Tuesday, February 5, 2008

Your (Radiologist's) Inner Fish

Fish Out of Water is the cover story for the February 2008 issue of Natural History by Neil Shubin. This article is adapted from Shubin's book: Your Inner Fish: A Journey into the 3.5-Billion-Year History of the Human Body, and gives a fascinating look at the evolutionary kinship between humans and fish. One of my favorite paragraphs from this article:

In many ways, humans are the fish equivalent of an old Beetle turned hot-rod. Take the body plan of a fish, reconfigure it to be a mammal, then tweak and twist that mammal until it walks on two legs, talks, thinks, and has superfine control of its fingers—and you have a recipe for trouble. In a perfectly designed world—one with no evolutionary history—we would not have to suffer from hemorrhoids or easily-damaged knees. Indeed, virtually every illness we suffer has some historical component that can be traced back from mammals to amphibians to fish and beyond.
One doesn't have to look too far to see possibly fishy behavior in radiologists. We constantly rely on the built-in image processing in our retinas to help us spot findings on medical images. We are also wary of this retinal intervention, as it sometimes produces visual illusions such as Mach bands, which may mislead our interpretations of these images. The physiological basis for this retinal activity can be traced back to some very early denizens of the sea.

Lateral inhibition -- an edge-enhancement phenomenon which takes place in the retina of the eye -- may be one of the mechanisms responsible for our perceptions of Mach bands. Lateral inhibition was studied by H. K. Hartline in the eye of Limulus polyphemus, the horseshoe crab. Hartline later shared the 1967 Nobel Prize in Physiology or Medicine for his work on the primary physiological processes in the eye.

Hartline applied various visual stimuli to the lateral eye of Limulus and measured the electrical activity of a receptor unit using a pipette microelectrode. Among other things, he discovered that patterns of light projected onto the retina undergo "data processing" directly in the retina before these stimuli even reach the brain. This processing results in edge-enhancement or "crispening" of the image. It's not hard for me to imagine how an eye with built-in image-enhancement would have an evolutionary advantage over an eye which lacks it.

Although the human retina is far more complex than that of Limulus, the underlying process that produces lateral inhibition (and hence Mach bands) may be much the same. The practical applications of Mach band theory in the analysis of thoracic images have been extensively explored by Chasen. He states:
The concept of Mach bands contributes to a greater understanding of three-dimensional structures projected onto two-dimensional routine radiographic images of the thorax. Mach bands can help differentiate normal from abnormal anatomy and thus increase the diagnostic yield from such images. Mach bands can be seen on images that use transmitted or reflective light, including CT scout images (topograms) of the thorax.
However, edge-enhancement can be a double-edged sword. As a resident, I was taught to beware of Mach bands and other visual illusions when viewing medical images. Lane and co-authors summarize it thusly:
Although Mach bands often facilitate perception of roentgen density, misinterpretation of their significance may lead to errors in diagnosis.
In the context of a traumatized bone, a Mach band projected over that bone may be misinterpreted as a fracture line. The converse is also true -- a fracture line can be misinterpreted as a Mach band. Just such a case is shown in the wrist radiograph below. This image comes from a medicolegal case in which the radiologist thought that the fracture line (arrows) extending through the scaphoid bone was actually a Mach band due to the overlying edge of the radius.

However, lateral inhibition is not the only explanation advanced to explain Mach bands. Purves at al have suggested that these bands occur as a consequence of real-world luminance gradients. Whether or not this explanation is correct, the Purves Lab website offers a truly striking set of images demonstrating Mach bands and other visual illusions.

For now, I'll leave the heavy lifting on Mach band causality to the neuroscientists. Personally, I like the mental image of an ancient fish lurking deep in my brain, peering out of my eyes at my patients and their images.

No comments: