The mathematical CAT scan1. A numerical simulationEach image in a CAT scan is produced from a collection of 1dimensional Xray projections, taken from different angles in the same plane. Typically, the object to be scanned lies through the center of a large, vertical doughnut; the Xray source and the Xray detector rotate inside the doughnut, one opposite the other. For each angle sampled, the detector gives a 1dimensional Xray record: it sweeps across and registers, at each position, how much radiation is getting through. This gives, for each angle, a 1dimensional picture which can be interpreted as a graph of the Xray penetrability of the object along a family of lines at that angle. In a CAT scan the information we want is a 2dimensional picture: what the Xray penetrability is at each point of a planar crosssection through the object. The information we get is a collection of 1dimensional records, each one giving the Xray penetrability of the object along a family of parallel lines. The mathematical operation that allows the reconstruction of a function of two variables from the knowledge of its totals along lines is called the Radon Transform. In more mathematical terms, the principle of the Radon Transform is that function of two variables can be reconstructed if its integral is known along any line in its domain. In terms of approximations, it can be reconstructed to a given degree of accuracy if these integrals are known for an appropriately dense family of lines. Here is a crude but illustrative example. A twodimensional ``object'' is constructed on a grid by blacking out a certain number of squares. We can think of this as assigning density 1 to the blackened squares and density 0 to the others. A shape like the letter ``C,'' for instance, can be given by the following pattern of 0s and 1s. This density function can be represented by a 3dimensional graph, as shown on the right.
The integral of the density along a line in the domain can be represented here by adding up the number of ones hit by a line through the grid. Instead of taking all possible lines, we will take eight families defined by the grid. The rows of the grid give one family. For this family the sums are 0,0,0,8,8,4,4,4,10,10,0,0,0 as we move down through the grid. The columns of the grid give a second family, with sums 0,0,0,7,7,4,4,2,2,7,7,0,0 as we scan from left to right. Two more families are given by the diagonal lines in the grid. For example, the upward diagonal starting at the lower left corner has sum 3, while the downward diagonal starting at the upper left corner has sum 4. The remaining four families are given by the lines in the grid of slope 2, 2, 1/2 and 1/2. A line of slope 2 goes over one square and up 2 at each step, and a line of slope 1/2 goes over 2 and down 1 at each step. For example, the line of slope 2 through the upper left corner of the grid has sum 2. (If this grid were a chess board, the families would correspond to straightline moves by rooks, bishops and knights.) These eight families of lines are illustrated in this table: each family is labelled by its slope, and each line is labelled by the number of ones it intercepted in the object. In each family one of those lines is picked put in red.

Welcome to the These web essays are designed for those who have already discovered the joys of mathematics as well as for those who may be uncomfortable with mathematics. Search Feature Column Feature Column at a glance 
Comments: Email Webmaster 
© Copyright
, American Mathematical Society


