Multi-algorithm microorganism culture comprehensive identification method and system based on convolutional neural network

Abstract

The invention belongs to the technical field of bioinformatics, discloses a multi-algorithm microorganism culture comprehensive identification method and system based on a convolutional neural network, and aims to improve the analysis efficiency, realize automatic flow, reduce manual intervention, remarkably improve the bacterial colony counting, drug sensitivity analysis and strain identification efficiency and meet the requirements of high-throughput experiments. The deep learning technology and the image processing algorithm are applied, so that the accuracy of bacterial colony recognition, inhibition zone recognition and strain identification is remarkably improved, and misjudgment and missed judgment are reduced. Functional diversity is enhanced, counting, drug sensitivity analysis and strain identification functions are integrated, and diversified analysis requirements such as analysis of different types of culture media and bacterial colonies are met. The algorithm is optimized, the adaptability to different types of culture media and bacterial colonies is improved, and the application range is expanded, for example, accurate analysis is carried out on bacterial colonies with different colors and forms. New strain information can be identified and recorded, and important data is provided for microbiological research.

Description

Technical Field

[0001] The present invention belongs to the technical field of bioinformatics, and in particular relates to a multi-algorithm microbial culture comprehensive identification method and system based on convolutional neural network. Background Art

[0002] In the field of microbial analysis, existing technologies mainly include manual counting, traditional automated counting and flow cytometry. Manual counting relies on manual operation, is inefficient, and the results are easily affected by subjective factors, and cannot provide detailed information. Traditional automated counting has a single function, limited recognition ability, and poor adaptability to different types of culture media and colonies. Existing products can meet certain requirements, but there are still some unsatisfactory aspects, such as limiting visible light acquisition, strict requirements for image acquisition conditions, and the inability to analyze infrared and ultraviolet light, as well as the inability to identify fluorescently stained colonies, which can only be used for fermentation processes. Some products even require special culture media to meet the results of analysis, and flow cytometry is expensive and cannot provide morphological information. New instruments interfere with the recognition results of special marks on heteromorphic colonies or culture dishes, such as markers and labels. The problems with these technologies make it difficult to meet the needs of efficient, accurate and comprehensive analysis in the fields of medical care, food safety, environmental monitoring and cosmetics.

[0003] Medical field: Traditional pathogen detection and drug sensitivity analysis rely on manual operation, which is time-consuming and the results are easily affected by human factors, leading to delayed diagnosis and treatment. Existing automated equipment has a single function and cannot perform colony counting, bacterial species identification and drug sensitivity analysis at the same time, making it difficult to provide comprehensive diagnostic information.

[0004] Food safety: Traditional methods rely on manual counting to detect microbial contamination, which is inefficient and inaccurate, making it difficult to effectively ensure food safety. Existing automated equipment has a low degree of automation, making it difficult to fully evaluate microbial indicators in the food production process.

[0005] Environmental monitoring: Existing technologies make it difficult to quickly and accurately detect the types and quantities of microorganisms in environmental samples, resulting in inefficient environmental pollution assessments and ecological research and poor reliability of results.

[0006] Detection of microbial residues in cosmetics: Traditional methods are difficult to effectively detect microbial residues in cosmetics, especially the limited ability to identify and record new strains of bacteria, which leads to hidden dangers in product quality control and safety assessment.

[0007] Through the above analysis, the problems and defects of the prior art are as follows:

[0008] (1) In the field of microbial analysis, traditional colony counting, drug sensitivity analysis and bacterial species identification methods mainly rely on manual operations, which are not only inefficient but also easily affected by subjective factors, resulting in unstable and unreliable results. In addition, manual operations are difficult to meet the needs of high-throughput experiments.

[0009] (2) Although existing automated analysis equipment can improve analysis efficiency, they can often only complete a single task and lack versatility. They also have limitations in processing complex images and identifying microcolonies, inhibition zones, or similar colonies.

[0010] (3) Existing methods are prone to misjudgment or omission when dealing with different types of culture media and colonies, resulting in inaccurate analysis results. At the same time, it is also necessary to solve the impact of labels or markers on the culture dish itself on the identification results, as well as the identification of irregular colonies and the division of adhesions. Summary of the invention

[0011] In view of the problems existing in the prior art, the present invention provides a multi-algorithm comprehensive identification method for microbial culture based on convolutional neural network.

[0012] The present invention is implemented in this way: a multi-algorithm microbial culture comprehensive identification method based on convolutional neural network includes:

[0013] Step 1, data set acquisition;

[0014] In the task of microbial culture image recognition, the acquisition of data sets is a crucial first step; the data sets should contain different types of culture media, colony morphology, color, size and other diversity;

[0015] Step 2: Data labeling;

[0016] Data annotation is a key step in training deep learning models;

[0017] Step 3, data preprocessing;

[0018] Data preprocessing is an important step to improve model training results;

[0019] Step 4, feature extraction;

[0020] Feature extraction is to extract useful features from the processed images for subsequent model training;

[0021] Step 5: Model training;

[0022] Model training is carried out using a convolutional neural network deep learning model;

[0023] Step 6, model adjustment;

[0024] Model adjustment is to further optimize the model based on the training and validation results.

[0025] The multi-algorithm microbial culture comprehensive identification system based on convolutional neural networks acquires images through a high-resolution camera or microscope. After preprocessing such as grayscale conversion, enhancement and denoising, it uses a deep learning model to perform colony detection, drug sensitivity analysis and bacterial species identification. The system includes a colony counting module (image capture → preprocessing → colony detection → feature extraction → counting), a drug sensitivity analysis module (image capture → preprocessing → inhibition zone detection → size measurement → drug sensitivity analysis) and a bacterial species identification module (image capture → preprocessing → differentiation between new and old colonies → terminal colony circle selection → bacterial species identification → memory of new bacterial species). Through data enhancement, network structure optimization and hyperparameter adjustment, the system achieves efficient and accurate microbial identification and analysis.

[0026] Another object of the present invention is to provide a computer device, characterized in that the computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the multi-algorithm microbial culture comprehensive identification method based on convolutional neural network.

