Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Dual-Encoder Transformer-Based Multimodal Learning for Ischemic Stroke Lesion Segmentation Using Diffusion MRI

Muhammad Usman, Azka Rehman, Muhammad Mutti Ur Rehman, Abd Ur Rehman, Muhammad Umar Farooq

TL;DR

This work tackles automated ischemic stroke lesion segmentation from diffusion MRI by benchmarking CNN- and transformer-based architectures on ISLES 2022 and introducing a dual-encoder TransUNet that preserves modality-specific representations from DWI and ADC. By incorporating three-slice contextual information, the approach achieves superior Dice scores, with the three-slice dual-encoder configuration reaching 85.4% and outperforming single-slice and single-encoder baselines. The findings highlight the value of multimodal fusion and limited 3D context in accurately delineating heterogeneous stroke lesions, suggesting strong potential for automated clinical decision support. The framework establishes a robust baseline for multimodal diffusion MRI segmentation and points to future extensions, including full volumetric segmentation, external cohort validation, additional imaging modalities, and uncertainty-aware predictions.

Abstract

Accurate segmentation of ischemic stroke lesions from diffusion magnetic resonance imaging (MRI) is essential for clinical decision-making and outcome assessment. Diffusion-Weighted Imaging (DWI) and Apparent Diffusion Coefficient (ADC) scans provide complementary information on acute and sub-acute ischemic changes; however, automated lesion delineation remains challenging due to variability in lesion appearance. In this work, we study ischemic stroke lesion segmentation using multimodal diffusion MRI from the ISLES 2022 dataset. Several state-of-the-art convolutional and transformer-based architectures, including U-Net variants, Swin-UNet, and TransUNet, are benchmarked. Based on performance, a dual-encoder TransUNet architecture is proposed to learn modality-specific representations from DWI and ADC inputs. To incorporate spatial context, adjacent slice information is integrated using a three-slice input configuration. All models are trained under a unified framework and evaluated using the Dice Similarity Coefficient (DSC). Results show that transformer-based models outperform convolutional baselines, and the proposed dual-encoder TransUNet achieves the best performance, reaching a Dice score of 85.4% on the test set. The proposed framework offers a robust solution for automated ischemic stroke lesion segmentation from diffusion MRI.

Dual-Encoder Transformer-Based Multimodal Learning for Ischemic Stroke Lesion Segmentation Using Diffusion MRI

TL;DR

This work tackles automated ischemic stroke lesion segmentation from diffusion MRI by benchmarking CNN- and transformer-based architectures on ISLES 2022 and introducing a dual-encoder TransUNet that preserves modality-specific representations from DWI and ADC. By incorporating three-slice contextual information, the approach achieves superior Dice scores, with the three-slice dual-encoder configuration reaching 85.4% and outperforming single-slice and single-encoder baselines. The findings highlight the value of multimodal fusion and limited 3D context in accurately delineating heterogeneous stroke lesions, suggesting strong potential for automated clinical decision support. The framework establishes a robust baseline for multimodal diffusion MRI segmentation and points to future extensions, including full volumetric segmentation, external cohort validation, additional imaging modalities, and uncertainty-aware predictions.

Abstract

Accurate segmentation of ischemic stroke lesions from diffusion magnetic resonance imaging (MRI) is essential for clinical decision-making and outcome assessment. Diffusion-Weighted Imaging (DWI) and Apparent Diffusion Coefficient (ADC) scans provide complementary information on acute and sub-acute ischemic changes; however, automated lesion delineation remains challenging due to variability in lesion appearance. In this work, we study ischemic stroke lesion segmentation using multimodal diffusion MRI from the ISLES 2022 dataset. Several state-of-the-art convolutional and transformer-based architectures, including U-Net variants, Swin-UNet, and TransUNet, are benchmarked. Based on performance, a dual-encoder TransUNet architecture is proposed to learn modality-specific representations from DWI and ADC inputs. To incorporate spatial context, adjacent slice information is integrated using a three-slice input configuration. All models are trained under a unified framework and evaluated using the Dice Similarity Coefficient (DSC). Results show that transformer-based models outperform convolutional baselines, and the proposed dual-encoder TransUNet achieves the best performance, reaching a Dice score of 85.4% on the test set. The proposed framework offers a robust solution for automated ischemic stroke lesion segmentation from diffusion MRI.
Paper Structure (20 sections, 1 equation, 3 figures, 2 tables)

This paper contains 20 sections, 1 equation, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Problem Formulation
  4. Dataset Description
  5. Preprocessing Pipeline
  6. Baseline Segmentation Models
  7. Customization of TransUNet Architectures
  8. Single-Encoder TransUNet
  9. Dual-Encoder TransUNet with Single-Slice Input
  10. Dual-Encoder TransUNet with Three-Slice Input
  11. Training Strategy
  12. Data Augmentation
  13. Implementation Details
  14. Results
  15. Evaluation Metrics
  16. ...and 5 more sections

Figures (3)

  • Figure 1: Single-encoder TransUNet architecture for multimodal ischemic stroke lesion segmentation. DWI and ADC inputs are fused via early concatenation and processed through a shared encoder, transformer module, and decoder.
  • Figure 2: Dual-encoder TransUNet with single-slice input. DWI and ADC are processed by independent encoders, and modality-specific features are fused at the bottleneck before transformer decoding.
  • Figure 3: Dual-encoder TransUNet with three-slice input. Three consecutive slices per modality provide limited inter-slice context, improving lesion boundary delineation while maintaining slice-wise training efficiency.