Kinetics of bacterial cellulose formation in soybean-boiled wastewater medium during fermentation utilizing real-time image processing

The bacterial cellulose (BC) fermentation is affected by environmental growth conditions such as pH, temperature, and turbidity. During the fermentation, a real-time image processing method was applied to observe the BC growth by using soybean-boiled wastewater and coconut water as a mixture. Cellulose will be produced and discharged into the medium throughout the fermentation process to form BC sheets, which will gradually become visible. The purpose of this study is to investigate the correlation between the growth parameters of BC and to analyze the formation of BC kinetically, using the Gompertz model on medium without additional nitrogen source and medium with additional nitrogen sources during the fermentation process. The strongest correlation result was found between time and thickness of BC formed without an additional nitrogen source, and with sodium glutamate as the nitrogen source. The Gompertz model equation was suitable for predicting the kinetics of BC formation time and thickness based on the available data. Both mediums exhibited three clusters that represented the adaptation stage, the exponential stage, and the stationary stage during the fermentation process. The addition of a nitrogen source to the medium resulted in thicker BC sheets compared to the medium without this addition.


Introduction
Bacterial cellulose (BC) is produced purely during synthesis without any other biopolymers, such as lignin, hemicellulose, or pectin.This differentiates it from plant cellulose [1].Its high crystallinity gives BC appealing physical properties, such as viscoelasticity, tensile strength, and extensibility [2].Because of its properties, BC can be employed in numerous fields, including the food, pharmaceutical, and cosmetic industries.Nata de coco is a jelly-like culinary dessert made from fermented coconut water [3].For BC formation on nata, additional nitrogen and carbon sources, which act as fermentation stimulants, are necessary during the fermentation process [4].The parameters, such as temperature, pH, culture medium and oxygen levels, influence the formation of BC throughout the fermentation process [5].
The thickness of the BC formation is typically determined during the harvesting process, which involves a thickness between 1-1.5 cm.The fermentation process takes approximately 6, 8, or 11 days [6,8].In BC fermentation, a tray sealed with paper is usually employed since the thickness of the BC formation is not easily observable, and it would be challenging to monitor the fermentation process.Image processing techniques have been extensively utilized in agriculture, primarily in product sorting, and have been adapted to monitor BC growth during fermentation [9].The synthesis of BC results in layer known as "nata," which can be observed via a USB camera as they thicken during the fermentation process.The thickness of the BC layer is calculated by an image processing algorithm and stored, alongside data detailing the parameters of BC growth throughout fermentation, in a database.These parameters include sensor measurements for turbidity, temperature, and pH [9].
The addition of yeast extract and ammonium sulphate initiated BC formation at hour 61 in the medium with ammonium sulphate and at hour 65 in the medium with yeast extract.However, the study did not compare the difference with the medium without nitrogen addition [10].In this study, mediums without an additional nitrogen source and with additional nitrogen are used to compare the results of thickness on both mediums.Sodium glutamate was added as a nitrogen source to the combination medium to assess the thickness of BC formation.Kinetic models are crucial tools in designing and controlling biotechnological processes for better comprehension of microbial growth behaviour using mathematical models [11].This study aimed to establish a correlation between growth parameters of BC, including pH, temperature, and turbidity.Additionally, real-time image processing with the Gompertz model was employed to conduct a kinetic analysis of BC formation during fermentation.

Materials
The medium fermentation of BC, as identified in a previous study [6], was composed of 70% coconut water and 30% soybean-boiled wastewater.During medium fermentation, additional nitrogen sources which are 0.5% sodium glutamate [7], 5% sucrose [8], and 0.2% acetic acid were added to bring the acidity level to 4. The medium was then cooked and slowly heated to a low temperature of around 60 o C. Afterwards, the medium was cooled to a room temperature of approximately ≤30 o C before being transferred to a fermenter.To start the BC fermentation, the water medium was poured into a fermenter, and a 20% starter of Acetobacter xylinum was added and then shaken until blended.The fermentation was done without stirring, letting it settle at room temperature and covering it with newspapers to maintain aerobic conditions [12].

Fermenter
The fermenter's design relied on existing research [9,10].It was constructed using acrylic and tinted in a non-reflecting black color.Its dimensions were 35 centimetres in long, 25 centimetres in wide, and 6 centimetres in high.The fermenter designed to provide a large surface area for oxygen interaction due to its dimensions.The fermenter top made of newspaper to allow the passage of oxygen.The fermenter had a transparent area measuring 6 cm in width and 4 cm in height for a USB camera.The camera can be positioned in front of the transparent area.A 145-lux LED light is positioned at the front to illuminate the fermenter.

