Gliomas are brain tumors whose progression from low-grade to high-grade indicates a poor prognosis for the patient. The grade of a glioma is determined using invasive techniques, such as analysis of surgical specimens or biopsy. The University Hospital of Poitiers offers a non-invasive alternative through a multimodal magnetic resonance imaging (MRI) assessment that combines anatomical, diffusion, perfusion, and spectroscopy data. In this thesis, we propose to implement automatic classification tools based on the analysis of these multivariate images. To do so, we deploy innovative tools to analyze the behavior of automatic classifiers. This leads us to identify inconsistencies within the most widely used database for the binary classification task of glioma grade. We propose an alternative based on a consensus among five expert radiologists. We then create a complete processing pipeline ranging from acquisition to classification, and explore how multimodal data improves the performance of automatic classification.
Author : Paul Dequidt
Management team :
- Benoit Tremblais
- Pascal Bourdon
- Olfa Ben Ahled
- Benoit Gianelli
- Carole Guillevin
- Christine Fernandez
- Rémy Guillevin
Accepted publications: (1 other international journal currently under review)
- Paul Dequidt, Pascal Bourdon, Benoit Tremblais, Carole Guillevin, Benoit Gianelli, Claire Boutet, Jean-Philippe Cottier, Jean-Noël Vallée, Christine Fernandez-Maloigne, Rémy Guillevin, “Assigning a new gl oma grade label groundtruth for the BraTS dataset using radiologic criteria” , special issue IRBM, Innovation and Research in BioMedical engineering, Elsevier (IF 1,022) forthcoming
- Paul Dequidt, Pascal Bourdon, Benoit Tremblais, Carole Guillevin, Benoit Gianelli, Claire Boutet, Jean-Philippe Cottier, Jean-Noel Vallée, Christine Fernandez-Maloigne, Rémy Guillevin, “Assigning a new glioma grade label ground truth:for the BraTS dataset using radiologic criteria” IPTA 2020, International Conference on Image Processing, Paris, France, 9-12 nov 2020
- Paul Dequidt, Pascal Bourdon, Olfa Ben Ahmed, Benoit Tremblais, Carole Guillevin, Mathieu Naudin, Christine Fernandez-Maloigne, Rémy Guillevin, “Recent advances in Glioma grade classification using machine and deep learning on MR data” , IEEE ICABME, Tripoli-Lebanon, October 17-18, 2019