Multi-Source Automatic Breast Cancer Diagnosis



Breast cancer is one of the most common cancers in women, which affects approximately 10% of all women at some stage of their life in the western world [89]. Since the cause of the disease remains unknown, early detection and diagnosis is the key for breast cancer control, and it can increase the success of treatment, save lives and reduce cost. X-ray Mammography screening is the most common breast cancer detection method. However, X-ray has well-recognized limitations, and recently, other imaging methods like UltraSound (US) have been used as additional screening tools. Moreover, the probability of missing breast cancer with the combination of mammography and ultrasound is much smaller compared to mammography along, especially in women with dense breasts [37]. In order to eliminate the operator dependency and improve diagnostic accuracy, Computer Aided Detection and Diagnosis (CAD) systems are a valuable and beneficial means for breast cancer detection and classification. Thesis objectives This work aims at achieving results that exceed the current state-of-the-art in multi-modality breast cancer detection systems. The fusion of information from mammograms and US for an improved CAD system of the breast is the primary goal of the project. The main scientific and technical objectives are: ? Create a multi-modal database of breast images, manually annotated and stored with the corre- sponding medical reports. This database is to be made publicly available for the benefit of the research community; ? Study and acquire a deep understanding of the image processing and automatic learning techniques in use in the breast image field, and their critical assessment; ? Conduct research on new representations and formulations for image understanding in the breast cancer field; 1 2 ? Conduct research on machine learning algorithms for the specific application; ? Development of tools that implement the aforementioned methodologies, at least to the stage of prototypes; ? Integration of the new techniques in medical decision support systems currently being developed in the host institution. We anticipate that this effort will result not only scientific contributions but also in a CAD prototype with impact in real clinical practice.

TechReport Number


Cited by

No citations found