A disease-specific algorithm is a preselected method for image processing that is optimized to identify a specific disease. An example would be that if one knows that a patient is at risk for a pneumothorax, one could have the computer enhance the image so that any pneumothorax would become more conspicuous. While disease-specific image-processing settings do not yet exist, situation-specific image-processing algorithms are commonly used. The clearest example is the use of histogram equalization to enhance the visibility of tubes and lines within the mediastinum and upper abdomen. The settings used enhance this visibility, but with some probable loss of information for subtle disease in the lungs. In the past, optimization methods have emphasized the desire to find image-processing settings that maximize the value of the chest radiograph for all diseases based on both how common the diseases are and their importance to the patient. In the future, it will be possible to have a system in which each chest radiograph goes through several different image-processing methods, each optimized for detection of a specific group of diseases. It has been proposed that one would use the input of clinical information to decide which image-processing method should be applied to each film. There is an inherent risk in this method in that if one only looks for what is expected, then one may fail to detect an unsuspected disease until it is more severe. Thus, I think that more than one image-processing method should be used on all films rather than a single method tailored to look for a specific diagnosis. In the future, computer-detected disease patterns are likely to be used to adjust the image processing or display parameters so that the detected disease is emphasized. The research goal is to have the computer adjust the image so that the radiologist is unlikely to miss the disease rather than have the computer place an arrow on the image directing the radiologists attention to a specific location.
Was this article helpful?