Stroke localization and classification using microwave tomography with k-means clustering and support vector machine.

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Electromagnetics

Background:

• Accurate and rapid stroke classification is vital for patient survival, enabling prompt medication delivery.
• Existing methods may lack the speed or accuracy required for time-sensitive stroke treatment.
• Microwave-based techniques offer a potential non-invasive approach for brain imaging and analysis.

Purpose of the Study:

• To propose and evaluate a novel microwave-based framework for stroke localization and classification.
• To assess the efficacy of combining microwave tomography, k-means clustering, and support vector machine (SVM) for stroke diagnosis.
• To determine the classification accuracy, sensitivity, and specificity of the proposed method.

Main Methods:

• Utilized microwave tomography and the Born iterative method to compute the brain's dielectric profile.
• Employed k-means clustering on the dielectric profile amplitude as input for feature extraction.
• Developed and tested a support vector machine (SVM) classifier using MRI-derived head phantoms.

Main Results:

• The proposed framework achieved 88% classification accuracy in a two-dimensional setting.
• Demonstrated high performance with a sensitivity of 91% and a specificity of 87%.
• Receiver operating characteristic (ROC) curves were used to evaluate the framework's performance across various signal-to-noise ratios.

Conclusions:

• The microwave-based framework shows significant promise for accurate stroke classification.
• The integration of microwave tomography, k-means clustering, and SVM offers a viable approach for rapid stroke diagnosis.
• This technology could improve treatment outcomes for stroke patients by enabling faster and more precise medical interventions.