Image processing of medical radiographs for single image display

摘要

There are two reasons for the processing of medical images. First, the image processing should increase the likelihood that a physician will make the correct diagnosis from the image, and, second, the result should increase the efficiency and thereby lower the cost of the process of diagnosis. Each missed diagnosis is a lost opportunity for the patient to get well faster. Each inefficiency added by layers of image processing increases the cost to the health care system which is already burdened by high costs. Often, these two competing demands result in tradeoffs that are usually expressed covertly, rather than openly. It is the combination of engineers, radiographers and physicians that can reach the best compromise. Images can be considered as consisting of context and content. There are images that result in improved diagnosis of certain diseases in which the context is destroyed. There are situations in which the context is destroyed and the content is thereby lost. The image processing of certain images can be quite context specific. It may work in one part of the body because of the context, in this case the underlying anatomic structure, and fail in another because of a different context. Thus, the image processing must be specific to the area of the body studied and the diagnoses likely to be encountered. The goal, which we have been working toward is a single image display: a single display of the image digital data set that provides all of the clinically relevant information within that data set in an easy-to-see form.