Research
Machine and Deep Learning Radiomics in Breast CT for tumor malignancy prediction
Radiomics is a growing medical image analysis field that aims at the extraction of mineable information from radiographic images, with the goal of developing computerized support systems to improve clinical decision making. Among the clinical tasks and oncological areas, breast cancer diagnosis is one of the main areas where radiomics is most growing, where computer-aided diagnosis (CADx) systems can be developed to improve the diagnostic pipeline through automatic tumor malignancy assessment and reduction of negative biopsy rate. While CADx methods have already been investigated in several breast cancer imaging modalities, their application to dedicated breast CT remains to be investigated. Thanks to its full 3D nature (along with its high spatial and contrast resolution), the application of radiomic-based, computer-aided diagnosis techniques to breast CT may result in the acquisition of important information for breast cancer characterization, potentially improving diagnostic performance by an increase in sensitivity and/or a reduction in negative biopsy rates. Therefore, the goal of this project is to develop a computer-aided diagnosis system for breast tumor classification in breast CT imaging. For this task, state-of-the-art image analysis techniques will be employed to extract imaging biomarkers from tumor images, and machine learning and Deep Learning algorithms will be developed to process these biomarkers and identify tumoral patterns in a large dataset, ultimately resulting in a computerized system being able to recognize between benign and malignant lesions.
Breast tumors are analyzed through traditional machine learning-based pipelines, and deep learning-based artificial intelligence systems, to recognize image-based tumoral patterns and perform the diagnostic classification based on quantitative imaging biomarkers.
Researchers:
Key Publications:
