Using Artificial Intelligence to fight pediatric cancer

There is significant interest in the use of image processing techniques and machine learning to improve automation and facilitate clinical decision making. With our partners at Alder Hey Children’s Hospital, a Centre of Excellence in Cancer Treatment, we are working to automate the measurement of children’s brain tumours based on images of the brain produced by scanning machines such as Magnetic Resonance Imaging (MRI).

It is important to be able to accurately measure the size of a tumour over time when evaluating disease progression and informing treatment, but is typically a time-consuming process for clinicians and difficult to accurately reproduce.

There is much interest in using Artificial Intelligence (AI) / Machine Learning techniques to automatically recognise brain tumours within MRI images [1], but this is still a rapidly developing field which has primarily been focused on data from adults. Brain tumours in children are strikingly different to those in adults [2], and algorithms will require specialised training to identify the boundaries of different types of tumour, and how their volumes are changing over time, when the brain may also be simultaneously developing and growing rapidly.
One commonly used technique is to apply AI algorithms, such as Convolutional Neural Networks, which are given a set of ’training’ data - existing tumour scans which have been manually annotated by experts - and can automatically derive a computational model to classify new data. We anticipate that similar approaches can be used, making use of the image data and expert analysis provided by Alder Hey.

In this project we are developing software, hardware, and network infrastructures that will make it possible to automate the process of measuring brain tumours from MRI data so that it can be used directly by clinicians at Alder Hey. This will involve being able to successfully research, apply and optimise existing AI solutions to this new data, as well as developing new software that will allow the data collection, cleaning, and transfer processes to be automated. 



Back to: Computational Biology Facility