The holistic grail: possible implications of an initial mistake in the reading of digital mammograms

Abstract

In 1967 Ulric Neisser, studying how laypeople examined pictures, hypothesized that image perception occurs in two stages, a pre-attentive stage in which the entire image is processed in parallel, where a 'holistic' view of what is being displayed is formed, and a secondary stage in which items or groups of items are examined by focal attention. Later, the proponents of Neisser's theory suggested that the pre-attentive stage may bias the selection of the areas that will be subjected to further analysis. This is easily seen in those dual interpretation figures; once one 'sees' the figure in a certain way, it is very hard to instruct the eye-brain system to let go of that perception and 'see' the figure in the alternative way. In medical image perception, Harold Kundel and Calvin Nodine proposed a model of medical image interpretation that is based upon Neisser's two stages, and have become so convinced of the influence of the 'holistic' view on the subsequent reading of the image that they have recently questioned the traditional framework that determines how lesions are found. In other words, as opposed to the traditional view of SEARCH THE IMAGE - DETECT A POSSIBLE FINDING - IDENTIFY THE FINDING - DECIDE WHAT TO DO ABOUT THE IMAGE, Kundel and Nodine have recently suggested a new framework: DETECT A POSSIBLE FINDING - IDENTIFY THE FINDING - SEARCH THE IMAGE - DECIDE WHAT TO DO ABOUT THE IMAGE. In light of this significant switch, we decided to investigate what happens when the 'holistic' view is incorrect.