Skip to main content icon/video/no-internet

Attention and Medical Diagnosis

When a radiologist is presented with a medical image, be it a radiograph or the many hundreds of images generated from a computerized tomography (CT) or magnetic resonance imaging (MRI) scanner, he or she needs to make sense of the images, which are representations of the human body, and perceive pathology among the different ambiguous shapes, shades, and contours. Abnormalities are generally perceived quickly, as eye tracking has demonstrated, with pathology usually looked at within the first two or three fixations. It takes many cases and years of training to become proficient in interpreting medical images as perceptual discrimination is learned and acquired knowledge is converted into a variety of cognitive strategies and cognitive skills. This process is still not fully understood, but it seems practice with feedback is the only way to achieve clinically acceptable performance. Medical imaging has seen many developments in equipment, including the recent move away from film to digital technologies. Research has, however, demonstrated that error rates during the past 50 years remain resistant to these improvements and changes in technology. This demonstrates the importance of perceptual and cognitive factors in the performance of radiologists when interpreting medical images. This entry discusses eye movements, models of medical image perception, searching medical images, and decision processes.

Eye Movements

Eye tracking as a research methodology has provided some insight into the strategies and types of errors that are made by radiologists. Eye movements are not involuntary in visual search but can be described in terms of target selection, which in turn is related to the motivational state of the radiologist and to higher cognitive processes.

Errors can be search errors because of incomplete scanning, detection errors caused by failure to recognize visual features, and decision errors where the wrong decision is made. Most errors, however, are not caused by a failure to perceive but by a failure to recognize and make the correct decision. In mammography, about 30% of cancers are missed, and 70% of those missed cancers are actually looked at, although not identified as cancer.

Eye tracking has demonstrated different search patterns between novices and experts. For example, when looking at a chest radiograph, the expert will exclude large parts of the image during a search for lung nodules and concentrate on regions where lung cancer is more likely to occur. Compared with the novice, the expert will also make fewer fixations and have a greater distance between fixation clusters.

Reflecting the fact that there is currently no formal theory of optimal eye movement strategy in conducting visual search, there is no optimal strategy for radiologists, but eye tracking is useful for attempting to understand the way radiologists assimilate information from a medical image.

Models of Medical Image Perception

The global focal model of medical image interpretation, developed by Harold Kundel and Calvin Nodine, can actually be applied to any situation where domain-specific knowledge is important. According to this model, two forms of image analysis are performed sequentially. The first stage is the global or holistic impression (see color insert, Figure 12). This occurs instantaneously, in the same way as recognition of a familiar face happens. Any perturbations in the image are identified based on learned templates, and are flagged for subsequent searching before the first fixation. This is why abnormalities are usually looked at within the first few fixations. The second stage is when focal scanning takes place, which is essential for subtle lesions that are less conspicuous.

...

  • Loading...
locked icon

Sign in to access this content

Get a 30 day FREE TRIAL

  • Watch videos from a variety of sources bringing classroom topics to life
  • Read modern, diverse business cases
  • Explore hundreds of books and reference titles

Sage Recommends

We found other relevant content for you on other Sage platforms.

Loading