(1,2-diamino-4-nitrobenzene)dichloroplatinum(II)

After identifying specific virtues in the training program, we looked for themes to categorize them. It appeared helpful to use well known categories, already available in virtue ethical theory. We distinguish between cardinal, intellectual, and moral virtues. This classification dates back to Ancient Greek philosophers such as Plato and Aristotle [38 ]. Some of the virtues identified did not fit into any of these classic categories. We labelled them as professional virtues because these virtues resemble qualities deemed laudable in the profession and practice of the ENT surgeon. See Table 1 for more information about these virtue categories which are based on literature [9 38 39 (link)]. Although such categorization is also widely debated, these four categories served the purpose of providing an insightful overview. Furthermore, any categorization can lead to discussion, as it is almost never completely exhaustive.

Bitter gourd raw material processing. Fresh bitter gourd (Momordica charantia) procured locally in (Mpumalanga Province, South Africa) was sorted, washed thoroughly in running water to remove adhering foreign materials, cut on both ends, and thereafter blended (Milex, Sandton, South Africa). Grapes procured locally at Food Lovers Market were washed and blended using a juicer (Milex, Sandton, South Africa).
Response surface methodology design of experiments (DoE). The suitable ranges for inputs (Table 1 used in the DoE were obtained from the literature and subsequent preliminary experiments. Design-Expert software version 11.0.0 (Stat-Ease Inc., Minneapolis, MN, USA) was used to generate 20 experimental runs (Table 1).
Blended bitter gourd and grapes were mixed at different concentrations and monitored for pH, alcohol, total titratable acidity (TTA), and total soluble solids (TSS), and we concluded that, at a concentration of 35% bitter gourd and 65% grapes, the ratio would be suitable for enough sugar content for fermentation to take place as well as enough bitter gourd juice for its physiochemical properties to be investigated. The mixture was subsequently inoculated with Saccharomyces cerevisiae and Metschnikowia pulcherimma (Anchor Yeast, Lallemand, South Africa). Experiments were conducted in triplicates, and samples were withdrawn after each experimental run to test for alcohol (°P), pH, total titratable acidity (TTA) (% lactic acid), and total soluble solids (TSS) (g/100 g).
Determination of alcohol. A semi-quantification of alcohol content in the beverage was determined using a digital refractometer for brewing (Hanna Instruments (Pty) Ltd., Johannesburg, South Africa) by placing 1 mL of the beverage on the sample well and observing the reading. The Plato (°P) readings were recorded afterwards.
Determination of pH. A pH meter (Hanna Instruments (Pty) Ltd., Johannesburg, South Africa) was first calibrated with standard buffers of pH 4 and 7 and used to measure the pH of respective samples.
Determination of total titratable acidity (TTA). TTA was determined using the AACC Method 02-31.011. This entailed dissolving 10 g of the sample in 100 mL of distilled water. The solution was mixed, and a drop of 1% phenolphthalein was added. The prepared solution was titrated with 0.1 N sodium hydroxide until a faint pink colour was observed.
Determination of total soluble solids (TSS). The TSS value was determined using a digital refractometer (Hanna Instruments (Pty) Ltd., Johannesburg, South Africa) by placing the beverage on the sample well and observing the reading.
Function fitting and neural network construction. The input and output data from the RSM experiments were exported to MATLAB R2020a (MathWorks, Natick, MA, USA) software, and a basic code was used to design and run the neural network. A feed-forward neural network with an input layer and an output layer was used. Fermentation time—X₁ (h), fermentation temperature—X₂ (°C), and starter culture concentration—X₃ (%) were used as the neural network inputs, while alcohol content—Y₁, pH—Y₂, TTA—Y₃, and TSS—Y₄ were used as the neural network outputs. Input and output data from RSM (Table 2) was randomly divided for training (70%), validation (15%), and testing (15%). The Levenberg–Marquardt (LM) training algorithm was used to train, validate, and test the neural network until the desired coefficient of correlation (R2) was obtained.
Statistical analysis. All experiments were conducted in triplicates and expressed as mean ± standard deviation. Analysis of variance (ANOVA) was employed to determine the significance of the data using Design-Expert^® software version 11.0.0 (Stat-Ease Inc., Minneapolis, MN, USA). Significant F tests at (p < 0.05) levels of probability are reported.