Instrument Used
The image processing instrument used in this study is based on prior research [12].To measure turbidity, pH (acidity), and temperature, the sensors need to be placed through the lid's corner of the fermenter.The USB camera used in this study is from M-Tech brand, with a resolution of 640x480 pixels, which should be positioned in front of the transparent area of the fermenter to capture the thickness of bacterial cellulose during the fermentation process.The study employed the Gravity Analog Turbidity Sensor for Arduino-Water Quality by DFRobot, the pH Sensor Kit E-201C-Blue Grove and the Waterproof Digital Temperature Sensor DS18B20 to measure turbidity, pH, and temperature respectively.Three sensors connected to Arduino were used to monitor bacterial cellulose growth, and the data readings obtained were automatically converted into digital data that the Raspberry Pi 4 could read every 15 minutes [10].The Raspbian Operating System and various programming tools such as Python, OpenCV, and a MySQL database are pre-installed on the Raspberry Pi 4. The table was created using a MySQL database.It includes date, time, fermentation code, and conditions of the bacterial cellulose such as pH, temperature, turbidity, and digital images of the thickness algorithm results.

Telegram Bot
A Telegram Bot was set up to send messages containing data on the data collection period, fermentation code, BC layer thickness, temperature, pH (acidity), and turbidity of the BC medium to track BC growth.As the Telegram Bot was connected to a Python script hosted on a Raspberry Pi 4, the information was sent automatically when online.

Data Collection
The data collected relied on previous study [12], as shown in Figure 1.The Arduino transmitted the sensor data to the Raspberry Pi 4, including temperature, pH, and turbidity.The data from the USB camera was sent directly to the Raspberry Pi 4 using the BC method to calculate the thickness of the BC layer.All of this information is input and kept in the MySQL database of the Raspberry Pi 4 automatically every 15 minutes from the beginning of fermentation until the stationary phase.In addition, the data delivered to the Telegram account of the user every 15 minutes by using the Telegram bot.

Statistical Analysis
The statistical analysis, utilizing R and RStudio, examined to establish correlations among the parameters of BC growth.This examination incorporates Principal Component Analysis (PCA), clustering founded on the relationship between time data and thickness data, and the kinetic analysis of BC formation utilizing the Gompertz model.The GUI program's RStudio Desktop version 2022.07.1+554 is available for download at https://cran.rstudio.com,while R version 4.2.1 can be found at https://www.r-project.org[10].

Data collection during BC fermentation
The study employed two experimental designs, BC fermentation without the additional nitrogen source and BC fermentation using sodium glutamate as a nitrogen source.On medium without an additional nitrogen source, as shown in Figure 2, the BC sheet was formed after 68 hours (4080 minutes) and continued to increase during fermentation.There was a temperature increase of about 25-28 o C following the formation of the BC sheet, showing the BC inoculum's metabolic activity, which releases heat into the medium and causes a temperature increase.Temperature affects the setting of an adaptation pattern organism for its survival by influencing its physiology and homeostasis [13].The yellow graph displays the turbidity parameter during the fermentation process.It is well known that fluctuations appear at the beginning of the fermentation process and tend to remain steady until its the end.At the beginning of the fermentation process, pH was close to 4 and then fluctuated with a decreasing trend in the final stage of the fermentation process which resulted from the conversion of BC precursors to gluconic acid and production of acetic acid [14].On the medium that utilizes sodium glutamate as a nitrogen source, as illustrated in Figure 3, the BC sheet starts to form and continues to grow throughout the fermentation, which lasts 61 hours (3660 minutes).The temperature sensor reading results are presented in a blue graph that shows fluctuations at the beginning and the end of the BC fermentation process.Following the formation of nata layers, the temperature increased to approximately 24-28°C, resulting from bacterial metabolic activity.The pH (acidity) value remained between 3 and 4 during the fermentation process without significant fluctuations.During the initial inoculation, a high number of turbidities appear due to the inoculum, which is shaken with the medium and distributed across the fermenter [12].
Figure 3. Reading data for medium using sodium glutamate.