[0027] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to execute the steps of the multi-algorithm microbial culture comprehensive identification method based on a convolutional neural network.

[0028] Another object of the present invention is to provide an information data processing terminal, which is used to implement the multi-algorithm microbial culture comprehensive identification system based on convolutional neural network.

[0029] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0030] First, the present invention aims to improve the shortcomings of the existing technology, such as low efficiency, insufficient accuracy, single function and poor adaptability. By integrating multi-modal functions, applying deep learning technology and optimizing image processing algorithms, it provides efficient, accurate and comprehensive microbial analysis tools for the fields of medical care, food safety, environmental monitoring and cosmetics, effectively solving practical needs that cannot be met by existing technologies, and can record new strains to improve product quality control and safety assessment levels.

[0031] The present invention adopts deep learning technology and image processing algorithm to realize automatic counting, drug sensitivity analysis and strain identification of microbial colonies. First, the system obtains the colony image in the culture dish through an image capture device. Subsequently, the image is processed in a series of steps, including grayscale conversion, image enhancement, contrast enhancement and denoising, to improve the image quality and provide clear image data for subsequent analysis. Then, the processed image is subjected to colony detection and identification using deep learning technology, especially convolutional neural network model. The model is trained to accurately identify colonies in the image and distinguish the boundaries between different colonies. In order to further analyze the colonies, the system extracts a variety of features of the colonies, including area, perimeter, shape factor, edge smoothness and color consistency. These features provide more comprehensive information for colony identification and classification and help improve the accuracy of the analysis.

[0032] The multi-algorithm microbial culture comprehensive identification method based on convolutional neural network provided by the present invention has the following beneficial effects:

[0033] Improve analysis efficiency: Automated processes reduce manual intervention, significantly improve the efficiency of colony counting, drug sensitivity analysis and strain identification, and meet the needs of high-throughput experiments.

[0034] Improve analysis accuracy: The application of deep learning technology and image processing algorithms can significantly improve the accuracy of colony recognition, inhibition zone recognition and bacterial species identification, and reduce misjudgments and missed judgments.

[0035] Enhanced functional diversity: Integrates counting, drug sensitivity analysis, and bacterial species identification functions to meet diverse analysis needs, such as analysis of different types of culture media and colonies.

[0036] Improve adaptability: Optimize algorithms to improve adaptability to different types of culture media and colonies, and expand the scope of application, such as accurate analysis of colonies of different colors and morphologies.

[0037] Record new bacterial species: Able to identify and record information on new bacterial species, provide important data for microbiological research, and enrich the bacterial species database.

[0038] Remember new bacterial species: The system can identify new species of bacteria and remember and correctly label them when they are discovered later, improving the continuity and accuracy of bacterial species identification.

[0039] Second, as auxiliary evidence of the inventiveness of the claims of the present invention, it is also reflected in the following important aspects:

[0040] (1) The expected benefits and commercial value of the technical solution of the present invention after transformation are:

[0041] Improve the efficiency of microbial cultivation: Through automated and intelligent comprehensive identification methods of microbial cultivation, the efficiency of microbial cultivation can be greatly improved, labor costs and time costs can be reduced, and it is suitable for large-scale industrial production.

[0042] Precision medicine and biopharmaceuticals: In the field of precision medicine and biopharmaceuticals, accurate identification of microbial culture is a key link. The present invention can provide efficient and accurate microbial detection methods for drug development, vaccine production, etc., and has significant commercial value.

[0043] Food safety and environmental monitoring: Microbial detection plays an important role in food safety and environmental monitoring. The present invention can quickly and accurately identify harmful microorganisms in food or contaminating microorganisms in the environment to protect public health and environmental safety.

[0044] Market competitiveness: This invention fills the technical gap in automated identification of microbial culture, can provide technical barriers for enterprises, and enhance market competitiveness. It is expected to have broad commercial application prospects in the fields of biotechnology, medicine, food, and environmental protection.

[0045] (2) The technical solution of the present invention fills the technical gap in the industry at home and abroad:

[0046] Gaps in automated identification of microbial cultures: At present, there are still technical gaps in automated identification of microbial cultures both at home and abroad, especially in the comprehensive identification method combining multiple algorithms and convolutional neural networks. There is no mature technical solution yet.

[0047] Real-time monitoring and intelligent analysis gap: Existing microbial culture identification technologies mostly rely on manual microscope observation or single algorithm image processing, which cannot achieve real-time monitoring and intelligent analysis. The present invention fills this technical gap through multi-algorithm fusion and convolutional neural network.

[0048] Blanks in large-scale applications: Existing microbial identification technologies are difficult to meet the needs of large-scale industrial production. The present invention, through automation and intelligent means, can achieve efficient identification in large-scale microbial culture, filling the technical gap in this field.

[0049] (3) Whether the technical solution of the present invention solves the technical problems that people have been eager to solve but have not been able to solve successfully:

[0050] Automation and intelligence of microbial culture: For a long time, the identification of microbial culture has relied on manual operation, which is inefficient and prone to errors. The present invention solves this technical problem by combining convolutional neural networks and multiple algorithms to achieve automation and intelligent identification of microbial culture.

[0051] The problem of multi-dimensional identification: Traditional microbial identification methods usually rely on a single feature or a single algorithm, which is difficult to cope with the complex and changeable microbial culture environment. The present invention solves this problem by integrating multiple algorithms and being able to identify from multiple dimensions.

[0052] The problem of real-time monitoring and feedback: In the process of microbial cultivation, real-time monitoring and feedback are the key to improving cultivation efficiency. The present invention can achieve real-time monitoring and feedback through automated image acquisition and intelligent recognition, thus solving this technical problem.

[0053] (4) Whether the technical solution of the present invention overcomes technical prejudice:

[0054] Dependence on a single algorithm: Traditional technologies often rely on a single image processing algorithm (such as traditional image classification or feature extraction methods), resulting in limited recognition results. This invention breaks the limitations of a single algorithm and overcomes technical bias by combining convolutional neural networks and multiple algorithms.

