Title: Integration of radiographic and histological images for the diagnosis of glioblastoma

Authors: Fatiha Alim-Ferhat; Linda Ait Mohammed; Mohammed Abdelaziz

Addresses: Architecture and Multimedia Systems Division, Advanced Technology Development Center, Alger, Algeria ' Architecture and Multimedia Systems Division, Advanced Technology Development Center, Alger, Algeria ' Architecture and Multimedia Systems Division, Advanced Technology Development Center, Alger, Algeria

Abstract: As the number of cancer cases increases, the pathologist's task becomes increasingly difficult. To classify tumours and define their level of aggressiveness, pathologists are faced with analysing a large number of pathological images, hundreds of thousands of them. Computer-aided methods, including artificial intelligence, can potentially improve tumour classification. It makes sense to implement such a system by taking advantage of the two complementary MRI and histological data. This study proposes to use multiple input convolutional neural networks to predict glioma grade. The proposed method was validated using data from the CPM: RAD-PATH 2020, achieved satisfactory results. We propose a dual path residual convolutional neural network model that trains simultaneously from MRI and pathology images. With this approach, we achieve a validation accuracy of 81%, showing that combining the two image sources yields better overall accuracy.

Keywords: glioblastoma; digital pathology images; IRM; deep learning; tumour classification.

DOI: 10.1504/IJBET.2023.135395

International Journal of Biomedical Engineering and Technology, 2023 Vol.43 No.4, pp.329 - 337

Received: 09 Oct 2022
Accepted: 07 Jan 2023
Published online: 08 Dec 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article