Problem

In the radiological field, different protocols and settings are adopted in the process of acquiring x-rays depending on the part of the body being researched: for a skull x-ray, the quantity of X-rays irradiated will be different from a knee x-ray, for example. Machine setup is typically done by hand before each acquisition, but the process is tedious and could benefit from automation.

Solution

Using pose estimation techniques, applied to the images of an embedded camera in an x-ray machine, it is possible to identify and localize the parts of the patient’s body subject to x-ray. With the data collected, it is possible to automate, at least in part, the selection of protocols and settings to speed up the acquisition flow.