Accurate Diagnosis of Respiratory Viruses Using an Explainable Machine Learning with Mid-Infrared Biomolecular Fingerprinting of Nasopharyngeal Secretions

2025-04-12

AI Summary

Accurate and rapid identification of multiple respiratory viruses is critical for effective pandemic response. This study proposes an end-to-end diagnostic framework integrating attenuated total reflectance Fourier-transform infrared (ATR-FTIR) spectroscopy (4000–900 cm⁻¹) with a novel Rotary Position Embedding-Sparse Attention Transformer (RoPE-SAT)—the first application of RoPE-SAT to mid-infrared biological fingerprint analysis. The method enables precise multi-virus discrimination within 10 minutes. Spectral preprocessing employs standard normal variate (SNV) normalization and second-order derivative transformation, while Grad-CAM visualization provides biologically interpretable localization of virus-specific spectral bands—particularly Amide I/II, nucleic acid, and lipid regions. Evaluated on two independent clinical cohorts, the model achieves ≥94.40% sensitivity and specificity for discriminating influenza B, SARS-CoV-2, *Mycoplasma pneumoniae*, and healthy controls, demonstrating high robustness and biological interpretability.

Technology Category

• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Computer Vision: Multi-modal Vision
• Machine Learning: Transparent, Interpretable, Explainable ML

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Security and Privacy: Data transparency and provenance
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Accurate identification of respiratory viruses (RVs) is critical for outbreak control and public health. This study presents a diagnostic system that combines Attenuated Total Reflectance Fourier Transform Infrared Spectroscopy (ATR-FTIR) from nasopharyngeal secretions with an explainable Rotary Position Embedding-Sparse Attention Transformer (RoPE-SAT) model to accurately identify multiple RVs within 10 minutes. Spectral data (4000-00 cm-1) were collected, and the bio-fingerprint region (1800-900 cm-1) was employed for analysis. Standard normal variate (SNV) normalization and second-order derivation were applied to reduce scattering and baseline drift. Gradient-weighted class activation mapping (Grad-CAM) was employed to generate saliency maps, highlighting spectral regions most relevant to classification and enhancing the interpretability of model outputs. Two independent cohorts from Beijing Youan Hospital, processed with different viral transport media (VTMs) and drying methods, were evaluated, with one including influenza B, SARS-CoV-2, and healthy controls, and the other including mycoplasma, SARS-CoV-2, and healthy controls. The model achieved sensitivity and specificity above 94.40% across both cohorts. By correlating model-selected infrared regions with known biomolecular signatures, we verified that the system effectively recognizes virus-specific spectral fingerprints, including lipids, Amide I, Amide II, Amide III, nucleic acids, and carbohydrates, and leverages their weighted contributions for accurate classification.

Problem

• Develop rapid diagnostic system for respiratory viruses
• Combine ATR-FTIR spectroscopy with explainable machine learning
• Achieve high accuracy in virus identification

Innovation

• Combines ATR-FTIR with RoPE-SAT model
• Uses spectral bio-fingerprint for virus detection
• Employs Grad-CAM for interpretable classification