Share Email Print
cover

Proceedings Paper

A multi-image approach to CADx of breast cancer with integration into PACS
Author(s): Matthias Elter; Thomas Wittenberg; Rüdiger Schulz-Wendtland; Thomas M. Deserno
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

While screening mammography is accepted as the most adequate technique for the early detection of breast cancer, its low positive predictive value leads to many breast biopsies performed on benign lesions. Therefore, we have previously developed a knowledge-based system for computer-aided diagnosis (CADx) of mammographic lesions. It supports the radiologist in the discrimination of benign and malignant lesions. So far, our approach operates on the lesion level and employs the paradigm of content-based image retrieval (CBIR). Similar lesions with known diagnosis are retrieved automatically from a library of references. However, radiologists base their diagnostic decisions on additional resources, such as related mammographic projections, other modalities (e.g. ultrasound, MRI), and clinical data. Nonetheless, most CADx systems disregard the relation between the craniocaudal (CC) and mediolateral-oblique (MLO) views of conventional mammography. Therefore, we extend our approach to the full case level: (i) Multi-frame features are developed that jointly describe a lesion in different views of mammography. Taking into account the geometric relation between different images, these features can also be extracted from multi-modal data; (ii) the CADx system architecture is extended appropriately; (iii) the CADx system is integrated into the radiology information system (RIS) and the picture archiving and communication system (PACS). Here, the framework for image retrieval in medical applications (IRMA) is used to support access to the patient's health care record. Of particular interest is the application of the proposed CADx system to digital breast tomosynthesis (DBT), which has the potential to succeed digital mammography as the standard technique for breast cancer screening. The proposed system is a natural extension of CADx approaches that integrate only two modalities. However, we are still collecting a large enough database of breast lesions with images from multiple modalities to evaluate the benefits of the proposed approach on.

Paper Details

Date Published: 13 March 2009
PDF: 8 pages
Proc. SPIE 7264, Medical Imaging 2009: Advanced PACS-based Imaging Informatics and Therapeutic Applications, 72640X (13 March 2009); doi: 10.1117/12.810983
Show Author Affiliations
Matthias Elter, Fraunhofer-Institute for Integrated Circuits (Germany)
Thomas Wittenberg, Fraunhofer-Institute for Integrated Circuits (Germany)
Rüdiger Schulz-Wendtland, Friedrich-Alexander-Univ. Erlangen-Nürnberg (Germany)
Thomas M. Deserno, RWTH Aachen Univ. (Germany)


Published in SPIE Proceedings Vol. 7264:
Medical Imaging 2009: Advanced PACS-based Imaging Informatics and Therapeutic Applications
Khan M. Siddiqui; Brent J. Liu, Editor(s)

© SPIE. Terms of Use
Back to Top