Multi-view iterative random walker for automated salvageable tissue delineation in ischemic stroke from...

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Medicine

Background:

• Accurate identification of salvageable brain tissue (penumbra) is critical for effective interventional stroke therapy.
• Differentiating core and penumbra tissues using diffusion and perfusion MRI is essential for ischemic stroke treatment.

Purpose of the Study:

• To develop a fully automated method for lesion delineation and quantification of salvageable tissue in ischemic stroke.
• To improve the accuracy and efficiency of penumbra detection in multi-sequence MRI.

Main Methods:

• A novel one-shot multi-view iterative random walker (MIRW) method was developed for automated lesion delineation.
• MIRW utilizes hierarchical decomposition of multi-sequence MRI properties and automated seed point detection for iterative segmentation.
• The method processes axial, coronal, and sagittal MRI volumes for efficient tissue quantification.

Main Results:

• The MIRW method achieved a Dice Similarity Coefficient (DSC) of 83.5% for penumbra detection.
• The method demonstrated a computational time of 98.23 seconds/volume, indicating high efficiency.
• Performance was validated on challenging adult ischemic stroke datasets against manual ground-truth.

Conclusions:

• The MIRW method shows significant improvement over state-of-the-art techniques for penumbra detection on the ISLES benchmark dataset.
• Quantitative measures highlight the potential of MIRW for computational analysis and penumbra quantification in stroke patients.
• Accurate penumbra quantification is crucial for selecting appropriate candidates for recanalization therapy.