Global Visual Processing In The Lab And In The Clinic

Brigham And Women'S Hospital, Boston MA

Investigators

Linked publications & trials

Abstract

DESCRIPTION (provided by applicant): The proposed research seeks to illuminate the role of non-selective global image processing in normal visual perception and in medical image processing. In particular, we will investigate whether there are capacity limits to global processing and if it is an under-utilized component of the medical image understanding task. Recent behavioral studies demonstrate that within 200 ms of stimulus onset, observers can extract statistical and structural regularities in images (e.g. what is the mean size or orientation of this display?) even if visual selective attention is focused on another visual task. They can also use those regularities to answer questions about the categorical status of scenes (beach, urban street, etc) or the presence in the scene of some categories of objects (animal, vehicle). This global image processing is often portrayed as having essentially unlimited capacity to perform its particular tasks. However, our hypothesis is that its capacity is limited. The first aim proposes a series of psychophysical studies that will measure the number of global non-selective image properties available to observers at any one time when viewing real world scenes. Our second hypothesis is that global image processing underlies the diffuse sense of target presence reported by experts in a variety of specialized search tasks. The second aim proposes experiments to assess the reliability of the sense of pathology in radiology screening with the goal of developing adequate training programs and/or systems that can support physicians performing medical image processing. We will collect data on the accuracy of "first guess" responses to briefly presented images. Those data will be the input to a global scene processing computational model that will extract global features associated with normal and abnormal images. Once extracted, these global properties will be then tested for their utility in aiding and/or training radiology screeners. Relevance: In many cases, our ability to understand complex visual images not only involves careful scrutinizing of that image, but also rapid access to global structural and statistical regularities of the image. An improved understanding of global non-selective visual processing can be expected to have important health consequences for both the enhancement of training programs in cancer screening and the development of computer aided systems supporting professional diagnostic expertise.