Key result
Decision tree combining clinical and proteomic markers outperforms logistic regression for detecting OMI with 0.90 AUC.
Why the study?
Prompt recognition of occlusion myocardial infarction (OMI) is essential, but current pathways miss patients who lack electrocardiogram changes.
Does a machine-learning classifier combining proteomic biomarkers and clinical features improve diagnostic accuracy for occlusion myocardial infarction in patients undergoing coronary angiography?
Case-Control (n=130)
Does a machine-learning classifier combining proteomic biomarkers and clinical features improve diagnostic accuracy for occlusion myocardial infarction in patients undergoing coronary angiography?
Answers are grounded in this paper and cite their sources.
May improve pre-angiography OMI detection; extends ML classifiers beyond logistic regression in ACS diagnostics.
Background: Prompt recognition and treatment of occlusion myocardial infarction (OMI) is essential, yet current pathways miss a proportion of patients who have OMI as not all have electrocardiogram changes. This exploratory study aimed to determine if proteomic analysis combined with clinical factors could improve diagnostic accuracy in OMI patients. Methods: In this case-controlled exploratory study 368 proteins were analysed from patients having a myocardial infarction and controls with stable angina. Angiographic and clinical features were recorded. Proteins were analysed using a proximity extension assay. Machine-learning techniques of hybrid and forward feature selection algorithms followed by comparing decision tree and logistical regression analysis were used to indicate the optimal classifier of proteins and clinical factors to increase diagnostic sensitivity in OMI. Results: Plasma samples were obtained from 130 patients, 41 (31.5%) had a non-OMI and 16 (12.3%) had OMI. The other 73 (56.2%) had stable angina with no evidence of myocardial infarction. A combination of 19 clinical features and 87 biomarkers for OMI gave a detection of AUC=0.90 which was higher than identification of OMI by clinical features alone (AUC=0.84) although similar to biomarkers alone (AUC=0.91). The decision tree classifier that included combination of biomarkers and clinical factors reached statistical significance for detection for OMI (p<0.001) compared to the logistical regression tree classifier. Conclusion: In this study we created a classifier for the diagnosis of OMI through a combination of clinical factors and proteins following proteomic analysis. Further refinement with larger cohorts and focused prior feature selection are required for validation.
No takes yet. Share an insight, caveat, or question.
Knoery et al. (2023) conducted a case-control in Occlusion myocardial infarction (OMI) (n=130). Decision tree classifier combining clinical features and proteomic biomarkers vs. Logistical regression classifier was evaluated on Area under the curve (AUC) for detection of OMI using combined blood and clinical markers (p=<0.01). A decision tree classifier combining clinical features and proteomic biomarkers achieved an AUC of 0.90 for detecting occlusion myocardial infarction, significantly outperforming a logistical regression model.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: