Chapter 1 — CT Image Formation and Reconstruction

Why the path begins at image formation

A quantitative imaging result can look precise while its meaning was already lost before any feature was calculated. The first question is therefore not which model to fit or which radiomic feature to select. It is what kind of object the image is, and what happened between the patient’s passage through the gantry and the array of values presented to the analyst.

A CT slice is not a photograph and not a raw detector measurement. It is the output of a reconstruction algorithm. DICOM then carries that output together with the information needed to interpret it: the pixels are the payload, while the accompanying metadata provides part of the contract for what those pixels mean and how they are situated. Confusing those layers makes later measurements look more direct than they are.

The chapter follows one causal chain. A scanner first records transmitted x-ray information as projections. A reconstruction process turns those projections into an image, and its settings can change the result. The resulting image is then stored and distributed as one or more series. Only after those steps does a downstream analyst have an image on which to define a region and make a quantitative measurement.

That order matters because a change early in the chain can survive every later check that looks only at the final array. Two arrays may have the same dimensions and appear to show the same anatomy while representing different reconstructions or different amounts of spatial information. A reliable workflow consequently starts by identifying the image-formation process, not by treating the file as a transparent window onto the patient.

The next section develops the first distinction in that chain: what the detectors measure, how projections are formed, and why the visible slice is a computed object rather than a direct reading.