Method and system for rapidly identifying pathogenic bacteria

Abstract

The invention discloses a rapid pathogen identification method and system, and belongs to the technical field of medical information processing. According to the method, space-time growth discrimination operation, multi-scale structure perception operation and microscopic focusing classification operation are combined, the dynamic trend of bacterial colonies from appearing to diffusion is accurately captured, low-contrast textures of micron-scale initial spots and the complex structure of millimeter-scale mature bacterial colonies are considered, a refined space context correlation model is established, and the dynamic trend of the bacterial colonies from appearing to diffusion is obtained. The method solves the problems that a traditional single-frame analysis method is insufficient in continuous modeling of the growth process and early-stage features are confused, the detection sensitivity of early-stage minute colonies is improved, rapid and accurate identification of pathogenic bacteria types is achieved, and the identification requirements in clinical rapid detection and complex microbial environments are met.

Description

Technical Field

[0001] The present invention relates to the field of medical information processing technology, and more particularly to a method and system for rapid identification of pathogens. Background Art

[0002] The rapid identification of pathogens is crucial in clinical diagnosis and treatment, public health, and biosafety. Its core goal is to accurately interpret the growth status of colonies and identify their species based on microscopic image sequences. Although traditional methods such as microbial culture, biochemical tests, and PCR testing are accurate, they generally have problems such as long detection cycles, complex processes, high dependence on personnel, and difficulty in achieving high-throughput screening. In recent years, deep learning methods have been widely introduced into the task of pathogen image recognition. Common ideas include extracting colony color, edge, and texture features based on convolutional neural networks, and using deep neural networks for full-image modeling. Some studies have also attempted to introduce the Transformer architecture to improve the perception of local colony structure and growth trends through the attention mechanism.

[0003] Although the above methods have excellent pathogen classification detection capabilities, they still have the following three problems: (1) Lack of spatiotemporal modeling capabilities for the colony growth process: Bacterial growth has significant spatiotemporal continuity. Existing methods are mostly based on single-frame image feature analysis, and lack effective modeling of the colony's appearance process from nothing to something, its diffusion speed, and its spatiotemporal distribution patterns. (2) Insufficient multi-scale structure perception capabilities: The sizes of colonies at different growth stages vary greatly (from micron-level initial spots to millimeter-level mature colonies). Existing single-scale feature extraction models are difficult to take into account the feature differences at different stages, and there are problems with insufficient representation of the low-contrast texture of early tiny colonies and the complex structure of mature colonies. (3) Weak microscopic feature discrimination capabilities: Early classification of bacterial species relies on accurate modeling of microstructures such as fuzzy edges, diffusion patterns, and texture changes. Traditional classification models are difficult to fully utilize these detailed features, resulting in insufficient classification accuracy.

[0004] Therefore, how to propose a rapid pathogen identification method and system to improve the efficiency, speed and automation level of pathogen detection is an urgent problem that needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method and system for rapid identification of pathogens, which provides a high-precision, low-manual-dependency solution for automated microbial monitoring and analysis.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] In one aspect, the present invention provides a method for rapid identification of pathogens, comprising the following steps:

[0008] Acquire a bacterial colony time-series image sequence, perform a spatiotemporal growth discrimination operation on the bacterial colony time-series image sequence, and obtain spatiotemporal features by modeling the growth dynamics of the bacterial colony in the time dimension and the morphological distribution in the space dimension;

[0009] Performing a multi-scale structural perception operation on the spatiotemporal features to obtain response features at multiple scales, and obtaining a colony significant region feature map based on the response features at multiple scales;

[0010] Constructing a large visual-language model, inputting the colony significant region feature map into the large visual-language model, determining whether there are growing colonies, and outputting the corresponding growth interpretation result; if the growth interpretation result is growth, performing bacterial species identification; if the growth interpretation result is no growth, terminating subsequent analysis;

[0011] Input training data and optimize the model parameters in the strain identification process based on the loss function;

[0012] The trained model parameters are persisted and deployed to the actual system to achieve rapid identification of pathogens.

[0013] Preferably, a spatiotemporal growth discrimination operation is performed on the bacterial colony time image sequence, and spatiotemporal features are obtained by modeling the growth dynamics of the bacterial colony in the time dimension and the morphological distribution in the space dimension, including:

[0014] The bacterial colony time series image sequence As input, I t ∈R H×W×C , T represents the number of image frames, H, W, and C represent the image height, width, and number of channels, respectively; the time series image is composed of bacterial colony images at different times during the culture process;

[0015] The spatial features of each frame are obtained by modeling the change trend between adjacent frames in the time series image and the image spatial structure distribution and growth characteristics

[0016] Splicing the spatial features and growth characteristics Get spatiotemporal characteristics

[0017] Preferably, performing a multi-scale structure perception operation on the spatiotemporal features to obtain response features at multiple scales, and obtaining a colony significant region feature map based on the response features at multiple scales, including:

[0018] Multi-scale dilated convolution operations are introduced in parallel to process the spatiotemporal features. By setting different dilation rates d iThe hole convolution obtains the response features at multiple scales

[0019]

[0020] Among them, Concat(·) represents the concatenation operation; Indicates that the convolution kernel size is 3×3 and the void rate is d i The dilated convolution operation; i = 1, 2, ..., S; S is the set scale number;

[0021] Response characteristics at multiple scales Perform pixel-by-pixel significance scoring to obtain the colony significance region feature map S t :

[0022]

[0023] Among them, * is the convolution operation, σ is the Sigmoid activation function, W s 、b s are learnable parameters.

