A Parallel Framework for HCC Detection in DCE-MRI Sequences with Wavelet-Based Description and SVM classification

• Autor(a) principal: Pavan, Ana L. M. [UNESP]
• Data de Publicação: 2018
• Outros Autores: Benabdallah, Marwa, Lebre, Marie-Ange, Pina, Diana Rodrigues de, Jaziri, Faouzi, Vacavant, Antoine, Mtibaa, Achraf, Ali, Hawa Mohamed, Grand-Brochier, Manuel, Rositi, Hugo, Magnin, Benoit, Abergel, Armand, Chabrot, Pascal, Assoc Comp Machinery
• Tipo de documento: Artigo de conferência
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1145/3167132.3167167; http://hdl.handle.net/11449/185292
• topic: Medical image analysis; machine learning; DCE-MRI; liver; HCC; tumor detection; parallelization; wavelet image description

In this article, we propose a complete framework devoted to detect liver HCC (Hepato-Cellular Carcinoma) tumors within DCE-MRI (Dynamic Contrast Enhanced-MRI) sequences. Our system employs different phases of these hepatic image sequences (depending on time after contrast agent injection) to describe local patches with wavelet-based descriptors. By using a SVM (Support Vector Machine)-based classification, we are able to distinguish healthy patches from pathological ones. Moreover, thanks to a parallel image processing strategy, we are able to reduce significantly the running time so that our system may be utilized as a computer aided diagnosis tool in the future. Our experiments show that our contribution is an accurate system for HCC detection, with a small cohort of patients, but representing a high volume of image data to be processed. This work encourages us to conduct deeper researches for detecting complex HCC cases for larger patients cohorts.