Seminar : Radiomics: An innovative approach for tumor characterization by image analysis

Roua BEN SASSI of the fera une présentation des travaux effectués lors de son stage, vendredi 7 Juillet 2017, à 14h00 en C218, à Télécom physique Strasbourg (Illkirch).

Title : Radiomics: An innovative approach for tumor characterization by image analysis

Abstract : Currently, the standard method to assess the nature of a tumor (phenotype) consists of performing a biopsy and conducting an anatomopathological analysis on the extracted tissues. This method suffers from several disadvantages. The biopsy is an invasive procedure that can be complicated and sometimes even impossible to perform due to the difficulty of accessing the tumor. On the other hand, the anatomopathological analysis is restricted to a localized area of the tumor whereas the latter is formed by extremely heterogeneous tissues. To offset these problems, a new method, named radiomics, offers the possibility of characterizing the tumor using only the acquired images. Here, feature extraction algorithms are associated with machine learning algorithms on the segmented tumors to automatically determine the phenotype (nature, grade). As part of the project RADIAL, the work conducted in the internship aims at carrying out a literature review on the features and machine learning models used in the Radiomics methods and implementing the workflow of the project (Medical image acquisition → Radiomics features computation → Tumor characteristics determination using machine learning techniques).