In transmission computed tomography, relative X-ray attenuation measurements are made at various angles around a patient's body. These input projection data are reconstructed to yield a cross-sectional view of internal structure. If the body section contains material that severely attenuates the X-ray beam (e.g., surgical clips, lead fragments), high-density streaks that obliterate internal structure will be produced in the process of image reconstruction. This loss of diagnostic information renders the scan useless. A technique has been developed that removes this imaging artifact. The approach views the affected projection data as misinformation. These data are assigned new values, and image reconstruction is performed without changing existing computer hardware or software. Projection data for a head section containing a lead fragment were obtained by Monte Carlo simulation. Three methods of obtaining replacement data were examined. A nearest-neighbor pattern recognition technique yielded excellent results.