Image Reconstruction (I): Computerized X-Ray Scanners

Medical science tends to advancc incrementally, and full-fledged breakthroughs are rare. The discovery of the xray by Wilhelm Conrad Roentgen in 1895 and the subsequent development of the science of radiography is one notable example. In the last 3 years, a new x-ray device known as the CAT-scanner (for computerized axial tomography) has been appearing in an increasing number of hospitals and clinics. On the basis of their experience so far, many radiologists are saying that these computerized x-ray scanners are the greatest advance in diagnostic medicine since Roentgen's discovery, while others are only somewhat less effusive in their praise. CAT-scanners have had an indisputably marked effect on the way radiologists and surgeons diagnose their patients, but it is still too soon to evaluate what the overall contribution of the scanners to the quality of health care will be. The enthusiasm for CAT-scanners derives from their superior ability to detect abnormalities (lesions) in the brain as compared with such conventional neuroradiological techniques as standard skull xradiography (roentgenography), angiography, pneumoencephalography, and radionuclide scanning. Radiologists also cite the relatively noninvasive character of the scanners and their potential for reducing the cost of health care for patients who otherwise would be hospitalized. In the diagnosis of numerous abnormalities of the brain, radiologists at the Mayo Clinic have reported an overall error rate with CAT-scanning of 4 percent on 12,000 scans over a little more than 2 years, for example (1). Disorders visualized included brain atrophy, degeneration of the brain, hydrocephalus, cysts, tumors of the brain and the eye, infarcts (dead areas of the brain due to loss of blood supply), and hemorrhage (Fig. I). In addition, they find that CAT-scanning is applicable to all of the above-mentioned categories of abnormalities, whereas the other methods are each limited to certain ones only. In conventional x-radiography, the image obtained on a film after a diverging xray beam passes through the subject is a projection or shadow of everything standing between the x-ray source and the film. Thus, the image may contain many overlapping organs and tissues which are difficult to separate. In addition, whereas an observer can easily distinguish between air, soft tissue, and bone in an x-ray photograph, the same viewer cannot easily see the few percent difference in the attenuation of x-rays by normal and diseased tissue, even when overlapping images are not a complicating factor. The method embodied in computeriied x-ray scanners to overcome these difficulties is a specific example of a general mathematical technique called reconst ruction of images from projections. In principle, if x-ray photographs are made of' a person's head at an infinite number of angles, it is mathematically possible to reconstruct a full three-dimensional image of the skull and its contents from these projections. Such reconstructions can be made from a finite number of projections, but the reconstructed image is no longer exact. A number of researchers have made reconstructions of two-dimensional cross sections normal to an axis of rotation of (an object (transverse axial tomography) from x-ray photographs taken at equal angular intervals around the axis. This procedure overcomes the problem of overlapping, but the cumulative x-ray dose to a patient would be excessive. In addition, scattering of x-rays by parts of the patient's body would cause a loss of contrast, as it does in conventional x-rav radiography. The use of an electronic detector in place of the x-ray film together with a collimated, narrow x-ray beam and computer processing solves these problems. Since the detector records only a small region at a time, in order to duplicate the