families), (Wodicka et al, 1997; Lockhart et al., 
1996) doubled each synthetic oligonucleotide 
(Perfect Match PM) to produce a new 
oligonucleotide which is adjacent and identical 
to the PM partner except for a single base in a 
central position (MisMatch MM); the difference 
between the PM signal and the MM signal 
should give the constructive signal purified from 
the noisy cross-hybridization signal. The same 
aim was achieved by (Bernard et al., 1996) by 
spotting four identical targets in two opposite 
areas of the membrane to correct the intrafilter 
variation. 
The majority of the studies present in literature, 
e.g. (Lockhart et al., 1996), provide a dynamic 
range extending over three orders of magnitude 
of the signal intensity. 
The aim of the spot signal processing (filters, 
Fourier transform, etc.) is to determine a 
quantity which measures the expression level of 
a particular target in a cell population (single 
color mode) or the differential expression of a 
particular target in two different cell populations 
(dual-color mode). Therefore a wide, noisy and 
variable spot signal must produce a unique 
measurement, which may be analyzed 
afterwards. Increase in target spacing reduces 
overlapping among spot signals (and therefore a 
part of the spot noise is eliminated), but at the 
same time it reduces the numerousness in the 
panel of targets. 
(Wodicka et al., 1997), after discarding outlier 
spots, averaged pixels inside the synthesis feature 
rejecting the pixels lying in the proximity of 
bounds. In this manner the authors reduced the 
image to a simple text file containing for each 
feature some information such as target position, 
nucleotide sequence (if the sequence is known), 
signal intensity, ratios between PM and MM 
signals; this knowledge leads to the decision (by 
means of statistical criteria based on calculating 
the difference of a single-spot quantity from the 
same quantity determined for the entire panel of 
genes) of presence or absence of hybridization 
for the feature. 
The use of negative control genes could give the 
possibility to measure the background noise, 
since these genes should not hybridize and 
should produce the only background noise 
(Bernard et al., 1996). 
By using internal control genes of known 
expression level, the entire gene panel can be 
determined at last. 
5. IMAGE SIMULATOR 
An image simulator has been developed in order 
to study the fundamental image problems 
without having to repeat, each time the 
experimentation protocol and the reading step. 
In this manner appropriate images can be 
obtained, and by means of Fourier transform, 
analysis of the signal components can be carried 
out. These images can be subjected to statistical 
and numerical filters and can be studied by using 
mathematical functions. Usually the processing 
time is very long due to the necessity of 
adequate resolutions. The simulations of single 
spots are repeated using different condition for 
each single target. An object-oriented database is 
used for storing spot images and relative 
simulation parameters. 
The generation process of the single spot is 
simulated by means of the Monte Carlo (MC) 
method. The deposition area of the genetic 
material is thought as composed of countless 
point-shaped sources, which are subjected to the 
system PSF (Point Spread Function). The spot 
area is uniformly sampled a given number of 
times and each extracted sample becomes the 
source which has to be transformed according to 
the chosen PSF; the direct or rejection criterion 
can determine the image pixels which the signal 
reaches and the relative unitary increments 
which contribute to the signal construction. The 
simulation parameters are: image resolution, PSF 
and its parameters, radius of the deposition area, 
number (N) of samples of the deposition area, 
number (M) of samples relative to the extracted 
source and to the chosen PSF, adopted criterion 
(direct or rejection criterion). N and M take into 
account the different amounts of deposited 
genetic material. 
Figure 1 shows two examples of spot simulation 
with normal PSF and unitary standard 
deviation, with the same number of samples but 
with two different deposition areas. 
(a) (b) 
Figure 1 - (a) deposition radius = 3, (b) 
deposition radius = 2.