Twenty to thirty percent of all small lung tumors actually recorded in chest images are missed by radiologists. The source of these errors is in human perception and decision making. The long term objective of this program is to decrease observer error. A series of experiments is proposed to determine the role of peripheral and central vision in tumor detection and in the discrimination between tumor nodules and normal structures in the images. This will be studied by using a digital display that is coupled to eye position so that independent images can be presented to the foveal and peripheral vision in real time. The display will also be used to highlight selected areas of chest images during viewing. The effectiveness of this visual feedback in reducing error will be evaluated. ROC analysis will be used to quantify decision accuracy.
Showing the most recent 10 out of 12 publications