Commercial Feasibility of Deep Learning based Medical Image Registration

Various methods of imaging are used within medicine to provide different information for the diagnosis and treatment of various diseases. Image Registration is a valuable tool that establishes an alignment of images to one another and, as a result, permits meaningful comparison and visualisation of multiple images. For example, a patient’s response to treatment can be assessed by overlaying (fusing) images captured before and after treatment. Alternatively, anatomical and functional information can be displayed together, providing more information than if the two were viewed separately. For these uses and many others, such as automated quantification of certain statistics, an accurate registration is clearly important. While a number of registration schemes are currently available for use in clinical routines, this project aims to develop a new and more effective scheme.