Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Segmentation of Ischemic Stroke Lesions using Transfer Learning on Multi-sequence MRI

R. P. Chowdhury, T. Rahman

TL;DR

This work tackles automated segmentation of acute ischemic stroke lesions from multi-sequence MRI by employing transfer learning within a Res-UNet framework. It processes axial, sagittal, and coronal 2D slices with dedicated Hybrid Res-UNet models pretrained on ImageNet and fuses their predictions via a Majority Voting Classifier to form a robust 3D segmentation. Evaluated on the ISLES 2015 dataset, the approach achieves a Dice score of 80.5% and an accuracy of 74.03%, outperforming a DSN baseline and demonstrating the benefits of multi-plane ensemble fusion. The method offers a computationally efficient alternative to full 3D CNNs with potential for real-time clinical deployment, while future improvements could include weighted voting and attention mechanisms to further enhance boundary delineation.

Abstract

The accurate understanding of ischemic stroke lesions is critical for efficient therapy and prognosis of stroke patients. Magnetic resonance imaging (MRI) is sensitive to acute ischemic stroke and is a common diagnostic method for stroke. However, manual lesion segmentation performed by experts is tedious, time-consuming, and prone to observer inconsistency. Automatic medical image analysis methods have been proposed to overcome this challenge. However, previous approaches have relied on hand-crafted features that may not capture the irregular and physiologically complex shapes of ischemic stroke lesions. In this study, we present a novel framework for quickly and automatically segmenting ischemic stroke lesions on various MRI sequences, including T1-weighted, T2-weighted, DWI, and FLAIR. The proposed methodology is validated on the ISLES 2015 Brain Stroke sequence dataset, where we trained our model using the Res-Unet architecture twice: first, with pre-existing weights, and then without, to explore the benefits of transfer learning. Evaluation metrics, including the Dice score and sensitivity, were computed across 3D volumes. Finally, a Majority Voting Classifier was integrated to amalgamate the outcomes from each axis, resulting in a comprehensive segmentation method. Our efforts culminated in achieving a Dice score of 80.5\% and an accuracy of 74.03\%, showcasing the efficacy of our segmentation approach.

Segmentation of Ischemic Stroke Lesions using Transfer Learning on Multi-sequence MRI

TL;DR

This work tackles automated segmentation of acute ischemic stroke lesions from multi-sequence MRI by employing transfer learning within a Res-UNet framework. It processes axial, sagittal, and coronal 2D slices with dedicated Hybrid Res-UNet models pretrained on ImageNet and fuses their predictions via a Majority Voting Classifier to form a robust 3D segmentation. Evaluated on the ISLES 2015 dataset, the approach achieves a Dice score of 80.5% and an accuracy of 74.03%, outperforming a DSN baseline and demonstrating the benefits of multi-plane ensemble fusion. The method offers a computationally efficient alternative to full 3D CNNs with potential for real-time clinical deployment, while future improvements could include weighted voting and attention mechanisms to further enhance boundary delineation.

Abstract

The accurate understanding of ischemic stroke lesions is critical for efficient therapy and prognosis of stroke patients. Magnetic resonance imaging (MRI) is sensitive to acute ischemic stroke and is a common diagnostic method for stroke. However, manual lesion segmentation performed by experts is tedious, time-consuming, and prone to observer inconsistency. Automatic medical image analysis methods have been proposed to overcome this challenge. However, previous approaches have relied on hand-crafted features that may not capture the irregular and physiologically complex shapes of ischemic stroke lesions. In this study, we present a novel framework for quickly and automatically segmenting ischemic stroke lesions on various MRI sequences, including T1-weighted, T2-weighted, DWI, and FLAIR. The proposed methodology is validated on the ISLES 2015 Brain Stroke sequence dataset, where we trained our model using the Res-Unet architecture twice: first, with pre-existing weights, and then without, to explore the benefits of transfer learning. Evaluation metrics, including the Dice score and sensitivity, were computed across 3D volumes. Finally, a Majority Voting Classifier was integrated to amalgamate the outcomes from each axis, resulting in a comprehensive segmentation method. Our efforts culminated in achieving a Dice score of 80.5\% and an accuracy of 74.03\%, showcasing the efficacy of our segmentation approach.

Paper Structure

This paper contains 13 sections, 7 equations, 7 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methodology
  3. Proposed Method
  4. Encoder-Decoder Structure
  5. ResNet and Hybrid ResNet Architecture
  6. Majority Voting Classifier
  7. Loss Function
  8. Results and Discussion
  9. Dataset
  10. Evaluation Metrics
  11. Hyperparameters
  12. Experimental Results
  13. Conclusion

Figures (7)

  • Figure 1: Workflow Diagram of our Proposed Segmentation Method
  • Figure 2: Architecture Diagram of a Baseline UNet Model
  • Figure 3: Architecture Diagram of Hybrid ResNet 50-UNet
  • Figure 4: Flowchart Diagram of Majority Voting Classifier Algorithm
  • Figure 5: Graphs of Dice Score and Loss while training and validation of our proposed model after applying the Majority Voting algorithm
  • ...and 2 more figures