Correlation analysis
Data is collected from four sensors throughout the fermentation process.The R programming language is used to analyze the correlation between the parameters of time, thickness, turbidity, pH value and temperature.The correlation test is utilized to determine the relationship between time, thickness, temperature, turbidity, and pH during the fermentation process.According to the findings of the correlation test with the highest percentage, adjustments and kinetic analysis can be made.In the medium without an additional nitrogen source, shown in Figure 4(a), the highest correlation observed was between time and thickness at 95%.This was followed by a correlation between thickness and turbidity at 81%, and finally a correlation between time and turbidity at 71%.Cellulose production by bacteria during fermentation is positively correlated with fermentation duration and depends on bacterial viability and dissolved oxygen levels [13].
As shown in Figure 4(b), when utilizing sodium glutamate as a nitrogen source, the correlation with the highest degree was detected between time and thickness (96%), with thickness and turbidity following in at a 59% correlation, and finally, a 50% correlation between time and temperature.The results were found to be similar in the medium without an additional nitrogen source.Over an extended fermentation period, the BC can increase in size and produce a greater amount of BC sheet.The strongest correlation between time and thickness was consistent with previous studies, which used ammonium sulphate as nitrogen in soybean-boiled wastewater medium

Principal Component Analysis and clustering analysis relationship
Principal Component Analysis (PCA) was employed to substitute the original variables that exhibit a certain correlation, reduce the dimensionality of the data, and identify the primary causes of variability [14].In this study, Figure 5 In the medium without an additional nitrogen source, the combined proportion value of PC1 and PC2 was 89.77%, demonstrating the ability to encompass 89.77% of the data variability.The data group was analyzed using clustering data that had three stages based on the highest correlation and PCA analysis.The cumulative proportion values of PC1 and PC2 in a medium containing sodium glutamate as a nitrogen source were 84.59%.This shows that 84.59% of data diversity was obtained.The first stage of clustering data between time and thickness occurred during the initial fermentation, up to 61 hours (3660 minutes), as shown in Figure 6(b).The second cluster, indicated by the green line, exhibited a rise in the formation rank of bacterial cellulose (BC) and was called the exponential stage [11].After 120 hours (7200 minutes), the data indicated stationary growth of BC during fermentation with the use of sodium glutamate.

Kinetic model analysis
The growth rate of BC is explained by an equation model created using R programming, which analyzes data from both mediums.The Gompertz curve model has been used to demonstrate the relationship between time and thickness in the fermentation process of BC.According to the sigmoid model of the Gompertz model, which is explained in reference [10], the rate of growth of BC is determined.Examining the R 2 value using the Gompertz curve plot allows the precision of the relationship between thickness and expected pattern thickness to be assessed, as explained by Nugroho,et.al. [12].Details of the iterations and R values obtained when fermenting on both mediums are given in Table 1.  1 indicates that different equations were obtained for both mediums.According to [15], nonsignificant growth was indicated by the upper horizontal asymptote in the growth curve.In this study, the fermentation stopped after the thickness of the BC layer was stable.The asymptote line on the medium without additional nitrogen was 2.276409e+02, while on the medium using sodium glutamate, it was 2.470665e+02.This aligns with a previous study that showed that using ammonium sulphate in the soybean-boiled wastewater medium results in an asymptote line of 2.618510e+02 on the Gompertz curve model [10].The result indicates that the thickness of BC increases in the medium soybean-boiled wastewater with an additional nitrogen source.As depicted in Figure 7, in medium without supplementary nitrogen, the stationary phase begins 175 hours after inoculation with a 1.1 cm thickness of BC sheet.With the use of sodium glutamate as an additional nitrogen source in the medium, as demonstrated in Figure 8, the stationary phase begins after 163 hours with a BC sheet thickness of 1.2 cm.The thickness of BC sheet is greater when cultured on medium supplemented with a nitrogen source compared to that on medium without an additional nitrogen source.Nitrogen is the primary component and essential for cell metabolism and growth, significantly influencing BC synthesis [16].

Conclusion
The correlation between time and thickness proved to be the highest of other correlations growth of BC parameters in both mediums based on correlation test.Both mediums showed three stages during the fermentation process, including adaptation, exponential growth, and the stationary phase.The kinetics of BC growth during fermentation can be described by a sigmoidal model curve obtained from the Gompertz model analysis.The addition of a nitrogen source to the medium increased the thickness of the BC sheet and resulted in a faster attainment of the stationary phase compared to the medium without the additional nitrogen source.Further research is required to ascertain the most favourable carbon and nitrogen concentrations when utilizing a blend of soybean-boiled wastewater and coconut water medium.

