Purpose: CT scans often image an anatomical region of interest (ROI), such as the spine. Traditional reconstruction algorithms require roughly-equal signal levels throughout the axial plane of the body, a limitation which has stymied any attempt to reduce radiation dose to off-ROI tissues not of clinical interest. We propose a modified filtered back-projection (FBP) algorithm, Focused CT (FCT), that permits accurate reconstruction of a limited scan-FOV with a 90% or greater reduction of radiation exposure outside the clinical ROI.
Methods: FCT splits the Fourier transform into low-frequency and high-frequency components. The high-frequency component uses low-noise data from within the ROI (acquired at a typical clinical dose). The low-frequency component uses high-noise data from the entire axial plane, with the off-ROI data acquired at 10% of the typical dose. To simulate a single CT exam acquired at different doses, a cadaver was scanned twice, with CTDIvol values of 17 mGy and 1.7 mGy. Matlab was used to create several hybrid slices combining ROI data from the 17-mGy acquisition with off-ROI data from the 1.7-mGy acquisition. Projection data for these combined images were simulated, then re-reconstructed using the FCT algorithm. The resulting HU values were compared to those from the 17-mGy image alone: quantitatively by use of the mean-square error, and visually by creation of a difference image.
Results: ROI images generated with the 1.7-mGy off-ROI dose appeared identical to those using the 17-mGy dose throughout the axial plane. The mean-square error was negligible (0.0041). The difference image showed no perceptible structures within the ROI.
Conclusion: The FCT algorithm permits accurate reconstruction of an ROI, with off-ROI dose reduction of at least 90%. The development of a mathematical basis for “focused” reconstruction may lead to the engineering of a physical mechanism to reduce radiation dose to off-ROI tissues.
Not Applicable / None Entered.