Medical Image Computing

Computing for Diagnosis and Therapy

Adapted from an article requested by Postgraduate UK.

Medical Image Computing is an exciting, interdisciplinary area of research and the UK has some of the largest medical imaging research groupings in the world. In January 2005, a new Centre for Medical Image Computing (CMIC) was established at University College London, which combines excellence in medical imaging sciences with innovative computational methodology. The new group is one of the strongest in the world and will launch a new MSc in Medical Image Computing in 2007. The course director, David Atkinson, reports on medical image computing and how it is being used to benefit medical diagnosis and therapies.

Within a few months of the discovery of X-rays by Wilhelm Röntgen in 1895, X-ray images were used to plan and guide a surgical intervention. In the field of medical imaging, the themes of innovation in diagnosis and treatment have continued over the past century and the UK has gained two Nobel prize winners; Godfrey Hounsfield for the development of computer assisted tomography (CT) and Peter Mansfield for discoveries concerning magnetic resonance imaging (MRI). Hounsfield’s work used data from X-rays taken at different angles to perform tomography – the calculation of image slices through the body. MRI also produces three dimensional data but is based on magnetic fields and radio waves rather than X-rays. MRI and CT are complementary - each providing images sensitive to different types of tissue.
In addition to MRI and CT, we also have ultrasound imaging based on the reflection of sound waves and positron emission tomography (PET) and nuclear medicine that detect the decay of radio-isotopes. These isotopes are added to compounds that concentrate at sites of specific tissues or diseases enabling them to be visualized in imaging. Furthermore, medical images come from surgical microscopes, endoscopic examinations, photographs and video as well as emerging technologies such as optical tomography whereby light is shone through the head at multiple angles. This wealth of data brings new information and great opportunities to apply computing methods for enhancing diagnosis, helping plan, guide and assess therapies and for the investigation of fundamental questions about the working of the human body.

Computers in Imaging

Acquisition systems produce digital data that is converted to images and ultimately viewed by radiologists and surgeons. Present throughout this chain are computers, not just as convenient tools for storage, workflow and display, but also in the control of the scanning, the alignment of images, detection of different tissue types, calculation of quantitative measures, generation of new images such as maps of active regions in the brain and surgical guidance. Increasingly the hardware of scanners is under software control and we have the opportunity to produce algorithms that will intelligently enhance the image acquisition by responding in real time to patient data.

A key tool for generating new information is image registration, a field in which the UK has been very active in research. Registration algorithms compute the transformation needed to align, or warp, one image into another. With the ability to align images, comparisons between images taken on different days can reveal small changes in tissue size, for example the slow shrinkage of the brain over many years due to dementia. Finding the transformation from one image to another can be used to measure motion, for example a cine series of cardiac images can show abnormalities in heart wall motion due to tissue damaged by a heart attack. In another application of registration, images from different patients are aligned into one common space. A representative image or atlas can then be computed, against which new patients can be compared. Registration is used in the fusion of data from different scanners leading to enhanced understanding, for example, a PET image is sensitive to the uptake of glucose by tumours and an MRI image can show other anatomy in detail - combining the two provides a richer source of information to guide patient care.

The UK is especially active in research into the workings and connectivity of the brain. Brain activity and thoughts cause changes in blood flow and oxygenation that show up as subtle differences in MRI scans. After computer processing involving registration, segmentation of relevant structures and statistical analysis, maps can be made revealing brain regions that are active during various tasks. These scans and processing are called functional MRI. Diffusion weighted MRI provides an image contrast that is sensitive to cellular architecture. Using this technique, we can now infer the directions of fibre bundles within the brain; for example, we can see how the spinal column is connected to the motor cortex.) In addition to enhancing our fundamental understanding, this research has the potential to guide brain surgery in order to avoid severing crucial nerve connections.

Obtaining quantitative measures is vital for enabling patient information to be compared with existing knowledge or to assemble new measures of anatomy and physiology. UK researchers have developed methods for modelling shape to study structural and functional variation in health and diseased states. Research is also using shape information to find organs and bones in images and to guide surgery.

In surgery and interventions where a catheter is inserted through a blood vessel into the heart, images from previous scans can be used to help guide the operator during the procedure. Registration techniques, and the 3D tracking of equipment, are used to visually present images taken prior to the procedure, which are overlaid on the new surgical scene. The world’s first MRI guided cardiac catheter intervention took place recently at Guy’s Hospital, King’s College London.

