12 March 2009 Reverse hierarchy theory and medical image perception
Author Affiliations +
We are unsure about what information is extracted from an image to allow a decision about pathology to be made. Our knowledge of the interplay between top down processing or bottom up, local or global perception, perceptual or cognitive processes is uncertain. However recent research has emphasised the importance of the global or holistic look in medical image perception in which recognition of abnormalities precedes search. Reverse Hierarchy Theory [1] is a useful general theory that helps to explain this. It also enables us to understand what information is extracted from an image and how this relates to expertise. Essentially the theory states that perceptual learning begins at high levels areas and progresses down to lower level areas when better signal to noise is needed. So perceptual learning, defined as an improvement in sensory abilities after training, stems from a gradual top down guided increase in usability of first high then lower level task relevant information. Evaluation of the scan paths of groups of observers with different levels of expertise when undertaking a lung nodule perception task seems to be consistent with the theory. Experts' perception is generally immediate and holistic suggesting high level representations whereas those with an intermediate level of expertise tend to be more variable in their scan paths. Interestingly naïve observers have eye tracking metrics that are more similar to experts suggesting they take a common sense approach using perceptual skills we all have as they lack experience in being able to access the low level information from the chest radiograph.
© (2009) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
T. Donovan, T. Donovan, D. J. Manning, D. J. Manning, } "Reverse hierarchy theory and medical image perception", Proc. SPIE 7263, Medical Imaging 2009: Image Perception, Observer Performance, and Technology Assessment, 72631M (12 March 2009); doi: 10.1117/12.812956; https://doi.org/10.1117/12.812956

Back to Top