13 th Workshop on ADC Modelling and Testing Sep. 22-24, 2008, Florence, Italy SOM Approach in Monitoring and Diagnosis of Obesity-Hypertension Octavian Postolache 1 , Monica Ferreira 2 , Gabriela Postolache 3 , Pedro Silva Girão 1 1 Instituto de Telecomunicações/Instituto Superior Tecnico, Lisboa,Portugal 2 Hospital Santa-Maria, Lisboa, 3 Escola Superior de Saude, Universidade Atlântica, Oeiras Emails: poctav@alfa.ist.utl.pt, gabrielap@uatla.pt, psgirao@ist.utl.pt Abstract - Obesity-hypertension is an emerging concept in pathophysiology. Obesity and hypertension have been turned into an epidemic afflicting all the word, being among the mainly factors that have been burning the health costs. This paper reports a study whose purpose was to develop an objective method to better diagnose and manage this pathophysiology. A data management and data mapping system was developed. Kohonen topological mapping was used in the classification of obesity- hypertension considering clinical characteristics and laboratory results. Thus, the n-dimensional space of physiopathological parameters was converted into a 2D space of the following obesity classes: healthy subject, overweight, obesity class I, obesity class I-hypertension, obesity class II, obesity class II-hypertension, obesity class III, and obesity class III-hypertension. Transient changes in the individual state could also be analyzed using the proposed self-organizing map based model. Characteristics of the designed maps, such as topology and quantification errors, were studied. keywords: neural network, data-mapping, obesity-hypertension syndrome I. Introduction Obesity is rapidly turning into an “epidemic” afflicting much of the industrialized world. A study carried by the Institute of European Food Studies (IEFS) in 1997 in the 15 European Member States showed that the prevalence of obesity is higher in United Kingdom (12%), followed by Spain (11%), while it was lower in Italy, France and Sweden (7%) [1]. Medical expense for obese people is at least 25% higher than for normal weight people [2]. Obesity accounts for 2-6% of total health care costs in several developed countries; some estimates put the figure as high as 7% [3]. The true costs are undoubtedly much higher as not all obesity-related conditions are included in the calculations. There is a lot of knowledge on obesity, but thoroughly view of the phenomenon remains to be done. Despite a sustained preventive work against increase obesity and hypertension, the efforts to manage obesity and hypertension have been soundly defeated. A new perspective is needed for better diagnosis and management of obesity-hypertension. Few works on stratification of risk factors and associated clinical conditions with obesity have been done [4]. To deal with these issues, much research is needed to develop improved statistical methods. In the last years, bio-informatics approach to detect complex pattern and dynamics has been developed. Many researchers are exploring variations and modifications of logistic regression, and automated detection of informative combined effects (e.g. [5]). Additional explorations are being conducted in data mining and machine learning research. Data reduction involves a collapsing or mapping of the data to a lower dimensional space. Pattern recognition, on the other hand, involves extracting patterns from the data to discriminate between groups by using the full dimensionality of the data. Examples of pattern recognition methods include cluster analysis [6], cellular automata [7], support vector machines [8], self-organizing maps [9] and neural networks [10]. We used a self-organizing map (SOM) algorithm because of its ability to visualize multidimensional data in a two-dimensional format, and to make data reduction and abstraction by generating prototype vectors from measurement data. The SOM is especially suitable for exploratory data analysis of large data sets [11,12,13]. We focused on potential cumulative risk of hypertension and obesity. Organ damage and associated clinical condition in obese people increase with the extent of risk factor clustering. Furthermore, according to the ESH-ESC guidelines, hypertension induces high-added risk for target organ damage, diabetes, or associated clinical conditions [14]. Therefore it is worth to diagnose obesity-hypertension relation. The problem with the diagnosis and management of obesity is that the relationship between different items (e.g. laboratory results and/or symptoms) is not always well established, and that there exists a myriad of exceptions for every rule. Worldwide, 45% of all physicians reported never measuring waist