< Terug naar vorige pagina


Towards a CAD System for Breast Cancer Based on Individual Microcalcifications?

Boekbijdrage - Boekhoofdstuk Conferentiebijdrage

Microcalcifications (MCs) are small calcium deposits which are often present in breast tissue. Although they may be an early indicator of breast cancer, they are often present in healthy women. Judging on the malignancy of MCs based on their appearance on mammograms is a challenging task for radiologists as mammography projects a 3D object onto a 2D plane, as a consequence, superposition of tissue may alter their appearance. Although Computer Aided Detection and Diagnosis (CAD) systems have been developed to assist radiologists in detecting and characterizing MCs, they mainly focus on analyzing MCs' spatial configuration rather than their individual characteristics. In this study, we propose a CAD system for the characterization of individual MCs, whose shape is visualized in 3D by scanning breast tissue with micro-CT, a high resolution 3D imaging modality. By calculating a large amount of radiomic features we are able to classify MCs as benign or malignant with 75.88% accuracy, 62.13% sensitivity and 86.39% specificity, outperforming the state of the art and proving that there is a difference regarding the shape and the texture of individual MCs. Although this methodology is not applicable in vivo, its potential for use in intra operative imaging is high, as it can reduce the waiting time between tissue extraction and result of the anatomopathological investigation, facilitating and accelerating the procedure.

Boek: 2018 IEEE 20th International Conference on e-Health Networking, Applications and Services, Healthcom 2018
Pagina's: 1-5
Trefwoorden:Breast Cancer, Computer Aided Detection, Computer Aided Diagnosis, Microcalcifications, Radiomics
  • ORCID: /0000-0002-3601-3212/work/91494580
  • ORCID: /0000-0001-8042-6834/work/84646453
  • ORCID: /0000-0002-1986-2183/work/71466328
  • ORCID: /0000-0002-0688-8173/work/71188403
  • ORCID: /0000-0002-8687-2808/work/71042220
  • ORCID: /0000-0003-3345-4431/work/61830559
  • Scopus Id: 85058329321
  • DOI: https://doi.org/10.1109/healthcom.2018.8531134