[0055] Dependence on manual operation: Existing microbial culture identification technologies mostly rely on manual microscope observation, which is inefficient and prone to errors. The present invention reduces manual intervention and overcomes technical bias through automation and intelligent means.

[0056] Neglect of large-scale applications: Traditional technologies are difficult to meet the needs of large-scale industrial production, while the present invention can achieve efficient identification in large-scale microbial culture through automation and intelligent means, thus overcoming technical bias. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a flow chart of a multi-algorithm microbial culture comprehensive identification method based on a convolutional neural network provided in an embodiment of the present invention.

[0058] Figure 2 It is a structural block diagram of a multi-algorithm microbial culture comprehensive identification system based on a convolutional neural network provided in an embodiment of the present invention.

[0059] Figure 3 It is a simplified representation of the CNN steps provided in the embodiment of the present invention.

[0060] Figure 4 This is a microbial colony identification diagram provided by an embodiment of the present invention.

[0061] Figure 5 This is a colony feature enhancement diagram provided by an embodiment of the present invention.

[0062] Figure 6 This is a CNN preliminary preprocessing image diagram provided by an embodiment of the present invention.

[0063] Figure 7It is a simple effect and flow chart of colony counting provided by an embodiment of the present invention.

[0064] Figure 8 It is a simple effect and flow chart of the inhibition zone analysis provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0066] like Figure 1 As shown, a multi-algorithm microbial culture comprehensive identification method based on a convolutional neural network provided by an embodiment of the present invention includes the following steps (the effect step flow is as follows Figure 7 exhibit):

[0067] S101, data set acquisition;

[0068] In the task of microbial culture image recognition, the acquisition of data sets is a crucial first step; the data sets should contain different types of culture media, colony morphology, color, size and other diversity;

[0069] S102, data annotation;

[0070] Data annotation is a key step in training deep learning models;

[0071] S103, data preprocessing;

[0072] Data preprocessing is an important step to improve model training results;

[0073] S104, feature extraction;

[0074] Feature extraction is to extract useful features from the processed images for subsequent model training;

[0075] S105, model training;

[0076] Model training is carried out using a convolutional neural network deep learning model;

[0077] S106, model adjustment;

[0078] Model adjustment is to further optimize the model based on the training and validation results.

[0079] The data set is the basis for training the convolutional neural network model. In the task of microbial culture image recognition, the data set needs to cover various types of microbial culture media and their colony morphology. These colonies may have different shapes (round, irregular), colors (white, yellow, etc.), sizes, and edge characteristics. Therefore, the diversity of the data set directly affects the generalization ability of the model. Image data is obtained through microscopes, incubator imaging equipment, etc., and combined with possible variables in the actual culture environment to ensure that the data is representative enough, laying a solid foundation for subsequent steps.

[0080] Data annotation is the process of associating the collected image data with specific microbial categories and feature information. The accuracy of annotation determines the effectiveness of model training. At this stage, professionals use annotation tools to classify, mark and record the characteristics of the colonies, such as color, morphology, size and culture medium type. The attributes of each annotated image will serve as a label for the training model to support the supervised learning of the convolutional neural network. Standardized methods are used in the annotation process to ensure consistency and high quality of annotation.

[0081] Preprocessing is an important step to improve the model training effect, which mainly includes data cleaning, enhancement and normalization. The cleaning step removes low-quality, incomplete or noisy image data; the enhancement step expands the data set size by rotating, flipping, scaling, cropping, adding noise, etc.; the normalization process adjusts the image to a uniform size and normalizes it to match the input requirements of the convolutional neural network. In addition, data balancing technology is used to avoid model bias caused by uneven distribution of data categories. This process makes the data more suitable for feature extraction of neural networks.

[0082] The core function of convolutional neural networks is to automatically extract efficient features from images. Through the combination of multiple convolutional layers and pooling layers, low-level features (such as edges and textures) to high-level semantic features (such as colony morphology and color distribution) are gradually extracted from the original image. This hierarchical feature extraction method is more robust and adaptable than the traditional manual feature extraction method. The weights of the convolution kernel are continuously optimized through training, thereby capturing the key information of the colony and laying the foundation for subsequent classification tasks.

[0083] During the model training phase, the processed image data is input into the predefined convolutional neural network structure. The network calculates the predicted value through forward propagation and compares it with the actual label, and uses a loss function (such as cross entropy) to evaluate the prediction error. The network parameters are optimized through the back-propagation algorithm, and the weights are iteratively adjusted to minimize the error. During the model training process, the training set is used for learning, and the validation set is used to evaluate the performance of the model to prevent overfitting. Commonly used optimizers include Adam, SGD, etc., which are combined with the learning rate adjustment strategy to improve the training effect.

[0084] After the initial training is completed, the model is adjusted and optimized based on the performance of the validation set. This includes adjusting the network structure (such as the number of convolutional layers, the number of neurons), changing hyperparameters (such as learning rate, batch size), or introducing regularization techniques (such as Dropout, L2 regularization). If the model performance is insufficient, it may be necessary to expand the dataset or adjust the data annotation strategy. In addition, applying the pre-trained model to the microbial identification task through transfer learning technology can significantly improve the model effect of small-scale datasets. Finally, the optimized model is evaluated through the test set to ensure that its generalization ability and robustness meet actual needs.

[0085] The multi-algorithm microbial culture comprehensive recognition method based on convolutional neural networks combines the advantages of data preprocessing, automatic feature extraction and deep learning model optimization to achieve efficient and accurate microbial image classification and recognition. This method effectively improves the intelligent level of microbial culture analysis and provides important technical support for biological laboratories, medical diagnosis and environmental monitoring.

