A rapid diagnostic approach for COPD utilizing multimodal serum spectra integrated with machine learning algorithms.

Chronic Obstructive Pulmonary Disease (COPD) is a major global respiratory illness causing death and disability. Traditional methods lack consistent standards, often miss diagnoses, and cannot explore diseases' molecular relationships. Thus, there is a need for a diagnostic method that is both efficient and convenient. This study aimed to evaluate the potential of diagnosing COPD, Non-COPD (Pulmonary infection), and Healthy Group using serum fluorescence, Raman, and surface-enhanced Raman spectra (FS, RS, and SERS) algorithms combined with eight machine learning algorithms. The experiment reveals variations at each peak by examining the serum FS, RS, and SERS of COPD patients compared to the control group. The combination of serum RS or SERS with machine learning algorithms provides superior classification results compared to serum FS. Serum SERS and machine learning algorithms classify COPD and healthy individuals with over 0.98 accuracy. Serum SERS combined with the synthetic minority over-sampling technique (SMOTE) -gradient boosting (GB) algorithm achieves a three-classification accuracy of 0.84. In summary, the integration of serum SERS with SMOTE-GB machine learning techniques showed significant promise for COPD detection.