Picturing Predictions for Patients with Brain Tumors

Gliomas are the most common form of primary brain tumors. The treatment of glioma is about finding the right balance between adequate treatment of the tumor, while preserving the surrounding, sometimes infiltrated, healthy brain. Considering the spatial variation in brain functionality, the location of the tumor plays an important role determining the feasibility of extensive treatment. 

In this project we leverage the power of AI in the form of deep learning to extract valuable information from large datasets, including the location of the tumor and radiological properties on the MRI scans. The models developed in this project will help better inform clinicians to select the treatment best suited for their current patient, based on the experience of treatment decisions made in the past for similar patients. 

Furthermore, the models will help better inform patients on the expected outcome of the different treatment options. By using probabilistic models the uncertainty of predictions can be quantified, an important feature to better valuate the model’s predictions and combine the model’s prediction with other clinically relevant information and the patient’s preferences. 

The software developed in this project will be implemented in an interactive web application. This application helps to make these complex models and software accessible to a wider audience, giving patients all over the world access to the best available treatment. 

Next to optimizing treatment decisions for the individual patient, the software will also contribute to the development of new treatment standards. We will enable experts to self-evaluate and compare their treatment decisions to the decisions made by other experts. We will help identify treatment variations and analyze their associated outcomes by providing experts with the tools to interactively and intuitively analyze and visualize the aggregated data of well over a thousand patients.