[0086] like Figure 2 As shown, a multi-algorithm microbial culture comprehensive identification system based on a convolutional neural network provided by an embodiment of the present invention includes:

[0087] Colony counting module: Based on the colony identification results, the system counts the colonies and calculates the actual number of colonies in combination with the dilution concentration. This step ensures the accuracy of the counting results and can reflect the actual colony density. The specific steps are as follows:

[0088] Image capture: Capture images of colonies in culture dishes using a high-resolution camera or microscope;

[0089] Image preprocessing: grayscale conversion, image enhancement, contrast enhancement and denoising are performed on the captured images to improve image quality;

[0090] Colony detection and identification: Use the convolutional neural network model to detect and identify colonies in the processed images and distinguish the boundaries between different colonies;

[0091] Feature extraction: Extract features such as area, perimeter, shape factor, edge smoothness, and color consistency of the colony;

[0092] Colony counting: According to the colony identification results, the colony count is performed, and the actual number of colonies is calculated in combination with the dilution concentration;

[0093] Drug sensitivity analysis module: Use deep learning technology, especially convolutional neural network model, to detect the inhibition zone in the image; the inhibition zone is a sign that the pathogen is resistant to a specific antibiotic, and its size can reflect the sensitivity of the pathogen to the antibiotic; the system measures the size of the inhibition zone and compares it with the standard value to determine the sensitivity level of the pathogen to the specific antibiotic; finally, according to the size of the inhibition zone and the type of pathogen, the system performs drug sensitivity analysis to determine the sensitivity of the pathogen to various antibiotics and outputs the drug sensitivity results; the specific steps are as follows (the effect is shown as follows Figure 7 exhibit):

[0094] Image capture: Capture the image of the inhibition zone in the culture dish by a high-resolution camera or microscope;

[0095] Image preprocessing: grayscale conversion, image enhancement, contrast enhancement and denoising are performed on the captured images to improve image quality;

[0096] Inhibition zone detection: Use convolutional neural network model to detect inhibition zones in images;

[0097] Inhibition zone size measurement: The system measures the size of the inhibition zone and compares it with the standard value to determine the sensitivity level of the pathogen to a specific antibiotic;

[0098] Drug sensitivity analysis: Based on the size of the inhibition zone and the type of pathogen, the system performs drug sensitivity analysis to determine the sensitivity of the pathogen to various antibiotics and outputs the drug sensitivity results;

[0099] Bacteria species identification module: Use deep learning technology to distinguish between old and new colonies, and avoid misjudging dead colonies as new colonies; for streaking culture media, the system is specially designed with a terminal colony circle selection function, which can accurately identify and circle the terminal colony to improve the accuracy of bacteria species identification; draw a ruler on the image to facilitate users to observe and measure the size of the colony; use deep learning technology to identify the species of the colony, determine the type of colony, and record the information of the new species; the system can identify new species and remember and correctly mark them when they are discovered later; finally, mark the species name on the image to facilitate users to identify and distinguish different species; the specific steps are as follows:

[0100] Image capture: Capture images of colonies in culture dishes using a high-resolution camera or microscope;

[0101] Image preprocessing: grayscale conversion, image enhancement, contrast enhancement and denoising are performed on the captured images to improve image quality;

[0102] Differentiation between new and old colonies: Use deep learning technology to distinguish between new and old colonies to avoid misjudging dead colonies as new ones;

[0103] Terminal colony selection: For streaking culture medium, the system is specially designed with a terminal colony selection function, which can accurately identify and select terminal colonies;

[0104] Draw a ruler: Draw a ruler on the image to facilitate users to observe and measure the size of the colony;

[0105] Bacteria species identification: Use deep learning technology to identify bacteria in colonies, determine the type of colonies, and record new bacteria species information;

[0106] Remember new bacterial species: The system can recognize new bacterial species and remember and correctly label them when they are discovered later.

[0107] Another object of the present invention is to provide a computer device, characterized in that the computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the multi-algorithm microbial culture comprehensive identification method based on convolutional neural network.

[0108] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to execute the steps of the multi-algorithm microbial culture comprehensive identification method based on a convolutional neural network.

[0109] Another object of the present invention is to provide an information data processing terminal, which is used to implement the multi-algorithm microbial culture comprehensive identification system based on convolutional neural network.

[0110] The specific technical solution of the present invention is as follows:

[0111] The complete process of data set acquisition, annotation, training, and adjustment

[0112] 1. Dataset acquisition

[0113] In the task of microbial culture image recognition, the acquisition of the dataset is a crucial first step. The dataset should contain different types of culture media, colony morphology, color, size and other diversity to improve the generalization ability of the model. The following are the specific steps and precautions for dataset acquisition:

[0114] Equipment selection:

[0115] High-resolution camera: Choose a camera with high resolution (sensor pixels no less than 20 million pixels and area no less than 1 / 1.3 inches);

[0116] Microscope: For identification of microcolonies, a microscope is an essential tool; choose a microscope with autofocus and image stabilization capabilities;

[0117] Light conditions:

[0118] Even lighting: Make sure the dish is photographed under even light;

[0119] Light source selection: Use LED variable color temperature (2700K-7500K) cold light source with color rendering index Ra>90. The lamp beads support the emission of common ultraviolet, infrared and all visible light bands to avoid the influence of heat sources on the colony.

[0120] Shooting environment:

[0121] Dust-free environment: Shoot in a dust-free environment to avoid the impact of dust on image quality.

[0122] Stable platform: Use a stable shooting platform to avoid shaking during shooting.

[0123] Image parameter settings:

[0124] Resolution: Set an appropriate resolution to ensure clear image details.

[0125] Exposure time: Adjust the exposure time according to the lighting conditions to avoid overexposure or underexposure.

[0126] White Balance: Adjust the white balance to ensure the image colors are true.

[0127] 2. Data Annotation

[0128] Data labeling is a key step in training deep learning models. The accuracy and consistency of labeling directly affects the performance of the model. The following are the specific steps for data labeling:

[0129] Annotation tools:

[0130] Option: Use professional image annotation tools (such as LabelImg, VGG Image Annotator, etc.) for annotation.

[0131] Marking content:

[0132] Colony counting: Annotate each colony with its bounding box and category (e.g. colony type, color, etc.).

[0133] Drug sensitivity analysis: Annotate the boundary box and size of the inhibition zone.

[0134] Identification of bacterial species: mark the characteristics of the colonies (such as shape, color, edge smoothness, etc.) and mark the name of the bacterial species.

[0135] Labeling quality control:

[0136] Multiple rounds of annotation: Multiple rounds of annotation and cross-validation are performed to ensure the accuracy and consistency of annotation.

