Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Automatic Assessment of Nipple Position in Automated 3D Breast Ultrasound Images

: Schwaab, J.; Gubern-Merida, A.; Wang, L.; Günther, M.

Harz, M. (Ed.); Mertzanidou, T. (Ed.); Hipwell, J. (Ed.) ; International Society of Medical Image Computing and Computer-Assisted Intervention -MICCAI-:
MICCAI-BIA 2015, Proceedings of the 3rd MICCAI Workshop on Breast Image Analysis : Munich, Germany, 2015
München, 2015
Workshop on Breast Image Analysis (MICCAI-BIA) <3, 2015, Munich>
Fraunhofer MEVIS ()

Nipple position is an important basis for clinical analysis of breast images. It is used by clinicians to report on lesion localisation, and it is also
considered as an anatomical landmark for image processing algorithms such as
registration between images of different modalities. Automated 3D breast
ultrasound (ABUS) imaging complements X-ray mammography in breast cancer screening of women with dense breasts. In standard ABUS acquisition routine, the nipple is pinpointed by the technicians. In some cases, however, the
nipple is not visible in the image at all or the manual annotation is incorrect.
Consequently, computer aided detection (CAD) systems may fail due to wrong
assumptions concerning the nipple position. The aim of this study is two fold:
We improved an automatic nipple detection algorithm by incorporating prior
location knowledge using a probabilitic atlas. Secondly, using features
computed by this nipple detection algorithm, we developed a novel algorithm
that assesses the quality of the manual nipple marks given by technicians. The
proposed method for detection of incorrect nipple marks was tested on an
independent dataset of 380 ABUS volumes, resulting in sensitivity and
specificity rates of 0.90 and 0.89, with an area under the curve (AUC) value of
0.94. Furthermore, by incorporating prior location knowledge into the
automated nipple localisation algorithm, nipple detection rate was increased
from 0.82 to 0.85.