Analysis of urinary EVs for the development of AI-assisted application for prostate cancer detection

Prostate cancer screening often relies on PSA tests, which can give false alarms or lead to unnecessary treatments. This thesis tested a new, non-invasive method using nanosized particles from urine, called extracellular vesicles (EVs), analyzed with a special light-based technique, known as FTIR spectroscopy. Combined with patients’ health data and machine learning, this approach greatly improved detection accuracy. In tests with 97 men, the best model identified cancer with 90% accuracy and outperformed PSA tests alone. This shows that combining urine-based FTIR signals with patient data and machine learning could make prostate cancer screening more reliable and less invasive.