For years, the way many readers judged a brain structure was to glance at the left, glance at the right, and say one looks smaller.
That is an opinion, not a measurement. And it hides a trap. Scanners are not perfectly symmetric in how they render the brain. A structure can appear a little smaller on an image simply because of the machine, not because of the patient.
So we do something more disciplined. We build a normal database, tuned to the specific scanner and matched to the patient's age and sex, and we ask where this brain actually falls against real reference values.
Now the statement changes. Not this looks smaller to my eye, but this measures below where it should for a person like this, on this machine.
The eye is a wonderful tool for pattern. It is a poor tool for magnitude.
When the difference between normal and abnormal is small, precision has to come from the data, not the impression.