[0024] Preferably, the colony significant region feature map is input into the visual-language large model to determine whether there is a growing colony and output a corresponding growth interpretation result, including:

[0025] The colony salient region feature map is input into the vision-language model, and the colony growth probability p is output. grow ;

[0026] Set the growth judgment threshold τ grow , if p grow >τ grow , it is judged that there are growing colonies and the growth judgment result is output as growth; otherwise, it is judged that there are no growing colonies and the growth judgment result is output as no growth.

[0027] Preferably, if the growth interpretation result is growth, bacterial species identification is performed, including:

[0028] Perform a microscopic focusing classification operation on the significant region feature map and output the bacterial species identification result. The process is as follows:

[0029] By introducing a local perception mechanism, the edge morphology, texture, fuzzy boundary and tiny diffusion structure in the salient region feature map are fine-grained modeled to output micro features F det , which is expressed as follows:

[0030]

[0031] Among them, MicroFocus(·) represents the micro-focus classification operation; ⊙ represents element-wise multiplication;

[0032] The microscopic features F det Input to the target detection head to realize the positioning of different colonies, predict the probability distribution of each colony belonging to each type of bacterial species, and output the recognition results:

[0033] B i =Regressor(Softmax(F det ))=[x i ,y i ,w i ,h i ]

[0034] y type(i) =argmax(Softmax(W cls ·F det +b cls ))

[0035] Among them, Regressor(·) is a bounding box regression function, which outputs the corresponding multiple colonies containing the center point (x i ,y i ) and width and height (w i ,h i )’s predicted bounding box parameters B i , W cls 、b cls is the trainable parameter of the fully connected layer, Softmax(·) is the normalized probability function, and the output of the bacterial species recognition result y type(i) The bacterial species category with the highest probability in each bounding box.

[0036] Preferably, the loss function includes the vision-language large model loss function Bounding Box Regression Loss Function and the micro-focus classification loss function

[0037] The vision-language large model loss function is the cross entropy loss function, which is used to constrain the consistency between the colony growth judgment results output by the vision-language large model and the true label;

[0038] The bounding box regression loss function Used to constrain the regression accuracy of the detection box coordinates and the true bounding box;

[0039] The micro-focused classification loss function The cross entropy loss function is used to constrain the matching relationship between the strain category prediction results output by the micro-focused classification operation and the actual strain label;

[0040] The visual-language model loss function The bounding box regression loss function and the microscopic focused classification loss function The average value will be used as the final loss value to optimize the model parameters in the process of strain identification.

[0041] On the other hand, the present invention also discloses a rapid pathogen identification system for implementing the above-mentioned rapid pathogen identification method, comprising:

[0042] A spatiotemporal growth discrimination module obtains a bacterial colony time-series image sequence, performs a spatiotemporal growth discrimination operation on the bacterial colony time-series image sequence, and obtains spatiotemporal features by modeling the growth dynamics of the bacterial colony in the time dimension and the morphological distribution in the space dimension;

[0043] A multi-scale structure perception module performs a multi-scale structure perception operation on the spatiotemporal features to obtain response features at multiple scales, and obtains a colony significant region feature map based on the response features at multiple scales;

[0044] The growth interpretation module constructs a visual-language model, inputs the colony significant region feature map into the visual-language model, determines whether there are growing colonies, and outputs the corresponding growth interpretation result; if the growth interpretation result is growth, the bacterial species is identified; if the growth interpretation result is no growth, the subsequent analysis is terminated;

[0045] Parameter training module, inputs training data and optimizes model parameters in the strain identification process based on the loss function;

[0046] The model deployment module persists the trained model parameters and deploys them to the actual system to achieve rapid identification of pathogens.

[0047] Preferably, the growth interpretation module includes a microscopic focus classification unit, which performs a microscopic focus classification operation on the significant region feature map and outputs a bacterial species identification result.

[0048] It can be seen from the above technical solutions that, compared with the prior art, the present invention provides a method and system for rapid identification of pathogens, which has the following beneficial effects:

[0049] (1) By adapting the spatiotemporal growth discrimination operation and multi-scale structure perception operation of the vision-language large model, the growth trend of the colony is identified from the continuous image sequence, solving the problems of small size, low contrast and multi-scale missed detection in the early stage, and achieving fast and accurate growth interpretation.

