Computer-processed medical images open new horizons in surgical techniques and in research into the mechanisms of disease
Modern medicine has been enormously advanced by techniques for the creation and analysis of images. The most powerful modern scanning techniques, particularly X-ray computer tomography (CT) and magnetic resonance (MR) imaging, have emerged only in the past 30 years.
Computers play an integral role the creation of scan images. But they can do much more: computer analysis of images plays an increasing role in many aspects of research and in the management of individual patients. My neurological research has taken advantage of these methods, which complement human visual analysis. The methodological advances we make in techniques that help to solve one clinical problem have opened possibilities in other clinical conditions. What is more, the insight gained from the success in making computers solve the problems of analysing images throws light on possible mechanisms of human vision.
An exciting development is in Creutzfeld-Jakob disease, where we are evaluating the role of MR scanning in helping early diagnosis, particularly in new-variant CJD. In MR scans of some CJD patients, there seem to be subtle changes in the brightness of parts of the brain. We are working on a collaborative project with the CJD Surveillance Unit in Edinburgh and centres in France, Belgium and Germany, using computer methods to enhance the scans and quantify the changes. We are also using post mortem brains to establish if the areas that show up as abnormal on in vivo MR scans are the most severely involved pathologically when studied under the microscope.
Computer processing plays a central role in planning some types of therapy. In a collaboration with a Hampshire company, we have developed a method for neurosurgical planning and guidance that uses stereo video imaging. Before an operation, scans are processed by computer to form a 3-D reconstruction of key structures: the scalp, the skull, the brain surface, the surgical target (usually a tumour), blood vessels and other structures. The surgeon plans the operation using graphical simulations. When the patient reaches the theatre, stereo video cameras survey the operating scene, and computers analyse the video images. Before the incision is made, the computer creates a model of the patient's scalp and matches this model to pre-operative scans. Then the computer tracks an instrument so that its three-dimensional position is known with high accuracy. The live position can be superimposed on pre-operative scan data to relate the instrument to structures detected pre-operatively. Next, the video display from the cameras can be enhanced by overlaying outlines of pre-operative structures to help to guide the surgeon.
Medical image computing plays a big role in research into the mechanisms of disease and in development of treatments. With multiple sclerosis patients in a trial of the drug interferon beta, we have used computers to measure changes in the scan appearances of MS lesions over time. Our results show how multiple new lesions can appear almost simultaneously, followed by periods of quiescence. This throws light on the mechanisms of disease and on the puzzling relationship between relapses in symptoms and the appearance of individual lesions on scans.
The increasing recognition of how computer processing of medical images is opening new horizons in managing individual patients and in medical research has led to a steady increase in research in this field.
Alan Colchester is reader in neurosciences and computing at the University of Kent at Canterbury and a consultant neurologist at Guy's Hospital in London.