Three-Biomarker Joint Strategy for Early and Accurate Diagnosis of Acute Myocardial Infarction via a Multiplex Electrochemiluminescence Immunoarray Coupled with Robust Machine Learning

Figures

• 

  Figure 1. Schematic of the proposed three-biomarker joint strategy for early and accurate AMI diagnosis. (A) ciGold-based ECL immunoarray coupled with a custom-built high-sensitivity ECL imaging system, including the chemiluminescent immuno-Gold nanoassembly, a single electrode based immunoarray chip, and a high-sensitivity imaging system. (B) Raw images of ECL emission from cTnI standard samples. (C) Calibration curves for three biomarkers. (D) Three-biomarker data from 260 patient serum samples was used to construct a machine learning model with 100% sensitivity and 100% specificity.

• 

  Figure 2. (A) Distribution of 4 cardiac biomarkers for chest pain patients with and without AMI. Solid lines represent population means, while the dark and light shaded regions represent 1 and 2 standard deviations from the mean, respectively. Individual dots represent individual subjects. Black dashed lines represent the optimal cutoff for AMI diagnosis as determined by ROC analysis. (B) SVM classification of AMI and non-AMI patients using three biomarkers. AMI patients who would have been misdiagnosed if relying solely on cTnT are circled in red. (C) ROC curves for classification using SVM (red, based on three biomarkers simultaneously), and using individual cardiac biomarkers (black): cTnT (solid line), cTnI (dashed line), h-FABP (dash-dotted line), and copeptin (dotted line).

• 

  Figure 3. SVM substantially outperforms traditional cTnT diagnosis of AMI at early time points. Dots represent cTnT concentrations or SVM-computed score for individual AMI patients within 12 h of symptom onset, while black dashed lines show the optimum ROC cutoff values. Red lines and shaded regions show the mean ±2 standard deviations of these values for the non-AMI population, highlighting that at early time points many AMI patients lack elevated cTnT values. These lead to misdiagnoses (blue points with red circles) in the case of cTnT testing, while three biomarker based SVM model accurately classifies these patients

• 

  Figure 4. Percentage of cTnT and SVM misclassification for AMI at early time points.