Towards Rapid and Low-Cost Stroke Detection Using SERS and Machine Learning.

Area of Science:

• Nanotechnology
• Biomedical Diagnostics
• Machine Learning

Background:

• Stroke impacts millions globally, requiring rapid diagnosis for effective treatment.
• Current hospital imaging delays treatment, highlighting the need for portable diagnostic tools.
• Differentiating stroke types (hemorrhagic vs. ischemic) is critical for appropriate patient management.

Purpose of the Study:

• To develop a proof-of-concept for a rapid, low-cost, point-of-care diagnostic assay for stroke type differentiation.
• To utilize silver nanostars (AgNS) and Surface-Enhanced Raman Spectroscopy (SERS) with machine learning (ML) for plasma analysis.
• To enable pre-hospital stroke diagnosis and potentially improve patient outcomes.

Main Methods:

• Silver nanostars (AgNS) were incubated with human plasma, spiked with glial fibrillary acidic protein (GFAP) as a biomarker for hemorrhagic stroke.
• Surface-Enhanced Raman Spectroscopy (SERS) was employed to analyze the spectral fingerprints of plasma samples on an aluminum foil substrate.
• Machine learning (ML) models, including ensemble modeling and feature engineering, were used to classify stroke types based on SERS spectra.

Main Results:

• The assay successfully differentiated between hemorrhagic and ischemic stroke mimics within 15 minutes.
• Optimized AgNS-plasma incubates, controlled ratios, and a low-cost substrate were key innovations.
• The integrated ML model achieved rapid and precise stroke predictions within seconds, identifying stroke-specific protein profiles.

Conclusions:

• This SERS-based assay with ML offers a promising approach for rapid, pre-hospital stroke diagnostics.
• The developed method demonstrates potential for low-cost, point-of-care application, aiding immediate clinical decision-making.
• Further development could significantly improve patient outcomes by enabling faster stroke type identification.