Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Stroke Locus Net: Occluded Vessel Localization from MRI Modalities

Mohamed Hamad, Muhammad Khan, Tamer Khattab, Mohamed Mabrok

TL;DR

The paper addresses the challenge of localizing the occluded vessel in ischemic stroke from MRI data, where dedicated vascular imaging is not always available. It proposes Stroke Locus Net, an end-to-end pipeline with a segmentation branch (nnU-Net) for lesion masks and arterial-atlas analysis to infer the stroke source, plus a generation branch that synthesizes MRA from MRI using a pGAN and a vessel-segmentation module, with MRI–MRA fusion for visualization. A key contribution is integrating a probabilistic arterial atlas for territory-based source identification and adopting a loss for MRA synthesis that combines $L_{L1}$, $L_{perc}$, and $L_{condGAN-k}$ terms to enable MRI-to-MRA translation, along with validation on the ATLAS and IXI datasets. The results demonstrate the feasibility of MRI-only vascular localization and visualization, but also highlight limitations in MRA realism and the need for stroke-patient paired data to drive clinical adoption.

Abstract

A key challenge in ischemic stroke diagnosis using medical imaging is the accurate localization of the occluded vessel. Current machine learning methods in focus primarily on lesion segmentation, with limited work on vessel localization. In this study, we introduce Stroke Locus Net, an end-to-end deep learning pipeline for detection, segmentation, and occluded vessel localization using only MRI scans. The proposed system combines a segmentation branch using nnUNet for lesion detection with an arterial atlas for vessel mapping and identification, and a generation branch using pGAN to synthesize MRA images from MRI. Our implementation demonstrates promising results in localizing occluded vessels on stroke-affected T1 MRI scans, with potential for faster and more informed stroke diagnosis.

Stroke Locus Net: Occluded Vessel Localization from MRI Modalities

TL;DR

The paper addresses the challenge of localizing the occluded vessel in ischemic stroke from MRI data, where dedicated vascular imaging is not always available. It proposes Stroke Locus Net, an end-to-end pipeline with a segmentation branch (nnU-Net) for lesion masks and arterial-atlas analysis to infer the stroke source, plus a generation branch that synthesizes MRA from MRI using a pGAN and a vessel-segmentation module, with MRI–MRA fusion for visualization. A key contribution is integrating a probabilistic arterial atlas for territory-based source identification and adopting a loss for MRA synthesis that combines , , and terms to enable MRI-to-MRA translation, along with validation on the ATLAS and IXI datasets. The results demonstrate the feasibility of MRI-only vascular localization and visualization, but also highlight limitations in MRA realism and the need for stroke-patient paired data to drive clinical adoption.

Abstract

A key challenge in ischemic stroke diagnosis using medical imaging is the accurate localization of the occluded vessel. Current machine learning methods in focus primarily on lesion segmentation, with limited work on vessel localization. In this study, we introduce Stroke Locus Net, an end-to-end deep learning pipeline for detection, segmentation, and occluded vessel localization using only MRI scans. The proposed system combines a segmentation branch using nnUNet for lesion detection with an arterial atlas for vessel mapping and identification, and a generation branch using pGAN to synthesize MRA images from MRI. Our implementation demonstrates promising results in localizing occluded vessels on stroke-affected T1 MRI scans, with potential for faster and more informed stroke diagnosis.

Paper Structure

This paper contains 16 sections, 1 equation, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Stroke Lesion Segmentation
  4. Vessel Localization Methods
  5. Medical Image Generation
  6. Arterial Atlases in Stroke Diagnosis
  7. Methodology
  8. Segmentation Branch
  9. Lesion Segmentation using nnUNet
  10. Arterial Atlas Analysis
  11. Generation Branch
  12. MRA Synthesis
  13. Vessel Segmentation
  14. MRI-MRA Fusion
  15. Results and Discussion
  16. ...and 1 more sections

Figures (5)

  • Figure 1: Stroke Locus Net Architecture Diagram
  • Figure 2: Fused image max (top), fused image mean (middle), fused image sum (bottom)
  • Figure 3: Multi-slice visualization of stroke lesion mask (red) overlaid on the arterial territory atlas and the MRI scan.
  • Figure 4: 3D visualization of quantitative vessel segmentation performance across modalities
  • Figure 5: Comparison of synthetic MRA generation with ground truth MRA and input MRI