< Back to previous page

Project

Improving breast cancer screening through dynamic big data analytics of Quantitative Imaging Biomarkers

Breast cancer still represents a major health problem. Mortality rates, both in Flanders and Slovenia, are similar and above the European average. While both regions have successful breast screening programs, the overall detection rates are still too low. Two main factors contribute to missed cancers: “dense breasts” which lead to a five-fold occurrence of breast cancer and “tissue masking”, resulting from complex glandular tissue structure. These challenges will be addressed by combining approaches pioneered by the two lead investigators from Flanders and Slovenia. An improved screening methodology will be developed by (1) integrating dynamic changes from one screening time point to another (longitudinal assessment) and (2) harnessing the power of big data analytics to extract cancer-predictive features (spatial assessment). A retrospective study will identify hand-crafted (radiomics) and machine-learning based (Convolutional Neural Networks) features on a data base of 9000 women in Flanders and Slovenia. These biomarkers will then be validated prospectively on at least 100 patients per year at each site. The international quantitative imaging framework will be leveraged (Networks of Imaging eXcellence (NIX)-Alliance) to provide the infrastructural backbone for imaging biomarker extraction and testing/fine-tuning. Our results will improve current breast screening schemes with an objective, risk-based stratification procedure based on a personalized risk score.

Date:1 Jan 2021 →  Today
Keywords:breast cancer
Disciplines:Scientific computing not elsewhere classified, Medical imaging and therapy not elsewhere classified, Diagnostic radiology