Radiomics for Identification of Active Bone Marrow from CT: An Exploratory Study

Samanta Rosati, Gabriella Balestra1, Pierfrancesco Franco2, Christian Fiandra2, Francesca Arcadipane2, Patrick Silvetti2, Umberto Ricardi2, Elena Gallio3

  • 1POLITECNICO DI TORINO
  • 2Università di Torino
  • 3A.O.U. Città della Salute e della Scienza di Torino

Details

10:00 - 17:00 | Mon 29 Oct | Foyer | A1P-E.5

Session: Cognitive Computing & Deep Learning in Life Sciences

Full Text

Digital Library Website

Abstract

During radiotherapy treatment of anal cancer it’s important to avoid active bone marrow (actBM). Actually, actBM is identified from PET and MRI and no approached have been developed for its identification from CT images. This study aims to use radiomics for detecting actBM on CT. Our approach is based on 36 first-order and texture features used as input for a Decision Tree to discriminate between active and inactive BM. Results show that actBM segmentation from CT overlaps the identification from PET for more than 75% in 4 out of 5 patients in lumbosacral and iliac structures.