Densitometry Measurements in Autoradiograms

The simplest (and wrong) way of measuring the density of a region in autoradiograms is shown in the figure below, which is a film autoradiogram of a rat brain slice incubated with S-35 labeled RNA probe.

1. Press F3 to enter area selection mode.
2. Select the region of interest using the mouse. The area selection mode can be set in the Configuration Dialog as: `Single', `Multiple', `Polygon', or 'Point to point'. The default mode is `Single', in which a single area is selected by freehand drawing of an outline around the region with the mouse. In Polygon mode, the area is selected by dragging small boxes to form a polygon around the area. This has the advantage that you can interactively change the area selection. Multiple mode is similar to Single mode, except that you can select any number of unconnected areas for each measurement. Point-to-point mode is similar to the drawing mode in xfig.
3. Release the mouse button and then click inside the area to tell tnimage which region is the inside of the area of interest. The selected region will be surrounded by a crawling dashed line.
4. Select ``Color...Histogram'' to create a histogram plot. The graph will contain at least one maximum corresponding to the most probable pixel value, as shown in the figure above.

This is a crude way of estimating the density of a region and is often employed by users of Photoshop and similar programs. The result can be made more accurate by clicking on ``Smooth'' to convert the histogram into a smoother curve, or dragging the mouse on the histogram to produce a summation of all the pixel values, which is shown on the graph as ``Area''. However, all of these methods are much less accurate than using spot densitometry, because it is necessary to manually exclude part of the histogram from the calculation to exclude the contribution from the lighter background. In a good image, there may be two distinct peaks separated by a clearly-defined minimum; but, as shown in the above example, most often there is only a single large peak and an indistinct shoulder.

It is strongly recommended to use ``Manual irregular'' mode in ``Spot densitometry'' (Sec. 7.9.2) to analyze images of this type, rather than this histogram method.

A disadvantage of film autoradiograms is that it is often necessary to expend considerable effort in setting up 12- or 16-bit CCD cameras and expensive light boxes to acquire good quality images. If you use film, a Northern Light light box, costing about $3000, which uses an electronic servo to maintain precisely-regulated light intensity, is recommended to obtain reproducible results. The images can be converted to colorful pseudo-color pictures that give the impression of smooth gradations of labeling. However, this perception is often misleading because it misrepresents the physical process, which is actually a discrete, binary exposure of individual photographic grains in the film or autoradiographic emulsion.

A second source of inaccuracy in film autoradiograms is that using the most probable value introduces errors if the illumination behind the autoradiogram changes. Also, unlike spot densitometry, this method cannot provide sub-pixel-value accuracy.

A better approach is to use a microscope with a 1$\times$ objective, attached to a microscope camera. Under the microscope, what originally appears to be smooth gray areas can be seen to be collections of black grains surrounded by unexposed regions. This sort of image is easily analyzed by the histogram technique just described, and gives a large, sharp peak corresponding to the black exposed grains. This obviates the need for acquiring 12- or 16-bit images and using precisely controlled illumination.

The difference in quality between a photomicrograph and an image acquired from a CCD camera with a zoom lens makes it worth the additional effort. Not only is pixel counting of exposed grains more accurate, it is more direct inasmuch as it is closer to a measurement of the actual physical process of exposing silver grains in the film.

Of course, the ideal method for imaging radiolabeled sections is to forget about film and use a phosphorimager. Nowadays, these are available for $10,000 or less and far surpass film autoradiography in terms of speed and resolution.