Ph.D. Dissertation:

The functional low-fat yogurt formulation containing Mentha Pulegium L. extract and Spirulina platensis powder by using artificial neural network – genetic algorithm

The functional low-fat yogurt formulation containing Mentha Pulegium L. extract and Spirulina platensis powder by using artificial neural network – genetic algorithm Nowadays, yogurt is the most commonly used dairy product in the world due to its unique properties and its health-promoting value. Due to the changing lifestyle of people and also, problems such as various diseases, consumption of bioactive components and beneficial microorganisms called probiotics in the diet have been considered. Therefore, enriching the yogurt with these compounds can play an important role in promoting consumers health. In the present study, at first, extraction of Mentha pulegium L. and Spirulina platensis were performed by maceration, ultrasound and microwave techniques and after determination of total phenol content, total flavonoid content and free radical scavenging activity, the optimum method for each plant was determined by SPSS software. Finally, the extract obtained from each optimum method was added to the milk samples to produce yogurt. Secondly, the physicochemical, microbial, rheological, microstructural and sensory properties of probiotic yogurt enriched with different levels of Spirulina and oregano extract (0, 0/2, 0/4, 0/6, 0/8 , 1 % v/v) during21 days of refrigeration time (5?c) were evaluated. In order to predict the experiments results, the multilayer perceptron neural network with two inputs (extract concentration, storage time) and one output using MATLAB R2013a software was used. The learning model of Levenberg-Marquardt, the learning cycle 1000, and correlation coefficient (R2) and mean square error (MSE) were created and evaluated for network. Genetic Algorithm was used to optimize the nerurons number in the hidden layer of artificial neural network. The results of experiments showed that the effect of all studied parameters in analysis of variance on all variables in yogurt samples was significant (p<0/05). For pH values, the correlation coefficient (r) was 0/996. The pH value of treatments containing M. pulegium extract was higher than Spirulina extract. Coefficients of Determination (R2) for all pH values of samples containing Spirulina extract (0/1 %) were more than control and sample (0/71297 %) and M. pulegium extract (0/45741 %). In case of Spirulina and M. pulegium extract, storage time and extract concentration were the more effective factors in predicting pH value, respectively. R2 for acidity was 0/999 and the acidity level of treatments containing Spirulina was significantly higher than M. pulegium extract. The R2 for all data in the neural network model was very high and for both extracts, the storage time had the highest effect during predicting the acidity changes. The effect of all parameters on syneresis was significant (p<0/05) and the correlation coefficient (r) was 0/995. The highest R2 (0/1 %) was related to Spirulina extract. For sensitivity analysis, storage time had the highest effect during prediction of syneresis changes and for storage time was 40 and 45% respectively. The correlation coefficient for water holding capacity (WHC) was 0/999. The R2 of Spirulina extract was more than control sample, and this sample was higher than M. pulegium extract and for both extracts, storage time had the highest effect during prediction the WHC changes. By increasing the shelf life up to the 14th day, the viscosity values of treatments increased and then, decreased. The treatment containing 1 % of Spirulina extract had the highest viscosity. The highest R2 For all data, belonged to the control sample (0/94863 %) and between two treatment groups belonged to Spirulina extract (98845/0 %). In both cases, storage time (21 days) had the highest effect during the prediction of viscosity changes. For inhibitory activity, the R2 was 0/979. The treatment with the highest level of than M. pulegium extract showed the highest activity. Antioxidant activity (AA) increased as the concentration of M. pulegium extract increased and the R2 was high. The extracts concentration had the most effect in predicting the changes of free radical scavenging activity. In evaluating the yogurt samples microstructure by scanning-electron-microscope (SEM), the compaction of casein - gel network increased with increasing the concentration of Spirulina extract and the whey-origine pores and very fine, which a lot of whey stays in the network. The microstructure of the treatments containing 1 % of Spirulina extract was similar to the control sample. In the treatment containing 1 % of M. pulegium extract, casein particles were dispersed in the system and no accumulated mass was observed. This treatment showed the best network structure among the treatments containing different concentrations of M. pulegium extract. As the shear rate increased, the viscosity decreased significantly. At moderate shear rates, the viscosity decrease was less intensity. The trend of changes in viscosity and shear stress versus shear rate indicates pseudoplastic behavior. Yogurt samples showed the non-newtonian and shear-thinning behavior. About Lactobacillus delbrueckii ssp. bulgaricus count, the most significant effect was related to storage time and the correlation coefficient was 0/847. The control sample showed the lowest L. bulgaricus count and the treatment containing 0/4 % M. pulegium extract showed the highest count. The R2 for the data at the test stage for all three groups was high (more than 83 %). In the case of Spirulina extract, the storage time and in the case of % M. pulegium extract, the extract concentration had the highest effect in prediction. Regarding the Streptococcus thermophiles count, the most significant effect was related to storage time. The mean count at concentrations of M. pulegium extract was significantly higher than Spirulina. The R2 of all data for Spirulina extract was highest and the extracts concentration had the highest effect in predicting the changes of bacterial count. As the concentration of Spirulina extract increased, the number of L. casei increased, the highest of which was 1% of this microalgae extract. Over time, the L. casei count declined significantly, although the lowest count on the final day was higher than the standard count of probiotics in yogurt. Therefore, yogurt can be called a probiotic and functional product. Concerning color scores, Spirulina extract had higher R2 (0/83099 %) than M. pulegium extract (0/74641 %) and control sample showed the highest R2 (0/95513 %). The highest aroma score at the end of the storage time belonged to the treatment containing 0/2 % Spirulina extract and the high extract concentrations (0/8 and 1 %) showed the lowest score. The R2 of data in test stage for Spirulina extract and control sample was more than 0/90%. For flavor score, the correlation coefficient was 1 and Spirulina extract obtained the highest score. The R2 for the control sample had the highest value (0/97701 %). Between two extracts, the R2 for Spirulina (0/79419 %) was higher than M. pulegium extract (76313/0 %). The effect of treatments on overall acceptance score was significant and the correlation coefficient was 0/1. The overall scores of treatments containing Spirulina extract were higher than M. pulegium extract. Spirulina extract had a higher R2 than the M. pulegium extract. In terms of color, aroma and flavour, extract concentration and overall acceptance, storage time were the most effective parametr in the sensitivity analysis. Fitting the network outputs for all the parameters showed that the studied model had good predictive power for the qualitative properties of probiotic yogurt samples.

Keywords: Spirulina platensis, Genetic algoritm, Mentha pulegium L. , Neural network, Lactobacillus casei