An investigation into the fabrication of a new mini gas sensor for medical diagnostic applications

This research was part of a larger Wellcome Trust funded project that aimed to create a non-invasive volatile diagnostic device for medical applications, particularly for the detection of Clostridium difficile (C. Diff) in human stool. The device (OdoReader) consists of a gas chromatography column to separate gaseous mixtures, a metal oxide sensor to detect and respond to individual analytes and an artificial neural network trained to recognise patterns indicative of positive and negative results. The aim of this MPhil research project was to fabricate the metal oxide gas sensor to work within this device and to conduct a pilot study to assess the potential for using OdoReader to detect urinary tract infections human urine. Sensors that were compatible with existing electronic components were not commercially available, making it essential to fabricate sensors in-house. A preliminary investigation into sensor fabrication techniques was conducted in order to assess the most simple and cost-effective process to fabricate the new sensor. Screen printing was selected and used to deposit conductive pastes onto alumina substrates to produce interdigitated electrodes and a heater element. Three sensor designs were subsequently produced, two designed to be housed in edge connectors (type A), and the third to be mounted onto a transistor can (type B). Initial tests showed that type A sensors would not withstand the operating temperature required (450˚C). Type B sensors required an extra process to join metal wires from the sensor substrate to the transistor can. This process consistently created strong mechanical connections that enabled the sensor to achieve and maintain the required operation temperature. For this reason this sensor and transistor can combination was selected. In order to give the sensor substrate its sensing properties, the deposition method of a metal oxide paste (to be applied over the interdigitated electrodes of the sensor substrate) was investigated. An even layer of the metal oxide paste was of extreme importance for this type of sensor and for this reason screen printing was one of the methods considered. For these trials, a custom built screen printing jig (for manual operation) was designed and fabricated. These trials proved unsuccessful for our purposes therefore a doctor blading technique was adopted. After some development this method satisfied the requirements for the sensor and has been used to coat over 50 sensor substrates. The mean electrical resistance (for 25 sensors) of the metal oxide coating was 116000Ω, S.D.38042Ω. The mean resistance of the heater element (for 26 sensors) was calculated at 10.8Ω at 0°C, S.D. 0.23Ω.The sensors met OdoReader sensitivity specifications for ≥150% to ethanol (50ppm) with a mean sensitivity (for 25 sensors) of 216% and a S.D. 46%.As this sensor was designed for medical diagnostic applications, a pilot study was conducted to assess the potential for the detection of bacteria in human urine and the identification of the urinary tract infection (UTI) causing bacteria through volatile organic compound (VOC) analysis. The sensor developed here was used in combination with a gas chromatography (GC) column (OdoReader) to assess urine inoculated with known bacteria. Five of the most common UTI causing bacteria were used to inoculate the stock urine and left to grow, with samples taken at hourly intervals for VOC analysis. The bacteria chosen for this study were; K. pneumoniae, S. saprophyticus, P. aeruginosa, E. faecalis and P. mirabilis. VOC analysis was undertaken using infected urine that had not been chemically treated and also infected urine treated with a sodium hydroxide solution. Using OdoReader the detection of bacteria in human urine (at UTI levels) was possible by looking at the retention times and percentage responses of significant peaks. It was also found that there is the potential to differentiate between bacteria by looking for the presence of peaks at certain retention time ranges