|Ph.D Student||Nakhleh Morad|
|Subject||Studying the Specificity of Disease`s Signature from|
Exhaled Breath: From Chemical Analysis to
|Department||Department of Nanoscience and Nanotechnology||Supervisor||PROF. Hossam Haick|
|Full Thesis text|
Volatile byproducts of diseases’ process usually diffuse into exhaled breath, altering the baseline spectrum of volatile organic compounds (VOCs) therein. In several diseases, the recognition of such VOCs in breath samples is used for non-invasive and safe diagnostic methods. The current study aims to explore the unique fingerprint of diseases in exhaled breath samples, as novel and easy-to-use in-vitro platform that enable diagnosis and follow-up of a large number of patients in a short time with a high precision as a point of care approach.
Over 800 breath samples were obtained from patients diagnosed with various systemic diseases, including chronic kidney disease, inflammatory bowel diseases, preeclampsia toxemia, pulmonary artery hypertension, tuberculosis, head and neck cancer, Parkinson’s disease, multiple sclerosis and others. Gas chromatography linked with mass spectrometry (GC-MS) analysis showed several remarkable changes in the VOC composition in each of the examined diseases, compared to the constituent healthy controls. Moreover, the sensors array analysis showed a clear discrimination not only between patients of each disease and the related healthy controls (binary phase), but also classification of each subtype of the disease.
A meta-analysis of all breath samples was conducted to investigate the possibility that each disease have a unique breath fingerprint. This analysis has shown 13 common VOCs that appear at distinctively different compositions in the various systemic diseases. Based on these findings, we developed an intelligent diagnostic tree-of-decisions in order to diagnose and classify the actual disease. The first part of the tree-of-decisions estimated whether a specific (blind) sample belongs to sick (including multiple diseases) or healthy subject with 90% sensitivity and 84% total accuracy. The second part discriminated between the different diseases with an accuracy of 86% in training phase and 78% in the blind analysis. The third, discriminated between a specific disease and the constituent control groups with accuracies that range between 78% and 98% (depends on the type of the disease). The same analysis showed the proposed approach is durable to common confounding factors (e.g., age, gender, ethnics, smoking, etc.)
The results presented in this thesis provide comprehensive evidence that different diseases have a unique breath fingerprint. Therefore a tailor made sensors array technology, sensitive to the specific breathprints could be developed and used as point-of-care diagnostic method. Combining different sets of sensors for each breath sample, an intelligent tree of decisions could be obtained for the diagnosis and classification of multiple internal and systemic diseases.