Wicaksono, Alfian (2021) Diagnosis of human disease by odour analysis employing machine learning. PhD thesis, University of Warwick.

Preview

PDF WRAP_Theses_Wicaksono_2021.pdf - Submitted Version - Requires a PDF viewer. Download (4Mb) | Preview

Request Changes to record.

Abstract

The scientific community has long been intrigued by the potential use of human waste odour in medical diagnosis and disease monitoring. Human waste odour analysis offers a fast and non-invasive way of testing and has the potential as an early disease diagnostic method. Cancer (particularly pancreatic and colorectal cancer), which has been known to be one of the deadly diseases of all time due to the late prognosis, may benefit from odour analysis as the current diagnostic method using highly invasive procedures such as colonoscopy. Other diseases, such as inflammatory bowel disease (IBD), which also employed colonoscopy in its diagnostic procedure, could also benefit from the application of human waste-based investigation.

In this work, data analysis pipelines employing five machine learning algorithms were developed for analytical instrument electronic nose (E-nose), field asymmetric ion mobility spectrometry (FAIMS), gas chromatography ion mobility spectrometry (GC-IMS), and gas chromatography time of flight mass spectrometry (GC-TOF-MS). These methods were used in 8 studies to investigate the potential of non-invasive diagnosis and monitoring of pancreatic cancer (PDAC), colorectal cancer (CRC), and IBD through urine, breath, and faecal odour analysis.

The PDAC studies included 285 subjects: 126 with PDAC, 45 with Chronic Pancreatitis (CP), and 114 controls, and utilised FAIMS, GC-IMS, and GC-TOF-MS to distinguish between groups based on the urinary VOC. All of the technologies were consistently able to distinguish between CRC and healthy control with an area under the curve (AUC) greater than 86%). FAIMS was also able to differentiate between early-stage PDAC from healthy and from late-stage PDAC (AUC 89% and 92%, respectively). However, urinary VOC analysis couldn’t differentiate PDAC from CP (AUC 0.58). Further studies investigating urinary and faecal VOC of patient for detecting CRC using FAIMS and GC-IMS included 728 subjects which consisted of 26 patients with CRC, 340 with adenoma, 32 polypectomy patients, 296 healthy control and 33 with other gastrointestinal diseases. Both FAIMS and GC-IMS methods were able to separate between CRC and healthy control with AUC above 82%. Patients with adenoma could only be separated from healthy control in faecal odour analysis. Interestingly, odour analysis showed that the faecal VOC profile of patients whom underwent polypectomy returned to its physiological state after three months of polypectomy. This demonstrated the potential of faecal VOC analysis for disease monitoring, especially in adenoma and CRC. Final studies investigated the urinary, breath, and faecal VOC of patients with IBD. Six hundred forty-one subjects in total were recruited: 350 IBD, 276 healthy control, and 15 irritable bowel syndromes (IBS)/Functional Abdominal Pain-Not Otherwise Specified (FAP-NOS). All three types of human waste were consistently able to detect IBD, both Crohn’s Disease (CD) and Ulcerative Colitis (UC), from healthy control. Faecal odour analysis was also able to separate between IBD patients with IBS/FAP-NOS, which share many common symptoms.

In this work, the potential of utilising urine, breath, and faecal odour-based methods was successfully demonstrated for the non-invasive diagnosis of PDAC, CRC and IBD. The potential use of odour analysis in healthy monitoring was also shown in PDAC patients and patients undertaking polypectomy. Further work in this area will need to focus on sampling and storing samples as well as machine learning approach in order to speed up the advances in the application of modern E-nose technology in a clinical setting. Furthermore, a large-scale multi-centre study involving patients with similar symptoms will push further the knowledge of human waste odour for diagnosis.

Item Type:	Thesis (PhD)
Subjects:	Q Science > Q Science (General) Q Science > QP Physiology R Medicine > RC Internal medicine R Medicine > RC Internal medicine > RC0254 Neoplasms. Tumors. Oncology (including Cancer)
Library of Congress Subject Headings (LCSH):	Odors -- Analysis, Diagnosis, Feces, Organic compounds, Machine learning
Official Date:	September 2021
Dates:	Date Event September 2021 UNSPECIFIED
Institution:	University of Warwick
Theses Department:	School of Engineering
Thesis Type:	PhD
Publication Status:	Unpublished
Supervisor(s)/Advisor:	Covington, James A., 1973-
Sponsors:	Lembaga Pengelola Dana Pendidikan
Format of File:	pdf
Extent:	xxii, 199 leaves : colour illustrations
Language:	eng

Request changes or add full text files to a record

Repository staff actions (login required)

View Item

Downloads