When I started this project, I was hoping to achieve three objectives that would enable us to:

1. Produce better diagnostic biomarkers using Machine Learning and Artificial Intelligence (AI) on magnetic resonance-based functional imaging.
2. Develop algorithms, which could automatically quantify metabolites that we measure with magnetic resonance spectroscopy in the brain, helping understand how brain tumours progress and to be able to clinically verify these findings.
3. Collect more data and undertake enhanced analytics that would allow us to improve our understanding and utility of imaging biomarkers in childhood brain tumours, aiming to integrate this information into clinical decision support tools that would assist clinical practice.

Characterising tumours with greater precision

Starting the project in 2004, with my partners at the University of Birmingham, we developed quantitative image analysis (including computational texture analysis on images) so that we can look at characterising tumours with greater precision and accuracy than afforded by the human eye.

We then combined this with in-vivo “biochemistry” information of a tumour to create a more comprehensive picture of the disease in individuals. This meant that, in addition to considering the typical features that an oncologist would usually take into account, such as the size and location, we were expanding the scope of analysis to include other factors such as the texture and density of tissue and the quantities of tumour-related metabolites, which helped improve the tumour classification at an earlier stage.

Historically, this type of diagnoses has been done through biopsies. However, these can be costly and carry a greater risk for young children, making an alternative, non-invasive approach even more important. With these imaging and metabolic biomarkers, identified through Machine Learning and AI, we can achieve this, while also building an enhanced imaging database , currently comprised of more than 1,500 patients in the UK.

Bringing together engineering technologists and clinical teams

While making a significant contribution to the field so far, as researchers, we are always hoping that more can be done. By bringing together engineering technologists and clinical teams, we have started to increase the interdisciplinary experience in this space – better integrating the human experience into medical technology and diagnostic imaging to improve patient outcomes. But where do we go from here?

We have seen how, through functional imaging and radiomics we can provide a faster and more accurate diagnosis of brain tumours in children, which has led to more personalised treatments for patients, ultimately improving chances of survival. However, technology is constantly evolving, and we don’t yet have the ‘perfect’ solution.

We have a strong starting point, but we want to continue to build our imaging databank to further enhance Machine Learning and make even more accurate and earlier diagnoses of brain tumours. We want to get to a point where clinical teams can support their patients earlier and don’t have to wait for diagnosis; a place where we have more information and are better equipped for challenging discussions with patients and their families so that more informed treatment decisions can be made.

Applying the technology to other types of tumours

Importantly, we’d also like to see this technology translated to other types of tumours in different parts of the body, such as prostate cancer, breast cancer and sarcomas. Compared to other types of cancer, paediatric brain tumours are rare and we have subsequently had to work with limited, and in some cases incomplete, data sets.

However, by using AI and Machine Learning, we’ve been able to start plugging these gaps to build more accurate diagnostic imaging which has the potential to positively impact other types of tumours where there are problems with the quality of data.

It’s an exciting innovation in the world of diagnostic imaging and we are eager to continue working with our research partners at the University of Birmingham and the Children's Cancer and Leukaemia Group to help build this approach and clinical support system into routine care.

Of course, blending engineering, science and medicine was going to have its challenges. However, WMG’s applied research and collaboration with the NHS and healthcare professionals provided a great environment to develop this research. Everyone involved in this project is united by one common goal and being able to drive this forward as the University is what has really enabled us to pursue truly transformative health innovations at scale.

Related article

• 'Characterisation of Paediatric Brain Tumours' case study