Classifying and localising future cancerous lesions

For many types of cancer, early detection is one of the most powerful tools to improve treatment success. An effective early cancer detection mechanism can be achieved through population based health screening programs, where clinicians generally use different types of medical image data (obtained from asymptomatic patients) to detect cancerous lesions, such as masses or polyps. An important research question is if it is possible to improve even more treatment outcomes with the detection of early disease processes from medical images before lesions become visible. To enable the answering of this question, we propose the development of new and sophisticated time-series forecasting optimisation methods that take longitudinal medical image data to predict the probability that a patient will develop cancer and to show the most likely image regions where the cancerous lesions will be localised.