Lo Sasso A., Bellantuono L., Facchini L., Amato E., Porcelli F., Frascella P., Bellotti R.
Prostate cancer stands as the most prevalent cancer among males in Italy, comprising 18.5% of all male cancer diagnoses. Nearly a quarter of men receive a prostate cancer diagnosis between ages 50 and 60. Detecting this cancer early could potentially save numerous lives.
Our research endeavors to investigate cancer onset by analyzing breath samples. By examining the abundance of approximately 250 Volatile Organic Compounds using gas spectrometry, we have studied patients afflicted with prostate cancer.
Following promising results from initial pilot studies on suspected and full-blown patients, we have progressed to standardizing this procedure for screening analyses across various hospital structures.In this study, we elucidate why relying solely on classical statistics struggles to effectively discern a patient disease state. We present a clustering model based on Self-Organized Maps (SOM), which aids in data labeling and error reduction in screening. Subsequently, we explain an artificial intelligence model capable of distinguishing patients categorized into three classes: No DiseasePatients, Suspected Patients, and Full-Blown Patients. We also discuss the algorithm’s performance with these patient groups.
