Design of a radiomic software to address an unmet clinical need: the early diagnosis of pancreatic tumor

Pancreatic cancer is one of the most aggressive tumors: both its incidence and its mortality rate are increasing over time. Nowadays, in the clinical practice there is essentially no certified device aimed at guiding clinicians in the early diagnosis of pancreatic tumors. The unmet clinical need arises precisely from the observation of the images detected by CT, MRI and other acquisition techniques: even if today medical imaging technologies are increasingly precise and efficient in faithfully reproducing the patient's clinical condition, it is not always possible to define with certainty the diagnosis of patients affected by pancreatic pathologies based only on observation. The problem is strictly connected not only to the diagnosis, but also to all further decisions based on it, and consequently to the therapeutic path of the patient and to its prognosis, which can vary significantly. Starting from the daily clinical practice, which highlighted the urgent need of an early and precise diagnosis of pancreatic tumors, passing through the study of radiomic features, extracted from medical images obtained from traditional acquisition methods, the aim of the thesis is to outline the fundamental characteristics and necessary requirements needed, from a clinical and regulatory point of view, in order to correctly design a radiomic software based on artificial intelligence, intended for clinical decision support in the pancreatic oncology field. The potential benefits arising from a radiomic software analysis can be achieved through multi-disciplinary collaboration for the development and demonstration of clinical evidence, to ensure the safety and effectiveness of the medical device software, who's applications could become increasingly indispensable for providing answers in clinical contexts of high complexity and uncertainty.