TEM Journal – Volume 2 / Number 2/ 2013. 115 www.temjournal.com Estimating SAD Low-Limits for the Adverse Metabolic Profile by Using Artificial Neural Networks Edith Stokic 1 , Biljana Srdic Galic 1 , Aleksandar Kupusinac 2 , Rade Doroslovacki 2 1 University of Novi Sad, Medical Faculty, Hajduk Veljkova 3, 21000 Novi Sad, Serbia 2 University of Novi Sad, Faculty of Technical Sciences, Trg Dositeja Obradovića 6, 21000 Novi Sad, Serbia Abstract – Cardiovascular atherosclerotic diseases represent the significant cause of death worldwide during the past few decades. Obesity is recognized as an independent factor for the development of the cardiovascular diseases. There is a strong correlation between the central (abdominal) type of obesity and the cardiovascular and metabolic diseases. Among a variety of anthropometric measurements of the abdominal fat size, sagittal abdominal diameter (SAD) has been proposed as the valid measurement of the visceral fat mass and cardiometabolic risk level. This paper presents a solution based on artificial neural networks (ANN) for estimating SAD low-limits for the adverse metabolic profile. ANN inputs are: gender, age, body mass index, systolic and diastolic blood pressures, HDL-, LDL- and total cholesterol, triglycerides, glycemia, fibrinogen and uric acid. ANN output is SAD. ANN training and testing are done by dataset that includes 1341 persons. Keywords – Artificial Neural Networks, Adverse Metabolic Profile, Obesity, Sagittal Abdominal Diameter. 1. Introduction It is well known that the risk of cardiovascular and metabolic abnormalities is determined by specific distribution of the adipose tissue. Abdominal (central) obesity is associated with dyslipidemia, impaired fasting glucose, insulin resistance and hypertension, which result in increased risk of cardio- and cerebrovascular diseases, and consequently premature death [1]. Adverse effects of the abdominal obesity have been supported by many studies of the metabolism and endocrine activity of adipocytes from different regions of the abdominal adipose tissue. Abdominal fat includes two morphologically and functionally different depots: subcutaneous (superficial) and deep, visceral (intraabdominal). The latter is located in the abdominal cavity and includes intraperitoneal (omental and mesenterial) adipose tissue, which makes 80% of the intraabdominal fat mass, and retroperitoneal adipose tissue, which makes 20% of the intraabdominal fat mass [2]. Visceral adipose tissue function plays a crucial role in the development of metabolic abnormalities and insulin resistance, mainly due to the direct access of intraperitoneal adipose tissue to the liver through the portal circulation [3, 4]. Body mass index (BMI) has been widely accepted as a simple and the most practical measure of fatness in clinical and epidemiological surveys, eventhough it does not distinguish fat from lean body mass. The values BMI ≥ 25 kg/m 2 correspond to the overweight, and values BMI ≥ 30 kg/m 2 correspond to obesity and indicate increased risk of obesity-related adverse health outcomes [5]. In Serbia 54.4% of adult suffers from excessive body mass, while 36.2% of the population is obese [6]. The highest prevalence of overweight and obesity is observed in the region of Vojvodina and is as high as 58.5% [7]. BMI does not provide sufficient information about fat mass. Therefore, body composition assessment is necessary for the diagnosis of obesity and prediction of its comorbidities. Several anthropometric indicators of abdominal obesity have been developed to measure abdominal adipose tissue mass. Sagittal abdominal diameter (SAD), or abdominal height was first demonstrated in 1988 by Kvist et al. to be a good correlate of visceral adipose tissue volume, observed by CT [8]. In 1994, Sjöstrom et al. proposed the use of sagittal abdominal diameter in the assessment of visceral fat mass [9]. Soon after, Richelsen and Pedersen confirmed its value in assessing the abdominal fatness and prediction of the metabolic risk profile [10]. In this paper, the multilayer feed-forward ANN with back-propagation as the training algorithm has been applied to estimating low-limits of the sagittal abdominal diameter SAD for the adverse metabolic profile. Our idea is to train ANN to predict SAD based on gender (GEN), age (AGE), body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), total cholesterol (TCH), HDL- cholesterol (HDL), LDL-cholesterol (LDL), triglycerides (TG), glycemia (GLY), fibrinogen (FIBR) and uric acid (UAC). We will test various