Volumetric Quantification of Stroke Lesions Using DW-MRI and 3D U-Net Approach

DOI

10.2991/978-94-6463-654-3_8How to use a DOI?

Keywords

Stroke; diffusion MRI; CNN; U-Net; ADC; Artificial intelligence; Deep learning

Abstract

Diffusion-Weighted Magnetic Resonance Imaging (DW-MRI) is an effective method for early ischemic stroke detection, often identifying strokes within the first few minutes of onset. Additionally, it enables the quantification of lesion volume, which is a critical factor in determining whether to proceed with endovascular thrombectomy, particularly beyond the 6-hour onset window. However, variability in clinical assessments of lesion volume can lead to missed opportunities for reperfusion therapy.

This research aims to develop Convolutional Neural Network (CNN) algorithms for the automated segmentation of acute ischemic lesions in DW-MRI. Specifically, it focuses on highlighting and evaluating the performance of the 3D U-Net segmentation method, emphasizing its effectiveness in accurately identifying and characterizing ischemic lesions.

The clinical utility of the method was evaluated in patients with confirmed brain infarction. To assess the effectiveness of the proposed approach, metrics including sensitivity and specificity were consistently applied across all segmentation methods. The reference data, serving as the ground truth, were predominantly determined through assessments by radiologists.

The results demonstrate that our proposed segmentation method provides a precise estimation of lesion volume compared to the current manual method used in clinical practice. Specifically, the 3D U-Net segmentation achieved the following results: a precision of 85.57%, a sensitivity of 80.69%, a specificity of 99.92%, an accuracy of 99.809%, and an F1 score of 85.19%. When compared to two other segmentation methods, our approach underscores its utility in achieving accurate lesion volume estimation.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Wanis Barreh
AU  - Ines Ben Alaya
AU  - Rayhane Ben Amor
AU  - Iyadh Riahi
AU  - Ridha Ben Salah
PY  - 2025
DA  - 2025/02/24
TI  - Volumetric Quantification of Stroke Lesions Using DW-MRI and 3D U-Net Approach
BT  - Proceedings of the  International Conference on Decision Aid and Artificial Intelligence (ICODAI 2024)
PB  - Atlantis Press
SP  - 93
EP  - 102
SN  - 2589-4919
UR  - https://doi.org/10.2991/978-94-6463-654-3_8
DO  - 10.2991/978-94-6463-654-3_8
ID  - Barreh2025
ER  -