The aim of this proposal is to study the role of attentional guidance during search through 3D volumes of images and to assess whether online feedback may aid search through this 3D space. Although it has become the state of the art in radiology, little is known about the role of attention in the visual search of volumetric data. Radiologists are faced with the difficult task of finding hard to detect heterogeneous targets that are often obscured by overlapping tissues. Despite ever-improving methods of providing medical images, the radiologists'perceptual and decision-making skills remain the final arbiter in this important process. When searching for a difficult target, one of the most challenging decisions an observer must make is when to terminate search in the absence of a target. This problem may be exacerbated by poor memory of the locations that have already been searched, leading to either perseveration on previously search locations or premature search termination. We will study the eye-movements of radiologists as they search these displays. Through an analysis of the scan paths and the errors, we hope to help radiologists search more efficiently. Next, we will then create experimental analogues of abdominal CT search tasks faced by radiologists. This will allow us to test by na?ve observers since radiologists are a limited pool of experimental observers. Finally, we will test the hypothesis that it is useful to provide online feedback about eye position and the history of those areas that have already been searched, and as an indication of areas that have been fixated repeated and might contain unidentified targets. Through an understanding of the cognitive requirements of performing these difficult visual searches, we hope to be able increase the efficiency and accuracy of 3D search.
Errors in medical imaging delay the detection of cancer, adversely affecting the prognosis of the patient. 3D imaging has become the state of the art in diagnostic radiology but there is a dearth of research on search behavior and the nature of the errors that are made with these types of images. This proposal aims to study radiologists as they search 3D volumes, to examine analogous behavior in na?ve observers with laboratory analogues of 3D medical imaging, and to test interventions that that may improve diagnostic accuracy in medical imaging.
