Non-invasive spectroscopic analytical molecular neural network-based profiling of saliva and endoscopic tissue biopsy for inflammatory bowel disease diagnosis

Current gold-standard for diagnosing and monitoring inflammatory bowel disease (IBD) is a costly, invasive and painful procedure. Recently, saliva analysis been emerging as a non-invasive biofluid due to its correlation with blood constituents at a molecular level. Herein, we developed an analytical system and validated the bio-diagnostic performance of Raman spectroscopic molecular profiling of saliva compared to colonic tissues in patients with ulcerative colitis and Crohn's, sensing biomarkers to rapidly detect and monitor IBDs. Via a combination of Raman spectroscopy and an advanced self-optimising Kohonen index Network (SKiNET), a non-destructive molecular profiling tool enabled rapid biosensing of salivary and colonic biopsy tissues. SKiNET was used to identify spectral markers and discriminate healthy and IBD classes. Spectral signatures were validated for 91 saliva and 46 tissue biopsy clinical samples, which were correlated with pathology markers and endoscopic evaluation. This yielded 87.5 % sensitivity and 89.0 % specificity for saliva and 83.9 % sensitivity and 84.6 % specificity for biopsies. Comparing the fingerprints of saliva to tissue biopsy for UC, similar spectroscopic signatures were found with equivalent intensities at primary peaks of 1003 cm⁻¹, 1445 cm⁻¹ and 1656 cm⁻¹ and for CD, significant variance throughout the spectral region with intensity changes observed at each of the primary peaks and additional shoulder bands. Raman spectroscopic biosensing of saliva is a reliable, non-invasive tool for diagnosing and differentiating UC from CD. The unique combination of analytical spectroscopy and advanced AI lays the platform towards improved, non-invasive and rapid detection of IBD and further in-situ disease diagnosis and monitoring.