British Journal of Radiology (2004) 77, S107
© 2004 British Institute of Radiology
doi: 10.1259/bjr/23442591
Image processing
A Jackson, PhD, FRCR, FRCP
Imaging Science and Biomedical Engineering, The Medical School, University of Manchester, Oxford Road, Manchester M13 9PT, UK
My original vision of this special edition was as volume that would sit on every office shelf to which radiologist could turn when faced with some new piece of analysis software. I think that the authors have admirably fulfilled that requirement and I hope that you will find the articles as informative and valuable as I have.
The relentless advance of technological capability in medical imaging has been matched by an exponential increase in the capabilities and complexity of image analysis techniques. As digital imaging methods such as CT and MRI have improved the radiologist is provided with high resolution, improved contrast and better anatomical demonstration. Simultaneously, and possibly more importantly, these high-quality images have become more amenable to complex secondary analysis procedures. Many clinical consultants are routinely using calculated images ranging from simple techniques such as image subtraction to very complex methods such as pharmacokinetic analyses of time course contrast enhancement data. There is a fundamental problem in the application of all these methods since the clinician is not trained or experienced in the methods associated with image analysis. In most cases analysis is performed by a piece of commercial software and the underlying analysis technique or algorithm is not explicitly described. Even if technical details were given it would be impossible in many cases for the clinician to understand. Unfortunately, most image analysis techniques are based on a series of underlying assumptions about the biological system being imaged and the signal and image formation characteristics of the imaging system. These assumptions are often forgotten as image analysis techniques pass into widespread usage. This is potentially dangerous since an analysis method designed to be used in a particular disease or situation may require assumptions that invalidate its use in other disease states.
The inescapable conclusion is that the clinical radiologist intending to use complex or advanced image analysis techniques in their routine practice must develop an understanding sufficient to enable them to identify potential problems. With the ever increasing range of demands on the clinical radiologist this is effectively impossible. Many of the contributors to this special edition of the British Journal of Radiology are collaborators in a large-scale multicentre research project "medical images and signals" (MIAS), which has been jointly funded by the Medical Research Council (MRC) and the Engineering and Physical Sciences Research Council (EPSRC) to examine the problems associated with the routine use of complex analysis techniques in medicine. The underlying principle of the MIAS project has been to develop methods to provide clinicians with "Information Not Data". This acknowledges that patient management would benefit considerably if the radiologist could be presented with an image that automatically identified, for example, areas of tumour progression or tissue change due to disease, particularly if these changes could be automatically and accurately quantified. In practice however the complexity of the analyses required to provide this sort of decision support facility has stymied their introduction into routine clinical practice and the majority of clinical image interpretation continues to be done by radiologists without computerized image analysis support. There is no doubt that this situation will change rapidly in the near future as analysis techniques become increasingly clinically relevant and robust.
In this special edition leading experts in areas of complex image analysis each review their area of expertise in a format that is aimed at the clinician. Each article is intended to give the clinician who is interested in the technique a detailed introduction to the area, an overview of the analysis problems and solutions that are being commonly applied and some idea of the potential developments to be expected. At first sight of these articles may appear to be off-puttingly technical and there are certainly quite a lot of equations in some of the articles. However, I hope you will persevere in trying to read them since all the authors have taken great efforts to make sure that the articles can be read by those who are unfamiliar with the mathematical and technical complexities of image analysis.
