; Katrin Hegenscheid; Christian Rosenberg; Georgi Wassilew; Stefan Frenzel; Katharina Wittfeld; Hans J Grabe; Marie-Luise Kromrey Journal: Healthcare (Basel) Date: 2021-12-24