Measuring quality in computer-processed radiological images

As digital imaging and digital image processing grow in their importance to medical and scientific applications, issues of image quality and utility in sensitive applications also gain importance. Analog images must be digitized if they are to be transmitted or stored in digital media or if they are to be subjected to digital image processing for assisting screening, diagnosis, and research. Digitization and digital signal processing such as lossy data compression, enhancement, or segmentation all entail changing an image from its original form. A key question for all concerned with radiology is to determine when such changes indeed improve the quality or utility of the resulting images. We discuss some of the issues that arise in demonstrating that one image is as good as or better than another in a specific medical application. Particular reference is made to mammograms.