| Popis: |
The ‘smell’ of illness, disease or age has been known for many centuries, mainly created by volatile organic compounds (VOCs). Dogs were first reported to detect cancer in 2004. Increasingly, the profiles of VOCs are being utilised as non-invasive diagnostic methods. The aim of the thesis was to develop and evaluate the performance of analytical methods to characterise the disease states of patients utilising selective discrimination, gas chromatography-mass spectrometry (GC-MS) and chemometrics. The primary analytical technique investigated was GC-Time-of-Flight-MS coupled with headspace solid-phase microextraction (HS-SPME-GC-ToFMS). A robust and sensitive method was developed by optimisation of all sample analysis parameters and was applied to clinical samples from bladder and prostate cancer patients and those with hepatic disorders. This evidence was obtained by quantifying an internal standard, present in every sample and blank throughout the studies. Based on these findings, large numbers of clinical samples were analysed with confidence. Statistically significant mathematical models were developed in partnership with Cranfield University to classify the diseased state of samples and clinically relevant controls. PLS-DA was determined as the best classifier. The results from the HS-SPME-GC-ToFMS studies were highly promising. Bladder cancer gave a mean accuracy of >80 % and even low-grade tumours gave a sensitivity of 73 %, superior to urine cytology. Higher clinical performance was obtained in the prostate cancer study, with BPH distinguishable from cancer. Hepatic disorders were better again (>86 %). Preliminary studies on sepsis detection also showed promise. Several recommendations were made to enable significant clinical results in the future based on analytical rigour. |
