Physiology of the Silhouette Sign.

The Physiology of the Silhouette Sign in the Chest X-ray.

Mechanism:
The work of Hartline and Ratcliffe work with Limulus, The Harvey Lectures 37:39 (1942), indicates that there is processing at low levels in the visual system. Indeed the optic nerve, although an extension of the brain, is not big enough to give a parallel transfer of all of the detail, which falls on the retina. Retinal processing may distort the unbiased perception of radiological images.


	Retinal processing gives rise to anomalous bands of density perceived at boundaries of density change between adjacent areas of homogeneous shading in the visual field. The process is related to the surrounding inhibition of neighbouring detectors, a necessary process in the management of an array of detectors.

When the process occurs in the eye, the resulting bands of density are called Mach bands. (after Ernst Mach).

By imitating the result of retinal processing of a contrast boundary, it is possible to give a false impression of contrast change at an edge. This is known as the Cornsweet illusion.
The right-hand panel should appear darker. If the illusion fails to work, try changing your distance from the picture.

Occasionally, an A-P chest in an elderly patient will project a skin-fold, which will imitate the Cornsweet Illusion and give a false impression of a change in grey-scale.

Here is another example, produced by an azygous lobe.

The perception of grey scale is not an absolute, but relative to neighbouring brightness.

Radiological images are large and the contained information frequently involves large areas of similar grey scale, which will influence the way some structures are seen.

The effect of neighbouring grey scale upon the perception of density will influence the retinal processing and the development of Mach bands. The brightness of the diagonal crossing bands of grey is identical, despite their appearance. This illusion works because of the grey scale context and its additional asymmetry.

Not only the context but also the size of the object can affect perception of its grey-scale.

Above 10 minutes of arc, the density of an object is more important than its size. Below this, the size of the object is more closely related to its salience.

Returning to the Cornsweet illusion, the curved structures will produce a curved plot of grey-scale. If this curve is not symmetrical, it will influence the development of the mach bands. Mediastinal boundaries are produced by curved structures. The interpretation of their resulting densities often rests on a false impression that the situation is symmetrical. The direction of X-rays, with their requirement to be tangential when producing a silhouette, will impose an assymetry on the resulting grey scale and will influence the resulting Mach bands.

This asymmetry in a curve of density on the retina produces a more visible white line on a concave boundary and a more visible black line around a convex boundary. These lines are not in the original image. This retinal processing enables the recognition of the original curvatures of mediastinal structures in obliqued views.

This model was constructed with density calculations assuming parallel rays. Geometrical factors with diverging rays may, theoretically, alter the density gradient, but will not change the asymmetry.

Cerebral processing will affect image perception.

The brain tends to recognize different objects as the same, when they resemble each other. In Radiology the process is known as the 'Aunt Minnie' effect ( Felsen). A new object may be categorized as part of a group of objects already experienced and its characteristics will be 'inherited' from the existing group. Object-oriented 'C' programmers will understand this point.

Thus, environment and experience will also affect perception of grey scale. The 'fact' that we know that the inside of a tube should be dark, enables us to see its inside as dark, even when we know the colour is the same shade.

Incidentally, did you notice how the visual system flips from one perception of the image to the other, and back again. Its another reminder that the systematic observer needs to compensate for any lack of multitasking with some parietal processing of visual function (including real and imagined objects in a possible differential diagnosis) and will deliberately look twice at a diagnostic image to allow another interpretation of the complex image that is the plain-film radiograph.

P. Edholm Acta Radiologica 23:161 (1983) has pointed out that intuitive visual responses can be inappropriate when applied to the projection of three dimensional objects, like barium coated colon loops in a radiograph.

In the recognition of a silhouette, contrast change is more perceptible at an edge as a result of retinal processing. This perception of contrast is enhanced by an additional process that may imitate the Cornsweet illusion. Extraction algorithms in digital image chains and film processing both enhance the subjective perception of contrast changes at an edge. In film development, it is called the Eberhard effect.

Some literature references are available.

return to silhouette intro
review physico-chemical factors
review geometrical factors

Ian Maddison, Nov 1994, revised Sept 2007