Abstract
Identifying noninvasive specific disease biomarkers that can provide valuable information about chronic obstructive pulmonary disease (COPD) progression, potential complications, or treatment response is of paramount importance. In this study, we propose the validation of an innovative near-infrared (NIR) device that utilizes near-infrared light reflectance techniques combined with data validation through a convolutional neural network for the detection of substance P in non-invasive saliva samples. We conducted an analytical observational cross-sectional study at a leading university hospital between January and March 2022, including patients with COPD and controls without the disease. Following the collection of clinical data, a saliva sample was obtained for the determination of substance P which was analyzed both by the NIR device and an Enzyme-Linked Immunosorbent Assay. Direct comparisons were made, and Bland-Altman plots were constructed to assess the level of agreement between the two measurements. The sample consisted of 102 subjects, 44 with COPD and 58 controls. The average differences between the two measurement methods yielded similar results with no significant differences between them, showing a value of 110.2 (16.1) pg/ml for the NIR device and 110.5 (16.7) pg/ml for the ELISA determination (p > 0.05). The Bland-Altman plots show a small difference and a level of agreement consistent with good measurement by the NIR device. The results of this study validate the efficacy of a NIR device combined with a convolutional neural network for detecting substance P in the saliva of COPD patients.