[0137] Annotation review: Review the annotation results to ensure the quality of annotation.

[0138] 3. Data Preprocessing

[0139] Data preprocessing is an important step to improve model training results. The following are the specific steps and parameter adjustments for data preprocessing:

[0140] Image reading:

[0141] Tools: Use OpenCV or PIL library to read images.

[0142] Grayscale conversion:

[0143] Purpose: Convert color images to grayscale images to reduce computational complexity while retaining key information.

[0144] Method: Use weighted averaging or directly take the green channel (because the green channel contains the most detail information).

[0145] Image Enhancement:

[0146] Purpose: To enhance the contrast and details of the image, making the colonies and background clearer.

[0147] Method: Use histogram equalization, contrast-limited adaptive histogram equalization (CLAHE), or contrast-limited adaptive histogram equalization (CLAHE).

[0148] Contrast Enhancement:

[0149] Purpose: To further enhance the contrast of the image and make the difference between the colonies and the background more obvious.

[0150] Method: Use contrast stretching, gamma correction or histogram matching.

[0151] Denoising:

[0152] Purpose: To remove noise from images and improve image quality.

[0153] Method: Use median filtering, Gaussian filtering or non-local means denoising (NLM).

[0154] 4. Feature extraction

[0155] Feature extraction is to extract useful features from the processed image for subsequent model training. The following are the specific steps and parameter adjustments for feature extraction:

[0156] area:

[0157] Purpose: To calculate the area of ​​colonies for colony counting and size analysis.

[0158] Method: Use the connected region labeling algorithm (such as 8-connected or 4-connected) of the binary image to calculate the area.

[0159] perimeter:

[0160] Purpose: To calculate the perimeter of the colony for shape analysis.

[0161] Method: Use edge detection algorithm (such as Canny operator) to extract the edge of the colony, and then calculate the length of the edge.

[0162] Form Factor:

[0163] Purpose: To calculate the shape factor of the colony for shape analysis.

[0164] Method: Calculate the shape factor using shape factor formulas (such as circularity, rectangularity, etc.).

[0165] Edge Smoothness:

[0166] Purpose: To calculate the smoothness of colony edges for shape analysis.

[0167] Method: Edge smoothness is calculated using edge smoothness formulas (such as edge curvature, edge fluctuation, etc.).

[0168] Color consistency:

[0169] Purpose: To calculate the color consistency of colonies for color analysis.

[0170] Method: Use color consistency formula (such as color variance, color histogram, etc.) to calculate color consistency.

[0171] 5. Model Training

[0172] like Figure 3 , model training is to use deep learning models such as convolutional neural networks (CNN) to achieve automatic detection and classification of colonies. The following are the specific steps of model training:

[0173] Data preparation:

[0174] Data collection: Collect a large amount of labeled colony image data to ensure the diversity and representativeness of the data set.

[0175] Data enhancement: Enhance the data (such as rotation, scaling, flipping, etc.) to increase the diversity of the data set and improve the generalization ability of the model.

[0176] Network structure design:

[0177] Convolutional layer: Design multiple convolutional layers to extract image features.

[0178] Pooling layer: Design the pooling layer to reduce the size of the feature map and reduce the computational complexity.

[0179] Fully connected layer: Design a fully connected layer to convert feature maps into classification results.

[0180] Activation function: Select a suitable activation function (such as ReLU, Sigmoid, etc.) to improve the nonlinear ability of the model.

[0181] Training strategy:

[0182] Optimizer: Select a suitable optimizer (such as Adam, SGD, etc.), adjust the learning rate and momentum parameters, and ensure model convergence.

[0183] Loss function: Select an appropriate loss function (such as cross entropy loss, mean square error, etc.) to ensure the accuracy of model training.

[0184] Regularization: Use regularization techniques (such as L2 regularization, Dropout, etc.) to prevent model overfitting.

[0185] Model Evaluation:

[0186] Validation set: Use the validation set to evaluate the performance of the model and adjust the model parameters to ensure that the model performs best on the validation set.

[0187] Test set: Use the test set to evaluate the final performance of the model and ensure the generalization ability of the model.

[0188] 6. Model Adjustment

[0189] Model adjustment is to further optimize the model based on the training and verification results to improve the performance of the model. The following are the specific steps for model adjustment:

[0190] Hyperparameter Tuning:

[0191] Learning rate: Adjust the learning rate to ensure the model convergence speed and stability.

[0192] Batch size: Adjust the batch size to balance training speed and memory usage.

[0193] Number of training rounds: Adjust the number of training rounds to avoid overfitting or underfitting.

[0194] Network structure adjustment:

[0195] Convolutional layers: Increase or decrease the number of convolutional layers, and adjust the kernel size and stride.

[0196] Pooling Layer: Adjust the type and parameters of the pooling layer, such as pooling window size and stride.

[0197] Fully connected layers: Increase or decrease the number of fully connected layers and adjust the number of neurons.

[0198] Data Augmentation:

[0199] Enhancement method: Add or adjust data enhancement methods, such as rotation, scaling, flipping, brightness adjustment, contrast adjustment, etc.

[0200] Enhancement Strength: Adjust the strength of data augmentation to ensure that the augmented data is still representative.

[0201]

[0202]

[0203] Part I: Complete steps of the method

[0204] Action Flow

[0205] Image preprocessing

[0206] Execution body: algorithm module

[0207] Step description: Determine the boundary of the culture medium: First, accurately identify the boundary of the culture medium from the image, which is the basis for subsequent analysis. Use the edge detection algorithm (Canny operator) to identify the outline of the culture medium. Figure 4 ,The microbial colony identification diagram shows the identification results of the culture medium boundaries.

[0208] Cropping: Based on the determined culture medium boundaries, unnecessary background parts are cropped out to keep only the area containing the colonies.

[0209] Convert to grayscale: In order to facilitate subsequent feature extraction and processing, color images are usually converted to grayscale images.

[0210] Image preprocessing CNN preliminary image convolution: Use convolutional neural network (CNN) to perform preliminary processing on the grayscale image to enhance the outline of the colony and other key features. Figure 6 ,The CNN preliminary preprocessing image shows the image effect after CNN processing.

