Developing a predictive model for rapid stroke diagnosis via blood using mid-infrared spectroscopy
Saiko Kino1, Tomohisa Ishida2, Kuniyasu Niizuma3
1Graduate School of Biomedical Engineering, Tohoku University, 6-6-05, Aza-Aoba, Aramaki, Aoba-ku, Sendai, Miyagi 980-8579, Japan.
Background:
The diagnosis of stroke requires rapid and objective methods for timely therapeutic intervention. However, the high incidence of stroke and limited healthcare resources, including imaging modalities, underscores the need for rapid diagnostic tools. Mid-infrared (MIR) spectroscopy is a comprehensive blood analysis method characterized by simplicity, speed, and cost-effectiveness. We analyzed plasma using MIR spectroscopy to evaluate its diagnostic value in differentiating patients with stroke from non-stroke individuals, identifying stroke subtypes, and predicting the volume of affected brain lesions.
Methods:
Peripheral blood plasma from 324 individuals (160 patients with stroke, 3 with transient ischemic attack, 8 with stroke mimics, and 153 controls) was analyzed using attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy. Partial least squares analysis of absorbance at each wavenumber was performed, and diagnostic accuracy was assessed by cross-validation.
Results:
The model distinguished patients with stroke from non-stroke individuals with 87.9 % accuracy and an area under the curve of 0.97. Among patients with stroke, ischemic stroke was identified with 73 % classification accuracy, outperforming hemorrhagic stroke. The spectral characteristics suggested structural changes in plasma proteins, indicating protein aggregation, along with marked increase in lipid concentrations. Lesion volume prediction yielded correlation coefficients of 0.61 and 0.34 in intracerebral hemorrhage and ischemic stroke, respectively.
Conclusions:
ATR-FTIR spectroscopy is a promising and minimally invasive approach for rapid stroke detection.
