Pancreatic cancer and AI: Predicting the risk of post-operative complications through machine learning
A project that, thanks to the synergy between the clinic and Artificial Intelligence, aims to predict the probability that a patient will encounter complications following pancreatic surgery, based on pre-operative CT images.
Humanitas is a one of the leading centers for the treatment of pancreatic cancer, with several active research projects. New technologies can be of great help in the field of research. Surgery for pancreatic cancer is very complex and is burdened with serious post-operative complications. For this reason the analysis of risk factor data through Artificial Intelligence programs is a crucial element in approaching patients with these tumors.
Thanks to clinical research, it has been possible to identify risk factors for the most insidious post-operative complication (pancreatic fistula). There are two physical aspects of the patient’s pancreas – the consistency of pancreatic tissue and the diameter of the pancreatic duct. Based on pre-operative CT scans, experienced surgeons and radiologists are able to predict these two variables. In addition, CT images can predict the presence of sarcopenia, an index of malnutrition and fragility, also related to the appearance of post-operative complications.
Aim of the research
We aim to be able to better predict the complications in order to evaluate the probability of their appearance and seriousness in the individual patient, to implement measures to limit them or, if necessary, even to postpone the intervention and promote a radiotherapy or chemotherapy treatment. The access to a large number of preoperative CT scans would allow the development of a tool capable of making the patient’s parameters objective in an automatic way. This process would lead to a score correlated to the surgical complexity and fragility of the patient. Therefore, from the preoperative CT scan, the outcome of the surgical procedure could be predicted and the risk of post-operative complications assessed.
Since pancreatic cancer is relatively uncommon, the number of patients does not allow for traditional imaging. To face this challenge, we will try to use Artificial Intelligence, through machine learning models.
Humanitas A.I. Health Center combines the network of doctors and researchers and a team of engineers specialized in Artificial Intelligence. 3 surgeons and 2 radiologists are currently involved in this project, as well as a strong team of data scientists and medical engineerings.
The medical skills and the expertise of the data scientists are integrated during weekly meetings where the focus is on the aspects related to the identification from CT parameters to be evaluated by radiomic technique, based on the clinical suggestions, the characteristics of the images and their processing possibilities.