Patient movement resulting from involuntary motion, cardiac pulsation, respiration and flowing blood can all blur MR images. Novel algorithms for correcting images are being developed at Imperial College London and UCL to aid diagnosis.

Medical and Technological Drivers

Faster scans at higher resolution and with better image quality are always in demand. Our enhanced understanding of the genome is driving a desire to observe changes at the molecular scale whilst wanting to consider the patient as a whole. A dream goal might be the biological equivalent of Google Earth that enabled zooming from a whole body image down to the cellular and then molecular levels.

Detector technology is improving and providing ever more data. In MRI, the numbers of coils used to receive the signal has increased by an order of magnitude. In CT, the detectors that acquire the X-ray signal at each angle now have many more elements, generating much more data. In PET imaging, modern scanners are now combined with a CT scanner and in ultrasound, micro bubbles injected as contrast provide harmonic signals in addition to the main signal. 

These large quantities of data present challenges for computer algorithms and are very demanding of memory requirements. The increased availability of 64-bit machines and the ability to connect large numbers of PCs to form a cluster are addressing these issues. For example, the CMIC group at UCL have their own 60 node, 64-bit cluster and there are e-science projects that use a national grid of computers. One e-science project called IXI is collecting brain images from 600 people at three different sites. The aim is for a doctor to be able to see at a glance the normal range of sizes and shapes of each brain structure, overlaid on the patient’s own scan, assisting diagnosis. To provide this information, the grid will access data that may be stored in distributed places and perform the necessary calculations on machines located anywhere on the grid network. Just like the electricity grid, the user draws resources without concern for where they are generated.

Imaging in clinical trials

Drug discovery and development can bring major advances in the treatment and management of disease. The costs from discovery to launch of a successful drug are estimated to be in the region of £1 billion when failures of other drugs are factored in. There is a big incentive to gauge the effectiveness of a trial drug early. In studies of dementia such as Alzheimer’s, disease progression can take decades and computing and imaging techniques are being developed to quantify small changes in brain volume early in a trial to predict outcome. This use of images to indicate underlying biology is termed “biomarkers” and can save time by halting trials early. The endpoint of a standard clinical trial may require waiting decades to observe complete disease progression, there is hope that imaging might act as a surrogate endpoint whereby a drug can be assessed more quickly thus saving money and enabling the drug to be made available sooner.

Thriving Industry

University research groups have started to spin-out companies to market algorithms and services for medical imaging. The university origins and links of these companies benefit Postgraduates. Examples include Siemens Molecular Imaging (formerly Mirada) in Oxford, iMorphics in Manchester and IXICO in London. PhD and MSc projects are also sponsored by companies such as Vision RT who are experts in real time 3D surface imaging for radiotherapy applications, Kodak who are involved in the whole imaging chain, DePuy who develop surgical technology, the Wellcome Trust and the major medical equipment manufacturers such as Philips, Siemens and GE Healthcare. Many of the global pharmaceutical and healthcare companies have research sites in the UK. Notably, a new £76 million Clinical Imaging Centre is under construction as a joint venture between GlaxoSmithKline and Imperial College to use imaging in drug discovery and development.

Postgraduate Opportunities

In 2007, UCL plans to start an MSc dedicated to Medical Image Computing. The course can be taken full or part time and scholarships are available to some students.

The UK Engineering and Physical Sciences Research Council recognised the strength of medical imaging by funding a six-year project that now links Imperial College London, Kings College London, Manchester, Oxford and UCL. These groups, and others in the country, provide opportunities for obtaining a PhD in medical image computing. A bi-annual summer school brings together an international teaching faculty to provide lectures and workshops for Postgraduates studying in the UK and overseas.


Within the UK, university research in medical image computing is well-funded, industrial activity ranges from start-up companies to global pharmaceutical organisations and there is substantial investment by the government in the healthcare sector. This creates a healthy environment for Postgraduate study and research and in a subject that brings together computing, medicine, healthcare, biology, maths, engineering and physics for applications that benefit healthcare and well-being.

David Atkinson
David Atkinson is a lecturer in the Centre for Medical Image Computing at University College London. Since 1996 he has been researching novel algorithms to improve magnetic resonance images. He is currently preparing a new MSc in Medical Image Computing at UCL for launch in 2007.