Deep Learning for Automatic Segmentation of Background Parenchymal Enhancement in Breast MRI.

Sylwia Nowakowska, Karol Borkowski, Carlotta Ruppert, Patryk Hejduk, Alexander Ciritsis, Anna Landsmann, Magda Macron, Nicole Berger, Andreas Boss, Cristina Rossi

Keywords: breast MRI, segmentation, 2D U-Net, BPE, BI-RADS, breast cancer

TL;DR: Automatic Background Parenchymal Enhancement segmentation from Breast MRI images for standarized classification according to BI-RADS.

Abstract: Contrast-enhanced breast MRI plays a crucial role in the care of women at high risk of developing breast cancer. Contrast agent uptake in the breast tissue, i.e., Background Parenchymal Enhancement (BPE), may be an indicator of a higher risk of developing breast cancer and may limit the detectability of lesions. Not only the degree, but also the area of enhancement are elements of importance in the decision-making process in each case. However, they rely on the visual assessment of the reader and thus suffer from poor reliability and reproducibility. In this study, we have developed and evaluated a deep learning (DL) multiclass algorithm for segmentation of both: the BPE area and the non-enhancing tissue. For training, validation, and testing 3441 slices were used. The mean Dice Similarity Coefficient (DSCmean) for the test set amounted to 0.76. Our results show that accurate BPE segmentation is feasible with DL for all classes of enhancement. Such an algorithm may be implemented as part of a pipeline for precise BPE classification or may find direct clinical application in the management of high-risk patients in breast MRI.

Registration: I acknowledge that acceptance of this work at MIDL requires at least one of the authors to register and present the work during the conference.

Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.

Paper Type: novel methodological ideas without extensive validation

Primary Subject Area: Segmentation

Secondary Subject Area: Application: Radiology

Confidentiality And Author Instructions: I read the call for papers and author instructions. I acknowledge that exceeding the page limit and/or altering the latex template can result in desk rejection.