References
[1] Cazon P and Vazquez M 2021 Bacterial cellulose as a biodegradable food packaging material: A review Food Hydrocolloids

Figure 2 .
Figure 2. Reading data for medium without additional nitrogen source.

Figure 4 .
Figure 4. Correlation analysis result in the(a) medium without additional nitrogen source (b) medium using sodium glutamate.

Figure 5 .
Figure 5. Distribution data (a) and clustering data (b) of medium fermentation without additional nitrogen source.

Figure 6 .
Figure 6.Distribution data (a) and clustering data (b) of medium fermentation using sodium glutamate.

Figure 7 .Figure 8 .
Figure 7. Gompertz model plot in medium without additional nitrogen source.
[2] Chen S Meldrum O W Liao Q Li Z Cao X Guo L Zhang S Zhu J Li L 2021 The influence of alkaline treatment on the mechanical and structural properties of bacterial cellulose Carbohydrate Polymers 271 [3] Rusdi R A A Halim N A Norizan M N Abidin Z H Z Abdullah N Ros F C Ahmad N Azmi A F M 2022 Pre-treatment effect on the structure of bacterial cellulose from nata de coco (Acetobacter xylinum) Polimery 67 (3) p 110-118 10 [4] Perez D Jaehde I Guillamon J M Heras J M Querol A 2021 Screening of Saccharomyces strains for the capacity toproduce desirable fermentative compounds under the influence of different nitrogen sources in synthetic wine fermentation Food Microbiology 97 pp 1-11 [5] Lahiri D, Nag M, Dutta B, Dey A, Sarkar T, Pati S, Edinur H A, Kari Z A, Noor H M, and Ray R R 2021 Bacterial Cellulose: Production, Characterization, and Application as Antimicrobial Agen Int.J. Mol.Sci 22 1-18 [6] Maloringan Y, G and Nugroho, D A 2017 Production of nata de coco using soaked soybean water as the alternative usage of Zwavelzuur Ammoniak (ZA) The 3 rd International Conference on Agro-industry KnE Life Sciences pp 43-50.[7] Ramana K V Tomar A and Singh L 2000 Effect of various carbon and nitrogen on cellulose synthesis by Acetobacter xylinum World Journal of Microbiology & Biotechnology 16 pp 245-248 [8] Nugroho D A and Aji P 2015 Characterization of Nata de Coco Produced by Fermentation of Immobilized Acetobacter xylinum Agric.Agric.Sci.Procedia 3 278-82.[9] Nugroho D A, Sutiarso L, Rahayu E S, and Masithoh R E 2020 Utilizing Real-Time Processing for Monitoring Bacterial Cellulose Formation During Fermentation agriTECH.40 118-123 [10] Astuti E N J, Nugroho D A, and Ahmadi T P 2023 Application of real-time image processing for monitoring bacterial cellulose growth in various nitrogen sources using soybean-boiled wastewater medium during fermentation IOP Conference Series: Earth and Environmental Science.1183 012064 [11] Rezvani F Ardestani F Najafpour G 2017 Growth kinetic models of five species of Lactobacilli and lactose consumption in batch submerged culture Brazilian Journal of Microbiology 48 pp 251-258 [12] Nugroho D A, Sutiarso L, Rahayu E S, and Mashitoh R E 2022 Kinetic observations of bacterial cellulose thickness formation using image processing approach during the fermentation process Food Research 6 pp 210-214 [13] Zahan K A, Nordin K, Mustapha M, and Zairi M N M 2015 Effect of Incubation Temperature on Growth of Acetobacter xylinum 0416 and Bacterial Cellulose Production Applied Mechanics and Materials 815 3-8 [14] Li M Li Y Huang X Zhao G Tian W 2014 Evaluating growth models of Pseudomonas spp in seasoned prepared chicken stored at different temperature by the principal component analysis Food Microbiology 40 pp 41-47 [15] Gonzales J M and Aranda B 2023 Microbial growth under limiting conditions-future perspectives Microorganisms 11 [16] Yodsuwan N Owatworakit A Ngaokla A Tawichai N Soykeabkaew N 2012 Effect of Carbon and Nitrogen Sources on Bacterial Cellulose Production for Bionanocomposite Materials 1 st Mae Fah Luang University International Conference

Table 1 .
Iteration, coefficient and R 2 values of the Gompertz equation.