Fully automatic brain tumor segmentation with deep learning-based selective attention using overlapping patches and multi-class weighted cross-entropy - CNRS - Centre national de la recherche scientifique Accéder directement au contenu
Article Dans Une Revue Medical Image Analysis Année : 2020

Fully automatic brain tumor segmentation with deep learning-based selective attention using overlapping patches and multi-class weighted cross-entropy

Résumé

In this paper, we present a new Deep Convolutional Neural Networks (CNNs) dedicated to fully automatic segmentation of Glioblastoma brain tumors with high-and low-grade. Where the proposed CNNs model is inspired by the Occipito-Temporal pathway which has a special function called selective attention that uses different receptive field sizes in successive layers to figure out the crucial objects in a scene. Thus, using selective attention technique to develop the CNNs model, helps to maximize the extraction of relevant features from MRI images. We have also treated two more issues: class-imbalance, and the spatial relationship among image Patches which are not addressed in the most state-of-the-art methods. To address the first issue, we propose two steps: equal sampling of images Patches and an experimental analysis of the effect of weighted cross-entropy loss function on the segmentation results. In addition, to overcome the second issue, we have studied the effect of Overlapping Patches against Adjacent Patches where the Overlapping Patches show a better segementation result due to the introduction of the global context as well as the local features of the image Patches compared to the conventionnel Adjacent Patches method. Our experiment results are reported on BRATS-2018 dataset where our End-to-End Deep Learning model achieved the state-of-the-art performance. The Mean Dice score of our fully automatic segmentation model is 0.86, 0.74, 0.74 for whole tumor, tumor core, and enhancing tumor respectively compared to the Dice score of radiologist that is in the range 74%-85%. Moreover, our proposed CNNs model not only computationally efficient at inference time, but it could segment the whole brain in average 16 seconds, in addition it has only 181,124 parameters. Finally, the proposed Deep Learning model provides an accurate and reliable segmentation result, and that make it suitable for adopting in research and as a part of different clinical settings.
Fichier principal
Vignette du fichier
Mostefa MedIA.pdf (2.81 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02533454 , version 1 (06-04-2020)

Identifiants

Citer

Mostefa Ben Naceur, Mohamed Akil, Rachida Saouli, Rostom Kachouri. Fully automatic brain tumor segmentation with deep learning-based selective attention using overlapping patches and multi-class weighted cross-entropy. Medical Image Analysis, 2020, ⟨10.1016/j.media.2020.101692⟩. ⟨hal-02533454⟩
202 Consultations
1142 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More