[0211] Colony feature enhancement algorithm: Apply specific algorithms to further enhance the characteristics of colonies for more accurate identification and classification. Figure 5 ,The colony feature enhancement diagram shows the effect of the colony features after enhancement.

[0212] Colony adhesion judgment execution body: algorithm module

[0213] Step description: Canny operator edge detection: Use the Canny edge detection algorithm to detect the edge information of the colony.

[0214] Determine whether the colonies are adhered: Determine whether there is adhesion between colonies by analyzing the edge detection results.

[0215] Segmentation and feature enhancement execution body: algorithm module

[0216] Step Description: Watershed segmentation algorithm: If the colonies are contiguous, they need to be segmented using the watershed segmentation algorithm to ensure that each colony can be analyzed and identified independently.

[0217] Feature enhancement processing: For colonies that have been separated or non-adhesive, feature enhancement processing can be performed directly to ensure that the characteristics of each colony are obvious enough and easy to distinguish.

[0218] Memory library comparison and new strain identification execution body: algorithm module

[0219] Step description: Extract memory comparison: compare the characteristics of the currently processed colony with the characteristics of known bacterial species stored in the memory bank.

[0220] Determine whether it is a new species of fungus: If the current colony characteristics do not match any known species in the memory library, then it may be a new species of fungus. At this time, the system will mark it as a "new species of fungus".

[0221] General labeling colony results: For all known bacterial species, the system will generate a general labeling result, including the number, morphology, size and other information of the colonies.

[0222] Result output execution body: algorithm module

[0223] Step Description:

[0224] The system summarizes the results of colony counts, drug sensitivity analysis, and bacterial species identification.

[0225] The results are output to the user interface, which displays a detailed analysis report.

[0226] Users can export reports for further analysis or archiving.

[0227] Part II: Inventive Points of the Present Invention

[0228] The core invention of the present invention is to realize the automation, efficiency and accuracy of microbial culture analysis by integrating multiple advanced technologies. The specific invention points are as follows: Multi-mode integration: The system integrates three modes of colony counting, drug sensitivity analysis and bacterial species identification to meet different experimental needs.

[0229] Users can select the corresponding mode according to the purpose of the experiment to achieve multifunctional integrated operation. Application of deep learning technology: The convolutional neural network (CNN) model is used for colony detection, inhibition zone detection and bacterial species identification, which significantly improves the identification accuracy.

[0230] After being trained with a large amount of data, the CNN model is able to process complex images and tiny colonies, reducing misjudgments and missed judgments. Optimization of image processing algorithms: Through preprocessing steps such as grayscale conversion, image enhancement, contrast enhancement, and denoising, the image quality is improved to provide a clear data basis for subsequent analysis.

[0231] The regional growing algorithm and watershed algorithm are used to process the adherent colonies to ensure the accuracy of colony counting. Feature extraction and analysis: Extract multiple features of the colonies, including area, perimeter, shape factor, edge smoothness, and color consistency, to provide comprehensive information for colony identification and classification.

[0232] The actual number of colonies is calculated in combination with the dilution concentration to ensure the accuracy of the counting results. Identification and memory of new strains: The system can identify and record the information of new strains, and remember and correctly mark the identification when it is discovered later, improving the continuity and accuracy of strain identification.

[0233] For streaking culture medium, a terminal colony selection function is specially designed to improve the accuracy of bacterial species identification.

[0234] The entire analysis process is automated, reducing manual intervention, significantly improving analysis efficiency, and meeting the needs of high-throughput experiments. The system can handle different types of culture media and colonies, with high adaptability and a wide range of applications.

[0235] Section 3: Evidence of beneficial effects

[0236] In order to verify the beneficial effects of the method of the present invention, we conducted a number of experimental comparisons and collected relevant data and effect curves. The following are the main experimental results and evidence:

[0237] Colony species identification

[0238] Experimental results:

[0239]

[0240] Colony count

[0241] Experimental results:

[0242]

[0243]

[0244] Inhibition zone determination

[0245] Experimental results:

[0246]

[0247] Through the above experimental comparison data and effect curves, it can be clearly seen that the method of the present invention has significant advantages in colony counting, drug sensitivity analysis and bacterial species identification. The present invention not only improves the accuracy and efficiency of the analysis, but also reduces manual intervention and errors, meets the needs of high-throughput experiments, and has broad application prospects and important practical significance.

[0248] 1. Specific application fields or related products of the present invention.

[0249] 1. Biopharmaceuticals and vaccine production

[0250] Application areas: Microbial cultivation is a key link in the production of biopharmaceuticals and vaccines. The method of the present invention can achieve real-time monitoring and accurate identification of the microbial cultivation process, ensuring the stability of the cultivation process and product quality.

[0251] Related products: Microbial culture monitoring system, automated microbial testing equipment, vaccine production quality control system.

[0252] 2. Food safety testing

[0253] Application field: In the field of food safety, microbial contamination is one of the main reasons for food spoilage and health hazards. The present invention can be used to quickly detect harmful microorganisms (such as Escherichia coli, Salmonella, etc.) in food to ensure food safety.

[0254] Related products: Food microbiological detectors, food safety rapid detection systems, and food production line microbiological monitoring equipment.

[0255] 3. Environmental monitoring and pollution control

[0256] Application areas: In environmental monitoring, the types and quantity of microorganisms are important indicators for assessing water quality, soil and air pollution. The present invention can be used to identify the types of microorganisms in the environment, help assess the degree of pollution and formulate treatment plans.

[0257] Related products: Environmental microbiology detectors, water quality microbiology monitoring systems, soil microbiology analysis equipment.

[0258] 4. Medical treatment and clinical diagnosis

[0259] Application field: In the medical field, microbial culture is an important means of diagnosing infectious diseases. The present invention can be used to automatically identify pathogenic microorganisms (such as bacteria, fungi, etc.) to improve diagnostic efficiency and accuracy.

[0260] Related products: Automated microbial culture and identification system, clinical microbiology testing equipment, infectious disease diagnosis assistance system.

