A machine learning approach to automated targeted analysis of raw gas chromatography-mass spectrometry data
A machine learning approach to automated targeted analysis of raw gas chromatography-mass spectrometry data
Angelika Skarysz* (presenting author), Yaser Alkhalifah*, Kareen Darnley§, Michael Eddleston¶, Yang Hu*, Duncan B McLarenǁ, William H Nailonǁ, Dahlia Salman†, Martin Sykora‡, C L Paul Thomas† and Andrea Soltoggio*
* Computer Science Department, Loughborough University, Loughborough, UK † Centre for Analytical Science, Loughborough University, Loughborough, UK ‡ Centre for Information Management, Loughborough University, Loughborough, UK § Clinical Research Facility, Western General Hospital, NHS Lothian, Edinburgh, UK ¶ Pharmacology, Toxicology & Therapeutics Unit, University of Edinburgh, Edinburgh, UK ‖ Edinburgh Cancer Centre, NHS Lothian, Edinburgh, UK
Full Poster:
Abstract:
The human breath carries hundreds of volatile organic compounds (VOCs), that comprehensively describe health conditions and can reveal pathologies. Thus, breath analysis has a potential to deliver a fast and non-invasive diagnostic platform. Gas chromatography-mass spectrometry (GC-MS) is an analytical method used to measure compounds in the exhaled air. The identification of the VOCs in GC-MS data requires labour-intensive and time-consuming preprocessing and analysis by domain experts. Presented research study explores the original idea of applying supervised machine learning to learn specific ion patterns of VOCs directly from raw GC-MS data. The proposed method adapts to machine specific characteristics, and once trained, can quickly analyse raw breath samples to automatically detect the target compounds, bypassing the time-consuming preprocessing phase. All considered machine learning tools: convolutional neural networks (CNNs), shallow neural networks (NNs) and support vector machines (SVMs) achieved high accuracy in experiments with clinical data. In particular, the CNN-based approach detected the lowest number of false positives. The results indicate that the proposed method is a promising tool to improve accuracy, specificity, and in particular speed of the detection of VOCs in large-scale deployment of breath-based diagnosis. Poster presents results published in SKARYSZ, A. et al, 2018. Convolutional neural networks for automated targeted analysis of gas chromatography-mass spectrometry data. IN: 2018 International Joint Conference on Neural Networks (IJCNN), Rio de Janeiro, Brazil, 8-13 July 2018.
Comments
Hi Mauro, thank you for your interest in Breath Biopsy. We have contacted you directly by email (Subject: Breath Biopsy-Owlstone Medical) to discuss you project further.
Is this product for sale? For years I thought that through the nose you could diagnose cancer, epileptic attacks and other diseases, today it occurred to me to look for and found this sensor. Is it possible to program it? What is the cost?
Please sign in to leave a comment.