Portal de Periódicos da CAPES

An Efficient Cascade of U-Net-Like Convolutional Neural Networks Devoted to Brain Tumor Segmentation

2023; Springer Science+Business Media; Linguagem: Inglês

10.1007/978-3-031-33842-7_13

1611-3349

Philippe Bouchet, Jean-Baptiste Deloges, Hugo Canton-Bacara, Gaëtan Pusel, L. Pinot, Othman Elbaz, Nicolas Boutry,

Machine Learning and ELM

A glioma is a fast-growing and aggressive tumor that starts in the glial cells of the brain. They make up about 30% of all brain tumors, and 80% of all malignant brain tumors. Gliomas are considered to be rare tumors, affecting less than 10,000 people each year, with a 5-year survival rate of 6%. If intercepted at an early stage, they pose no danger; however, providing an accurate diagnosis has proven to be difficult. In this paper, we propose a cascade approach using state-of-the-art Convolutional Neural Networks, in order to maximize accuracy in tumor detection. Various U-Net-like networks have been implemented and tested in order to select the network best suited for this problem.