[0261] 5.Agriculture and Ecology Research

[0262] Application areas: In agricultural and ecological research, microorganisms play an important role in soil improvement, plant growth and ecosystem balance. The present invention can be used to analyze soil or plant rhizosphere microbial communities, providing a scientific basis for agricultural production and ecological protection.

[0263] Related products: soil microbial analyzer, plant rhizosphere microbial detection system, agricultural microbial research platform.

[0264] 6. Industrial fermentation and bioengineering

[0265] Application areas: In industrial fermentation and bioengineering, microbial cultivation is a key step in producing enzymes, organic acids, biofuels and other products. The present invention can be used to optimize the fermentation process and improve production efficiency and product quality.

[0266] Related products: industrial fermentation process monitoring system, bioengineering microbial detection equipment, fermentation process optimization platform.

[0267] 2. Relevant evidence of the technical effects obtained by the embodiments of the present invention.

[0268] 1. Experimental data and comparative analysis

[0269] Experimental design: Under laboratory conditions, the method of the present invention was used to identify different types of microbial culture images and compared with traditional methods (such as manual microscope observation and single algorithm recognition).

[0270] Experimental results:

[0271] Identification accuracy: The method of the present invention has an accuracy of more than 95% in microorganism identification, while the accuracy of traditional methods is only 70%-80%. The identification rates of similar products are all greater than 93%.

[0272] Recognition speed: The method of the present invention can complete the recognition of a single image within 5 seconds, while the traditional method takes 5-10 minutes. The recognition time of similar products varies from 5 to 60 seconds depending on the product positioning.

[0273] Robustness: Under complex background and lighting conditions, the method of the present invention can still maintain a high recognition accuracy, while traditional methods perform poorly under complex conditions. Similar products have special requirements for light sources, backgrounds and even culture media.

[0274] 2. Practical application cases

[0275] Case 1: If Figure 7 , it can count the colonies on the diluted coating plate in the experimental environment and give the analysis results at the same time. If the culture data of the culture environment is given, more results can be output to give suggestions for the culture environment.

[0276] Case 2: If Figure 8 , the analysis and identification of the inhibition zone can be carried out in the laboratory environment, and the judgment and conclusion can be made through automatic identification of the inhibition zone and colony morphology.

[0277] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0278] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A multi-algorithm microbial culture comprehensive identification method based on convolutional neural network, characterized in that: The following steps are involved: Step 1, data set acquisition; In the task of microbial culture image recognition, the acquisition of a dataset is a crucial first step; the dataset should contain different types of culture media, colony morphology, color, and size diversity; Step 2: Data labeling; Data annotation is a key step in training deep learning models; Step 3, data preprocessing; Data preprocessing is an important step to improve model training results; Step 4, feature extraction; Feature extraction is to extract useful features from the processed images for subsequent model training; Step 5: Model training; Model training is carried out using a convolutional neural network deep learning model; Step 6, model adjustment; Model adjustment is to further optimize the model based on the training and validation results.

2. The multi-algorithm microbial culture comprehensive identification method based on convolutional neural network as claimed in claim 1, characterized in that: The dataset obtains: In the acquisition stage of microbial culture images, it is crucial to choose high-quality equipment; first, use a high-resolution camera to capture clear images; for the identification of tiny colonies, it is particularly important to equip the microscope with autofocus and image stabilization functions to ensure accurate capture of details; In addition, LED cold light sources are selected to provide uniform lighting. The light source has variable color temperature and color rendering index Ra>90, and supports ultraviolet, infrared and full-spectrum visible light, which helps to improve the contrast and clarity of the image. Shooting in a dust-free environment and using a stable platform to avoid shaking ensures that the captured images have high quality. After image acquisition is completed, the data needs to be professionally annotated to support model training; Use professional image annotation tools to classify and annotate the colonies and related features in the culture dish; The specific contents include: drawing a bounding box for each colony and marking its category for colony counting; marking the bounding box and size of the inhibition zone to support drug sensitivity analysis; extracting the morphological characteristics of the colony and marking the species name to complete the species identification; through these detailed annotations, the model can obtain accurate supervision information, thereby improving the recognition accuracy; To ensure the accuracy of the labeled data, strict quality control is required; the labeling process should include multiple rounds of labeling, and use cross-validation methods to verify the consistency of the labeling; after the labeling is completed, the labeling results are reviewed by a professional team to ensure the accuracy and consistency of the labeling of each colony, inhibition zone and strain; this multi-level quality control mechanism can effectively improve the reliability of labeled data and provide high-quality data support for the training of deep learning models.

3. The multi-algorithm microbial culture comprehensive identification method based on convolutional neural network as claimed in claim 1, characterized in that: The data preprocessing: Data preprocessing is an important step to improve model training results; the following are the specific steps and parameter adjustments for data preprocessing: Image reading: Use OpenCV or PIL library to read images; Grayscale conversion: use weighted average method or directly take the green channel; Image enhancement: using histogram equalization, adaptive histogram equalization, or contrast-limited adaptive histogram equalization; Contrast enhancement: using contrast stretching, gamma correction or histogram matching; Denoising: Use median filtering, Gaussian filtering, or non-local means denoising.

4. The multi-algorithm microbial culture comprehensive identification method based on convolutional neural network as claimed in claim 1, characterized in that: The feature extraction: Feature extraction is to extract useful features from the processed image for subsequent model training; the following are the specific steps and parameter adjustments for feature extraction: Area: The area is calculated using the connected component labeling algorithm of binary images; Perimeter: Use edge detection algorithm to extract the edge of the colony and then calculate the length of the edge; Shape Factor: Calculate the shape factor using the shape factor formula; Edge smoothness: Use the edge smoothness formula to calculate edge smoothness; Color Consistency: Calculates color consistency using the color consistency formula.

