Skip to main content Skip to navigation

Non-invasive visualisation of children tumours

Cancer is a leading cause of mortality in children, with the latest available statistics in the UK showing that between 2009 and 2011, an average of 1,574 children per year were diagnosed with cancer, of which 16% had died. Nevertheless, more children than ever are surviving childhood cancer; there are new and better medical technologies, drugs and treatments. But it remains devastating to hear that your child is diagnosed with cancer, and at times it can feel overwhelming. Professor Theodoros Arvanitis and doctoral researcher Ahmed Fetit of the Institute of Digital Healthcare (IDH) have been investigating the application of advanced computer algorithms that can provide intelligent decision support systems to aid clinicians with the characterisation of childhood brain tumours.

by Ahmed Fetit

When a child is suspected with a brain tumour, an MRI scan is usually undertaken to provide a detailed, high-quality visualisation of the inside of the brain. Initial tumour characterisation from the MRI scan is then done by an experienced radiologist. However, different brain tumours do not always demonstrate clear differences in visual appearance. Hence, using only MRI to provide a definite diagnosis could potentially lead to inaccurate results, so a biopsy sample is usually taken through invasive surgery, allowing tumour characteristics to be studied under the microscope.

The long-term goal of the work conducted by IDH researchers is to minimise surgery and improve decision support by using a technique called ‘texture analysis’. By analysing the texture of an MRI scan using advanced image analysis algorithms, quantitative measurements can be used to capture information that may be beyond human visual perception. Such information can be of potential clinical value and could support decisions made by radiologists. To ensure future clinical adoption of texture analysis, the researchers have demonstrated that the technique can be used for diagnostic classification of common tumour types, using over 120 clinical datasets obtained from three different hospitals across the UK.

A key collaboration with Professor Andrew Peet and Dr Jan Novak at Birmingham Children’s Hospital, and a large number of scientists and clinicians at Great Ormond Street and Nottingham University Hospitals, has allowed robust testing of the algorithms on clinical imaging data. This is important as it enables more information to be obtained from clinical datasets, which can be presented in a way that is more amendable to clinicians. Thanks to the Children’s Cancer and Leukaemia Group, such collaboration and analysis of multicentric datasets has been possible.