Predicting Clinical Outcome in Acute Ischemic Stroke Using Parallel Multi-parametric Feature Embedded Siamese Network

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neurology

Background:

• Stroke is a leading global cause of death and disability, with ischemic stroke being the most prevalent type.
• Multi-parametric magnetic resonance imaging (MRI) is crucial for acute stroke diagnosis, lesion characterization, and assessing reperfusion therapy efficacy.
• Predicting treatment outcomes is challenging due to variability in stroke characteristics and limitations of current diagnostic datasets, which are often scarce and imbalanced.

Purpose of the Study:

• To develop an automated model for predicting stroke treatment outcomes using multi-parametric MRI data.
• To address the challenges of limited sample sizes and high class imbalance in acute-stage stroke imaging datasets.
• To create a valuable tool for clinicians to better assess risks and benefits of reperfusion therapies.

Main Methods:

• Introduction of the parallel multi-parametric feature embedded siamese network (PMFE-SN), a novel deep learning architecture.
• Implementation of PMFE-SN designed to learn effectively from few samples and handle skewed multi-parametric MRI data.
• Definition and application of five evaluation metrics robust to class imbalance for performance assessment.

Main Results:

• The PMFE-SN model demonstrated superior performance across all defined metrics compared to state-of-the-art techniques.
• The model achieved significant predictive accuracy for both minority (0.67) and majority (0.61) classes, even with minimal training data.
• In contrast, traditional methods using handcrafted features yielded 0 accuracy for the minority class and 0.33 for the majority class.

Conclusions:

• The proposed PMFE-SN is highly effective for predicting stroke treatment outcomes from imbalanced, limited multi-parametric MRI data.
• This deep learning approach offers a promising solution for clinical decision-making in acute stroke management.
• PMFE-SN advances the field by enabling reliable predictions even with scarce data, overcoming a major hurdle in stroke imaging research.