Reliability of Different Implicit Calibration Data Sources

Determine the relative reliability of different naturally occurring training-data sources for implicit gaze calibration, including tool-tissue contact points, finger pointing at the monitor, tool tips used as cursors to indicate structures, and explicit verbal references to anatomical landmarks during surgical discussion.

Background

SurgGaze uses tool-tissue contact points during surgical dissection as implicit calibration samples, but the discussion identifies additional referential behaviors that could also provide training data. These include pointing at the monitor, using a tool tip to indicate an anatomical structure, and verbally referring to a landmark during discussion.

The paper does not compare these sources or establish whether they provide equally accurate calibration targets. Determining their relative reliability is therefore an unresolved problem relevant to scaling implicit calibration beyond manually selected tool-tissue contact points and to designing automated systems that can select among heterogeneous behavioral signals.