Back to the list
Congress: ECR24
Poster Number: C-14978
Type: EPOS Radiologist (scientific)
DOI: https://dx.doi.org/10.26044/ecr2024/C-14978
Authorblock: K. Geißler1, N. Uchiyama2, R. Grimm3, T. Suzuki4, U. Fischer5, D. H. M. Szolar6, H. Meine1, H. von Busch3, R. M. Murakami4; 1Bremen/DE, 2TOKYO/JP, 3Erlangen/DE, 4Tokyo/JP, 5Goettingen/DE, 6Graz/AT
Disclosures:
Kai Geißler: Nothing to disclose
Ms. Nachiko Uchiyama: Nothing to disclose
Dr. Robert Grimm: Employee: Siemens Healthcare GmbH
Dr. Takashi Suzuki: Employee: Siemens Healthcare GmbH
Mr. Uwe Fischer: Nothing to disclose
Herr Dieter H. M. Szolar: Nothing to disclose
Mr. Hans Meine: Nothing to disclose
Dr. Heinrich von Busch: Employee: Siemens Healthcare GmbH
MD Ryusuke Medical Murakami: Nothing to disclose
Keywords: Artificial Intelligence, Breast, MR, Computer Applications-Detection, diagnosis, Segmentation, Cancer
Purpose Breast cancer is the world’s most prevalent cancer in women [1]. Magnetic Resonance Imaging (MRI) is increasingly used for the screening of breast cancer [2]. These examinations contribute to an increasing workload for radiologists. In this study we evaluate the performance of a deep learning-based research application (Figure 1) to detect lesions on Dynamic Contrast-Enhanced (DCE) MRI which may help radiologists perform the task accurately and efficiently.  [fig 1]
Read more Methods and materials A research application for AI-based breast MRI analysis (BreastMARC, v1.6, Siemens Healthcare, Erlangen, Germany) is used to automatically segment the breast tissue, parenchyma, and lesions. Its data processing pipeline is depicted in Figure 2.  [fig 2] The breast tissue and parenchyma segmentation are carried out by deep learning models based on a 3D U-Net [3] and 2D U-Net [4] respectively. The image intensities are normalized by mapping their 2nd and 98th intensity percentiles to 0 and 1. For breast tissue segmentation the images...
Read more Results On the internal development dataset, the breast segmentation model achieved an average Dice score of 0.917 and the fibroglandular tissue segmentation model achieved an average Dice score of 0.787. The raw lesion detection model without false positive reduction achieved a lesion-based sensitivity of 0.88 while producing on average 3.78 false positives per case. The results on the external test set are as follows. In 115 of 116 patients, breast tissue segmentation was successful. Patient-based sensitivity, lesion-based sensitivity, average, and maximum number...
Read more Conclusion The AI-based lesion detection shows high sensitivity with an acceptable level of false positives, potentially supporting the radiologist’s workflow. By adjusting the cut-off value, false positives can be reduced but sensitivity also decreases. Thus, different operating points may be appropriate depending on the expected prevalence of cancer in the studied cohort.  
Read more References [1] Sung, H., Ferlay, J., Siegel, R. L., Laversanne, M., Soerjomataram, I., Jemal, A., & Bray, F. (2021). Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA: a cancer journal for clinicians, 71(3), 209-249. [2] Lee, M. V., Aharon, S., Kim, K., Sunn Konstantinoff, K., Appleton, C. M., Stwalley, D., & Olsen, M. A. (2022). Recent trends in screening breast MRI. Journal of Breast Imaging, 4(1), 39-47. [3] Çiçek, Ö., Abdulkadir, A., Lienkamp,...
Read more
GALLERY