Information Identification And Analysis Of Complex Antibiotic Samples

Water is the material basis and lifeline for all living things on the planet.In recent years,with the development and utilization of various antibiotics,antibiotics have been widely used in the prevention and treatment of microbial infections and the promotion of animal growth.It leads to the serious abuse of antibiotics,which has caused negligible negative effects on today's natural environment,especially pollution of environmental water bodies such as groundwater,surface water,rivers,lakes,etc.,leaving potential harm to nature and human society.Therefore,antibiotics as a new type of pharmaceutical-type environmental pollutants have attracted the attention of experts and scholars.The water environment contaminated by antibiotics is one of the complex chemical systems.Complex mixed solutions are the most common phenomenon in nature.In general,the internal relationships of complex chemical systems are extremely complex and cannot be accurately described with some clear methods,which has increased the difficulty of studying mixed antibiotics in the water environment.In this paper,a method for detecting the concentration of mixed antibiotics in a complex background is proposed by combining high-throughput technology and machine learning technology,and a device is initially established to realize this idea.Borrowing the rapid detection method of this experiment,simply input the spectral image recorded with antibiotic information into the established neural network model,and then obtain the type and concentration of the mixed antibiotic corresponding to the map,and then invert the original environment with a complex background Types and concentrations of antibiotics.The main research process and results are as follows:(1)Collection of visible spectrum big data in complex environmental solutions.First of all,according to the idea of controlling variables through a large number of experiments in the well plate experiment,we combined the color difference comparison and the solution absorbance curve to select a suitable single developer,and made two mixed developers according to a certain formula.Second,collect the environmental background solution and configure the antibiotic mother liquor needed for the experiment.Finally,based on the imaging principles of high-throughput and visible spectrum,largescale data collection of ternary mixed antibiotics and five-component mixed antibiotic experiments was performed through a simple spectrometer device built by ourselves.(2)Standardized processing and deep learning of color spectrum images.The big data sample set is processed through a series of image standardization processes.The color spectrum obtained from the experiment is processed into 224×224 RGB pictures with uniform numbering.The dimensional labels are input into the computer for deep learning at the same time,and a neural network model is established to predict the antibiotic concentration through pictures.(3)Comparative analysis of prediction effects under different experimental conditions.The data predicted by deep learning and the true concentration of antibiotics are subjected to linear regression analysis.From the perspectives of different composite coloration systems,different concentration ranges,different chip templates,and different complex background solutions,combined with digital analysis methods,the three and five mixed antibiotics are discussed separately.Comparison of prediction effects of prediction experiments under different experimental conditions.The results show that the prediction effect of the method in the five-element mixed antibiotic is better than that in the three-element mixed antibiotic.This detection method performed well in different composite systems,different concentration ranges,different chip templates,and background solutions,and changes in experimental conditions had little effect on the prediction fit effect of the method.(4)Image characteristics analysis and experimental scientific discussion.First,based on the image digitization analysis method,the feature information of the color map is extracted for comparison,and the repeatability of the experimental prediction method is verified by using the gray-scale change trend map of the sample image.Secondly,through the "gray-contour map",the specific distribution characteristics of the experimental spectral image and the specificity of the image were explored.Finally,the three-dimensional space image is used to display the vector information of the complex chip template,and the color information richness of the template is analyzed.The results show that the information is rich.