This paper is a contribution in the field of air pollution due to particulate matter (PM10) and ozone (O3) concentrations. In particular, the objective of this study is to determine a way of assessing exposure to outdoor pollutions. In the short term, we analyze their impact assessment on population (especially on children, the patients, pregnant women and elderly and on people who suffer from chronic cardio-respiratory diseases) in terms of morbidity and mortality by cardio-pulmonary diseases in Rome. The data used covered a period of two years: 2005 and 2006 and referred to hospital discharge records (HDR) that provide relevant information about the patient's diagnosis. In fact, HDRs measure the effects on population health due to exposure to physical, chemical and biological agents outside the human body. The data used on pollutants and meteorological variables coming from monitoring stations of the ARPAL (Environmental Protection Agency of Lazio Region) positioned in Rome. Environmental dataset examines about 90,000 hourly pattern data and is composed by pollutants variables and conventional meteorological variables: Carbon monoxide (CO), Nitrogen Oxide (NO), Nitrogen Dioxide (NO2), Mono-Nitrogen Oxides (NOx), Ozone, PM10; Temperature (T),Global Solar Radiation (GSR), Relative Humidity (RH),Pressure (P), Rain. The aim of this work is to analyze the annual trend of morbidity and mortality in hospital for cardio-respiratory diseases due to PM10 and ozone, by using neural net (NN) techniques. NN is an important tool to forecast, because it can work as universal approximators of non-linear functions such as those in the environmental systems, without a priori assumptions on its nature, by means of an accurate choice of the variables of the system and of the meaningful patterns, and data distribution. The NN capacity to learn non-linear functions is an important issue in our problem. Moreover, neural network does not require assumptions about the input variable distribution or absence of correlations between such variables. Consequently, NN can be used in evaluating cardio-respiratory diseases due to environmental systems. The selection of appropriate network architecture depends on the number of parameters, the network weights, on appropriate training algorithm and the type of transfer functions used. As architecture we used a 3-layer perceptron model with a single hidden layer, 10 hidden neurons and with sigmoid activation function that approximates nonlinearities. For transfer function, we used the Multi Layer Perceptron (MLP) with an error-back-propagation supervised learning rule. Thus it will be possible to increase the knowledge of levels of concentration of pollution in urban areas and inform the public about environmental risks on human health. By applying NN to the field of environmental epidemiology we want to: • examine the most injurious pollutants to human health and their mechanisms of dispersion • identify the factors and mechanisms of epidemic diffusion • model the cardiovascular and respiratory diseases by means the use of air pollution data coming from urban monitoring station • formulate a methodological suggestion to apply to Neural Net to predict the human health impact due to the different pollutants. Our results show that the NN is able to reproduce a good approximation the causal link between the concentration levels of PM10 and of O3 and the performances of the NN or the ability to reproduce the target, are strictly linked to the variables and patterns selections. As usually, we use 65% of random patterns during the training and the remaining 35% as test, never seen by NN during the learning phase. As evident, the NN predicted well the hospitalisations during all the days of period. The determination coefficient is very high (R2=0.93) and we underestimate the extremes values of admissions both in term of the lower than upper limits. The results for simulations of the respiratory hospitalisations are showed that the determination coefficient is a little worse respect to previous simulation (R2=0.92), even if levels are very significant. In Rome and in general all the Italian cities, high concentration levels of air pollution, are one of the key risk factors of public health. Our data show that in Rome the concentration of PM10 and ozone have exceeded the danger levels suggested by the European Guidelines (DIR2008/50/CE). One of more interesting aim of for the health and environmental question are the connections between the air quality data and the relative effects on the human health, i.e. the hospitalization. In general, while is well know that air quality impacts on the health, the quantification of this the effect is hard to simulate. This happen because the relation between the main variables (such as meteorological and pollutants ones) and health effects cannot be determined directly by deterministic models or by simplified statistical models. In this context, we use one of the most advanced non linear models, such as the Neural Network, to attempt to model the relations between the environment and meteorological data and health effects on the populations. The results obtained by NN model are very encouraging and suggest a way to modeling this complex relation. In fact, we obtained a very meaningful correlations (higher than 0.90) for both simulations. While the cardiac pathology is better reproduced by NN, the respiratory ones have needed more analysis in deep. However, both simulations exhibit that, to optimize the training phase, the choice of the input variables and the choice of patterns are the main factors to be considered for successful of intelligent methodology. By using neural networks, it was possible to determine numerically the association statistically meaningful between the cases of death or hospitalization and pollutants (Pm10 and/or O3). As last consideration, we underline that the results obtained seem encouraging to simulate the effects of air quality on the health through NN model, but at the same time indicate that further study are necessary in future as to extend the NN's prediction to more years and to consider also the spatial distribution of pollutant in relation with the local hospitaliza