Human volatile organic compounds and sensorial recognition for medical diagnosis

Detection of volatile organic compounds (VOCs) is a new frontier in rapid, accurate, and non-invasive medical diagnosis. This thesis is focused on the detection and the analysis of VOCs from bodily samples to diagnose medical pathologies. An electronic nose, as an artificial olfactory instrument that mimics the mammal olfactory system, is involved to recognize the volatile signature of health status. Thus, the sensing properties of Quartz microbalance (QMB) sensors coated with metalloporphyrins/corroles are employed with the pattern recognition algorithm. We also used gas chromatography-mass spectrometry (GC–MS) as a gold standard to recognize and evaluate the variation of detected VOCs between healthy patients and sick people. The protocols from collection to the measurement of VOCs from urine, serum, and exhaled breath, have been designed, implemented, optimized for safe and repeatable VOCs analysis. Clinical trials for colon, pulmonary, kidney cancers and infectious diseases as Tuberculosis and Covid-19 have been done. These metabolomic investigations are the first in the published scientific literature for kidney cancer and Covid-19 investigation through an electronic nose with urines/sera. And a scientific enrichment for the sensorial diagnostic of lung cancer based on urinary headspace. For Tuberculosis, the performances reach by the sensorial recognition of the disease and in the case of HIV co-infection stand beyond the actual non-invasive and nonsputum-based tests performances. A set of VOCs markers have been identified for all these pathologies except Tuberculosis and colorectal cancer, while the volatile pattern has been released for all these 5 diseases across different samples and patient conditions with an accuracy between 88% to 97%. Statistical machine learning algorithms have been prioritized to get more explicability among the metabolites and the different chemical-sensitive layers involved. As an outcome, the gas sensor array combined with these pattern recognitions offers accurate diagnostic in a brief time and safe mode. Further, to each disease and thanks to GC-MS analysis, more precise chemical layers for the sensors could be chosen to well map the class of VOCs related to each disease.