The Fourier reconstruction of a head section

The Fourier reconstruction may be viewed simply in the spatial domain as the sum of each line integral times a weighting function (to be chosen) of the distance from the line to the point of reconstruction. Ramachandran's linear interpolation, thought to be an approximation, is merely the choice of a particular weighting function. A modified weighting function simultaneously achieves accuracy, simplicity, low computation time, as well as low sensitivity to noise. Using a simulated phantom, we compare the Fourier algorithm and a search algorithm, very similar but not identical to one described by Hounsfield
⁴
under his conditions, The search algorithm required 12 iterations to obtain a reconstruction of accuracy and resolution comparable to that of the Fourier reconstruction, and was more sensitive to noise. To speed the search algorithm by using fewer iterations leaves decreased resolution in the region just inside the skull which could mask a subdural hematoma.