Using Artificial Intelligence and data
This study focuses on tailoring rectal cancer treatment with an artificial intelligence that learns from multiple hospitals’ data without data ever leaving those hospitals. The study will use privacy-preserving distributed learning of a Bayesian network for prediction of pathological response to chemo-radiotherapy in patients with rectal cancer.
Research project summary
In rectal cancer, the best treatment option at each point in the treatment pathway might change from patient to patient.
Cancer treatments such as surgery, chemotherapy and radiotherapy might cause severe side effects in some, whilst others do not experience them. Hence, for some aggressive therapy will be beneficial, whilst others will be better off with a less invasive treatment.
The researchers aim to build software that learns which treatment is best for each patient from the information of patients who have been in a similar situation. Unfortunately, this information is different for each patient and the information from different patients is stored in different hospitals that do not want to share their data.
Building a smart computer programme
In this project, the researchers propose to build a smart computer programme that will tell us how likely it is for a certain patient with rectal cancer to benefit from an extra radiotherapy dose and thereby avoid surgery.
The programme would learn to calculate this probability based on what has happened to similar patients in many hospitals. They propose to do this without patient information leaving the hospitals, by sending the smart programme to the hospitals to learn instead.
This PhD research will be undertaken by Akuli-Biche Osong, supervised by Dr Bermejo, Postdoctoral researcher, at Maastro Clinic in Netherlands.