Bladder cancer is the seventh most common cancer in the UK (2011), fourth most common cancer in men and thirteenth most common in women. The total annual cost of bladder cancers in the UK was £55.39 million. The cost per patient was £8349 per year, including diagnosing, treating and following up, as well as indirect costs like loss of earnings etc. Cystoscopy with biopsy is the current gold standard for diagnosing bladder cancer. However, it is expensive and invasive, and has difficulties in diagnosing early stage cancer or carcinoma in situ (flat tumour). Non-invasive diagnosis of bladder cancer like urine cytology and urinary biomarker tests have poor specificity therefore they are used as supportive tests rather than definitive tests. The aim of this study is to develop a detecting system for bladder cancer urinary volatile organic compounds (VOCs) biomarkers and explore its application in non-invasive diagnosis of bladder cancer. Urinary VOCs are a class of chemical compounds found in urine that generated from biological activities of cells, cancer cells have special metabolic alternations compare to normal cells, therefore having different VOCs profiles that can be used for discrimination. Compare to conventional mass-spectroscopy and e-nose techniques for VOCs detection, the fluorescence sensor array has good detection accuracy and much lower cost of use, it is especially ideal for developing non-invasive point-of-care device for bladder cancer diagnosis and surveillance purposes. This study successfully developed a fluorescence cross-response sensor array system for diagnosing bladder cancer by detecting the urinary VOCs. On this system, a distinguishing test of four urinary VOC biomarkers: ethylbenzene, hexanal, lauric aldehyde, and nonanoyl chloride, was undertaken and achieved a sensitivity of 77.75% and a specificity of 93.25%. In a proof-of-principle clinical trial involving 79 participants (38 bladder cancer patients and 41 healthy controls), this system using a PLSDA model successfully identified over 80% of urine samples (86.08% with leave-one-out, 81.76% with Mote Carlo cross-validation) with 77.42%-84.21% sensitivity and 85.82%-87.80% specificity. This study revealed the possibility of using low-cost optical sensor system for medical diagnosis purposes, and this could inform larger scale multi-centre trails of biogenic VOCs diagnosis and other possible clinical applications in the future.