Nodule Detection in Chest X-Rays with Eye Movements


(a) Robert H. Smith School of Business, University of Maryland, College Park, MD 20782 
(b) Department of Mathematics, University of Maryland, College Park, MD 20782  
(c) University of Maryland School of Medicine, Baltimore, MD 21201
(d) Kelley School of Business, Indiana University, Bloomington, IN 47405
Radiologists often miss nodules that may represent lung cancer on chest radiographs.  We investigated whether eye movements collected during the search for lung nodules by large samples of laypeople may provide information that could assist radiologists in their detection. For that purpose, we analyzed eye tracking data of over one hundred laypeople who reviewed fourteen chest X-ray images, of which seven contained a potentially cancerous nodule. From the eye movement data we estimated Regions of Interest that could cover the nodule (see the figure below). The study demonstrated that the eye movements recorded from laypersons contained information that may assist radiologists in the detection of nodules in chest X-rays. This has important implications for the application of crowd-sourcing in large scale screening programs to detect pulmonary nodules.
Estimated regions of interest on seven chest X-ray images with a nodule. Black dots indicate fixation locations, blue ellipses are the estimated ROIs, and solid gray circles represent the location of a nodule.
Michel Wedel
Robert H. Smith School of Business