Medical imaging aims to produce pictures of the body that reveal information about its normal internal structure and mechanisms, as well as changes that occur in disease (e.g. figure illustrates tracing fibre direction using diffusion MRI). Most medical imaging techniques work by solving an inverse problem that relates properties of the tissue, such as x‐ray absorption or water content, inside the body to measurements that an imaging device makes outside the body. For example, a magnetic resonance imaging (MRI) scanner measures the relative strengths of magnetic fields oscillating at different frequencies and infers an image of water density, which provides contrast between different types of tissue. Automated image processing methods are increasingly responsible for extracting useful information from such images and highlighting areas that may be abnormal or indicate disease.

For example, one of the most important medical image computing problems is registration: to compare two brain scans of the same patient at different times, we need to remove differences in the images that arise from different positioning of the head inside the scanner. Image registration aligns one image to the other so we can subtract them and look for changes.

Another key challenge is to compute reliable biomarkers. A biomarker is a feature of the image that shows the presence of a disease or the effect of a treatment. They may be very simple features, such as the image intensity in a certain region of the brain, or require sophisticated computations and comparisons with atlases, such as the degree of atrophy in grey matter of Alzheimer’s patients compared to the average brain. Modern medical imaging combines knowledge of biology, medicine, imaging devices and image analysis techniques to identify the most descriptive biomarkers.

BACK TO CGVI OVERVIEW