5. The multi-algorithm microbial culture comprehensive identification method based on convolutional neural network as claimed in claim 1, characterized in that: The model training: Model training is performed using a convolutional neural network deep learning model to achieve automatic detection and classification of colonies. The following are the specific steps for model training: Data preparation: Data collection: Collect a large amount of labeled colony image data to ensure the diversity and representativeness of the data set; Data enhancement: Enhance the data to increase the diversity of the data set and improve the generalization ability of the model; Network structure design: Convolutional layer: Design multiple convolutional layers to extract image features; Pooling layer: Design the pooling layer to reduce the size of the feature map and reduce the computational complexity; Fully connected layer: Design a fully connected layer to convert feature maps into classification results; Activation function: Select a suitable activation function to improve the nonlinear ability of the model; Training strategy: Optimizer: Select a suitable optimizer and adjust the learning rate and momentum parameters to ensure model convergence; Loss function: Select an appropriate loss function to ensure the accuracy of model training; Regularization: Use regularization techniques to prevent model overfitting; Model Evaluation: Validation set: Use the validation set to evaluate the performance of the model and adjust the model parameters to ensure that the model performs best on the validation set; Test set: Use the test set to evaluate the final performance of the model.

6. The multi-algorithm microbial culture comprehensive identification method based on convolutional neural network as claimed in claim 1, characterized in that: The model adjustments: Model adjustment is to further optimize the model based on the training and verification results to improve the performance of the model; the following are the specific steps for model adjustment: Hyperparameter Tuning: Learning rate: Adjust the learning rate to ensure the convergence speed and stability of the model; Batch size: adjust the batch size to balance training speed and memory usage; Number of training rounds: adjust the number of training rounds to avoid overfitting or underfitting; Network structure adjustment: Convolutional layers: increase or decrease the number of convolutional layers, adjust the convolution kernel size and stride; Pooling layer: adjust the type and parameters of the pooling layer, such as pooling window size and stride; Fully connected layers: Increase or decrease the number of fully connected layers and adjust the number of neurons; Data Augmentation: Enhancement method: add or adjust data enhancement methods, such as rotation, scaling, flipping, brightness adjustment, contrast adjustment, etc.; Enhancement strength: Adjust the strength of data enhancement to ensure that the enhanced data is still representative; Mark the name of the strain: Mark the name of the strain on the image to facilitate users to identify and distinguish different strains.

7. A multi-algorithm microbial culture comprehensive identification system based on a convolutional neural network that implements the multi-algorithm microbial culture comprehensive identification method based on a convolutional neural network as claimed in any one of claims 1 to 6, characterized in that: The multi-algorithm microbial culture comprehensive identification system based on convolutional neural network includes: Colony counting module: Based on the colony identification results, the system counts the colonies and calculates the actual number of colonies in combination with the dilution concentration. This step ensures the accuracy of the counting results and can reflect the actual colony density. The specific steps are as follows: Image capture: Capture images of colonies in culture dishes using a high-resolution camera or microscope; Image preprocessing: grayscale conversion, image enhancement, contrast enhancement and denoising are performed on the captured images to improve image quality; Colony detection and identification: Use the convolutional neural network model to detect and identify colonies in the processed images and distinguish the boundaries between different colonies; Feature extraction: Extract features such as area, perimeter, shape factor, edge smoothness, and color consistency of the colony; Colony counting: According to the colony identification results, the colony count is performed, and the actual number of colonies is calculated in combination with the dilution concentration; Drug sensitivity analysis module: Use deep learning technology, especially convolutional neural network model, to detect the inhibition zone in the image; the inhibition zone is a sign that the pathogen is resistant to a specific antibiotic, and its size can reflect the sensitivity of the pathogen to the antibiotic; the system measures the size of the inhibition zone and compares it with the standard value to determine the sensitivity level of the pathogen to the specific antibiotic; finally, according to the size of the inhibition zone and the type of pathogen, the system performs drug sensitivity analysis to determine the sensitivity of the pathogen to various antibiotics and outputs the drug sensitivity results; the specific steps are as follows: Image capture: Capture the image of the inhibition zone in the culture dish by a high-resolution camera or microscope; Image preprocessing: grayscale conversion, image enhancement, contrast enhancement and denoising are performed on the captured images to improve image quality; Inhibition zone detection: Use convolutional neural network model to detect inhibition zones in images; Inhibition zone size measurement: The system measures the size of the inhibition zone and compares it with the standard value to determine the sensitivity level of the pathogen to a specific antibiotic; Drug sensitivity analysis: Based on the size of the inhibition zone and the type of pathogen, the system performs drug sensitivity analysis to determine the sensitivity of the pathogen to various antibiotics and outputs the drug sensitivity results; Bacteria species identification module: Use deep learning technology to distinguish between old and new colonies, and avoid misjudging dead colonies as new colonies; for streaking culture media, the system is specially designed with a terminal colony circle selection function, which can accurately identify and circle the terminal colony to improve the accuracy of bacteria species identification; draw a ruler on the image to facilitate users to observe and measure the size of the colony; use deep learning technology to identify the species of the colony, determine the type of colony, and record the information of the new species; the system can identify new species and remember and correctly mark them when they are discovered later; finally, mark the species name on the image to facilitate users to identify and distinguish different species; the specific steps are as follows: Image capture: Capture images of colonies in culture dishes using a high-resolution camera or microscope; Image preprocessing: grayscale conversion, image enhancement, contrast enhancement and denoising are performed on the captured images to improve image quality; Differentiation between new and old colonies: Use deep learning technology to distinguish between new and old colonies to avoid misjudging dead colonies as new ones; Terminal colony selection: For streaking culture medium, the system is specially designed with a terminal colony selection function, which can accurately identify and select terminal colonies; Draw a ruler: Draw a ruler on the image to facilitate users to observe and measure the size of the colony; Bacteria species identification: Use deep learning technology to identify bacteria in colonies, determine the type of colonies, and record new bacteria species information; Remember new bacterial species: The system can recognize new bacterial species and remember and correctly label them when they are discovered later.

8. A computer device, characterized in that: The computer device includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the multi-algorithm microbial culture comprehensive identification method based on a convolutional neural network as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the multi-algorithm microbial culture comprehensive identification method based on a convolutional neural network as described in any one of claims 1 to 6.

10. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the multi-algorithm microbial culture comprehensive identification system based on convolutional neural network as described in claim 7.

Images