Accurate Segmentation of Ischemic Stroke Lesion Areas Based on Pre-trained UNets

DOI

10.2991/978-94-6463-540-9_90How to use a DOI?

Keywords

Machine Learning; Transfer Learning; Image Processing

Abstract

Due to the mortality and disabilities caused by ischemic stroke, it is of great significance to provide accurate segmentation during the treatment of ischemic stroke. In this study, pre-trained UNets were utilized to save the computational resource and provide accurate prediction of lesion area caused by ischemic stroke. More specifically, this study proposed four ImageNet-based pre-trained models as encoders for constructing UNets, which aim at applying non-medical priori knowledge to improve the efficiency and performance of each neural network. Additionally, a self-defined UNet was built as a baseline. All five models were trained on the ATLAS 2.0 dataset after a data filter and binary focal loss were used to mitigate the data imbalance. Finally, trainable parameters, training time and segmentation results from all five models were used for comparison. Experimental results indicate that pre-trained models achieve a recall rate of approximately 0.95 on average and consume only half of the time that self-defined UNet costs. Briefly, pre-trained models achieve a more competitive performance than that of the self-defined UNet and can deliver accurate segmentation results for patients suffering from ischemic stroke.

Copyright

© 2024 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  - Zhewen Guo
PY  - 2024
DA  - 2024/10/16
TI  - Accurate Segmentation of Ischemic Stroke Lesion Areas Based on Pre-trained UNets
BT  - Proceedings of the 2024 2nd International Conference on Image, Algorithms and Artificial Intelligence (ICIAAI 2024)
PB  - Atlantis Press
SP  - 901
EP  - 910
SN  - 2352-538X
UR  - https://doi.org/10.2991/978-94-6463-540-9_90
DO  - 10.2991/978-94-6463-540-9_90
ID  - Guo2024
ER  -