Answers are grounded in this paper and cite their sources.
Background Accurate assessment of wasp sting severity is clinically important. Serum Raman spectroscopy may provide molecular information complementary to routine clinical measurements. Methods Forty participants were included: 9 healthy controls, 12 patients with moderate envenomation, and 19 with severe envenomation. Raman spectra were acquired from dried serum samples and processed for analysis of the 600–1800 cm -1 fingerprint region. Clinical, Raman, and integrated models were evaluated using participant-level leave-one-participant-out cross-validation (LOPO-CV). The clinical model used CK, Cr, TBil, and LDH; the Raman model used PC1 and PC2; and the fusion model incorporated all six features. The primary task was three-class classification, with particular clinical interest in distinguishing moderate from severe cases. SHAP analysis was used to describe feature contributions. Results The clinical, Raman, and fusion models achieved accuracies of 0.750, 0.800, and 0.875, respectively, corresponding to 30, 32, and 35 correctly classified participants out of 40. Their respective multiclass AUCs were 0.918, 0.847, and 0.935. In the separate exploratory Moderate-versus-Severe binary analysis (n = 31), the fusion model achieved an accuracy of 0.903 and an AUC of 0.935. Paired comparisons did not establish statistically significant superiority of fusion over the clinical model. LDH and PC1 showed prominent contributions in SHAP analysis. Conclusion Integrating serum Raman spectroscopy with clinical biomarkers showed promising performance for severity stratification following wasp stings, supporting further investigation of this multimodal approach.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: