4.3. Digital Image Processing

Many methods for counting dental cementum annulations have been tested in previous studies: a) direct reading under transmitted light microscope (Lipsinic et al., 1986), contrast phase microscope (Großkopf, 1989, 1990; Cipriano-Bechtle et al., 1996), or fluorescence microscope (Kvaal and Solheim, 1995); b) reading from a photographic print (Stott et al., 1982; Condon and Charles, 1983; Naylor et al., 1985; Miller et al., 1988); c) reading from a slide projection on a screen (Condon et al., 1986; Wright, 1990); d) estimating the total number of annulations by comparing the mean thickness of each annulation to the whole thickness of the cementum (Kvaal and Solheim, 1995).

In an attempt to limit the role played by technical ability and past experience of the observer, Lieberman et al. (1990) used a sample of gazelle's teeth to test a computerised method of image analysis, which was able to bring out details of the chromatic and dimensional features of cementum. Using the research by Lieberman et al. (1990) as a starting point for our own study, the analysis of the histological dental sections of the Isola Sacra sample was carried out by applying the digital image processing (DIP) procedures developed at the Section of Anthropology of the "L. Pigorini" Museum of Rome (Macchiarelli et al., 1996a, b; Geusa et al., 1997b, d; see the following diagrams).





The digital image analysis system is composed of an optical microscope and a surgical stereo-microscope connected, via video camera, to a network of computer workstations (Pentium and Apple PowerMacintosh personal computers and an Indigo Silicon Graphic computer). The computer network was designed in order to maximise performance with respect to DIP power and versatility. The network system mixes different software/hardware workstations that have the capacity to operate the majority of the modern digital image analysis software. This tool assemblage, combined with a large on-line storage facility (greater than 50 Gb on hard disk), guarantees productivity and efficiency during all phases of data collection.

Each portion of the histological section with annulations was recorded using a JVC TK1281 video camera (currently, a Polaroid Digital Microscope Camera allows the digitisation of images up to a resolution of 1600x1200 pixels), assembled on a Leitz Laborlux S microscope with polarised transmitted light at 400x, connected to an AST Premmia GX P/90 personal computer with an analog-to-digital (A/D) converter board (Neotech ImageGrabber PC). Each image, captured as an average of 16 subsequent shots of the same field with the Neotech ImageGrabber PC ver. 1.2 software, has been saved in TIFF standard format at 8 bits greyscale (with a resolution of 764x573 pixels), and processed with an Apple PowerMacintosh 8100/80 computer, using Graftek Optilab 2.1 and/or NIH Image 1.62 image analysis software, or with a Pentium Pro 200 computer, using UTHSCA Image Tool software.

To improve the quality of the image, and thus facilitate the counting of annulations, the following standardised procedures have been applied using Graftek Optilab 2.1: a) rotation of the image in order to have the annulations systematically oriented vertically on the monitor, for the application of directional filters; b) intensification of light and dark areas by means of exponential transformation (Power-Y)), a look-up table function (LUT) able to modify the greyscale of 256 levels, lightening the light pixels (grey levels trending towards 255) and darkening the dark pixels (grey levels trending towards 0); c) increase in contrast by applying a non-additive West ® East directional convolution (Prewitt 3x3 w/image and/or Gradient 5x5 w/image kernels), a filter that gives each pixel a value derived from the weighed mean of the adjacent pixels, limiting the shades and highlighting the discontinuities (band separation).

The counting of the bands (arbitrarily, the dark bands have been chosen), comparable to the number of annulations, was performed by both reading directly from the monitor and by means of a line profile, a specific software option. In the latter case, the software processes a digital diagram of the grey levels over a line drawn by the researcher across the annulations. In resultant diagram, the minima represent the dark bands, while the maxima represent the light ones. In order to reduce the disturbances caused by thin lines not related to the annulations (cutting marks, micro-lesions, etc.), each plot profile has been filtered with a base 5 running average. Moreover, the software NIH Image and UTHSCSA Image Tool perform both three-dimensional and mean line profiles (i.e., considering an area instead of a line), which minimise these disturbances.