Eye-tracker data analysis in cephalometric landmarking

Eye trackers are becoming increasingly popular and a large amount of data is being produced in the form of eye gaze points and related information. It is hence important to visualize and to analyse sequences of gaze points. Comparison of scanpaths from different subjects can support several types of investigations, e.g., detect similarities and differences between subjects or build predictive models of user performance. In this work we present an eye-tracking study for a real cephalometric landmarking task: on 15 x-rays of the skull with a latero-lateral orientation 5 medical experts have to locate 22 landmarks that, in some cases, are difficult to locate because of overlapping anatomical regions or non-optimal image quality. The set of 22 landmarks represents the most used method adopted in daily clinical practice. This eye-tracking study analyses the eye-tracker data collected during the landmarking of 1,650 landmarks. The objective of this study is to present a gaze point statistical analysis in order to identify a strategy that can predict a good performance in terms of precision in landmarks localization, and speed.