[0050] (2) Microscopic focused classification operations enhance the perception of early weak features (texture, edge, microstructure), improve the accuracy of colony classification, overcome the difficulty of early feature recognition, and ensure the efficiency and accuracy of colony species identification.

[0051] (3) The growth interpretation agent (which takes continuously collected colony time-series image sequences as input, introduces a large visual-language model as the basic perception unit, and integrates spatiotemporal growth discrimination operations and multi-scale structure perception operations) collaborates with the colony recognition agent (which introduces microscopic focus classification operations) to achieve full process automation from growth judgment to species recognition, reduce manual intervention, improve efficiency and consistency, and adapt to clinical rapid diagnosis scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0053] Figure 1 A flow chart of the method provided by the present invention;

[0054] Figure 2 This is a diagram of the system architecture provided by the present invention. DETAILED DESCRIPTION

[0055] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0056] On the one hand, if Figure 1 As shown, the present invention proposes a method for rapid identification of pathogens, comprising the following steps:

[0057] S1. Obtain a time-series image sequence of the colony and perform spatiotemporal growth discrimination on the time-series image sequence. By modeling the growth dynamics of the colony in the temporal dimension and the morphological distribution in the spatial dimension, spatiotemporal features are obtained, including:

[0058] S1.1: Time-sequential image sequences of colonies As input, I t ∈R H×W×C , T represents the number of image frames, H, W, and C represent the image height, width, and number of channels, respectively; the colony time series image is composed of colony images at different times during the cultivation process.

[0059] S1.2: Perform spatiotemporal growth discrimination on the bacterial colony time image sequence, and obtain the spatial features of each frame by modeling the change trend between adjacent frames and the image spatial structure distribution. and growth characteristics

[0060]

[0061] Among them, F space (·) is the spatial structure feature extractor, F temporal (·) is the time series modeling module, and Δt is the time window length.

[0062] S1.2: Stitching spatial features and growth characteristics Get spatiotemporal characteristics

[0063]

[0064] Among them, Concat(·) represents the concatenation operation.

[0065] S2. Perform multi-scale structural perception operations on spatiotemporal features to obtain multi-scale response features. Based on the multi-scale response features, a colony salient region feature map is obtained, including:

[0066] S2.1: Spatiotemporal characteristics Perform multi-scale structure perception operations and introduce multi-scale dilated convolution operations in parallel to model the structure of the colony area in detail at multiple receptive field scales.

[0067] By setting different void ratios d i The dilated convolution is used to obtain the response features at multiple scales.

[0068] in, Indicates that the convolution kernel size is 3×3 and the void rate is d i The dilated convolution operation is performed; i = 1, 2, ..., S; S is the set scale number.

[0069] S2.2: Yes Perform pixel-by-pixel significance scoring to obtain the colony significance region feature map S t :

[0070]

[0071] Among them, * is the convolution operation, σ is the Sigmoid activation function, W s 、b sare learnable parameters.

[0072] S3. Build a large visual-language model, input the colony saliency region feature map into the large visual-language model, determine whether there are growing colonies, and output the corresponding growth judgment result; if the growth judgment result is growth, perform bacterial species identification; if the growth judgment result is no growth, terminate subsequent analysis.

[0073] Specifically, the specific process of this step is as follows:

[0074] S3.1: Construct a semantic prompt template Prompt: "Judge whether there is colony growth in this area" and map the colony saliency feature map S t It is embedded as input into the visual-language model VLM to perform the growth interpretation task, as shown below:

[0075] p grow =Softmax(VLM(Prompt,S t ))

[0076] Among them, VLM(·) represents the input into the visual-language model VLM; Softmax(·) is the normalized probability function, which outputs the predicted probability of the two semantic categories of "no growth" and "growth" (i.e., the colony growth probability p grow ).

[0077] S3.2: Setting the growth judgment threshold τ grow ∈[0,1]. If p grow [1]>τ grow , it is determined that there are signs of colony growth in the current time series image, and it is judged that there are growing colonies. The growth judgment result is output as growth, and the bacterial species identification is performed; otherwise, it is considered that there are no growing colonies, and the growth judgment result is output as no growth, and subsequent analysis is terminated.

[0078] Furthermore, if the growth interpretation result is growth, bacterial species identification is performed, including:

[0079] Perform microscopic focusing classification on the salient region feature map and output the bacterial species identification results. The process is as follows:

[0080] By introducing the local perception mechanism, the edge morphology, texture, fuzzy boundary and small diffusion structure in the salient region feature map are fine-grained modeled to output the micro-feature F det , which is expressed as follows:

[0081]

[0082] Among them, MicroFocus(·) represents the micro-focus classification operation; ⊙ represents element-wise multiplication;

[0083] The microscopic feature F det Input to the target detection head to locate different colonies, predict the probability distribution of each colony belonging to each species, and output the recognition results:

[0084] B i =Regressor(Softmax(F det ))=[x i ,y i ,w i ,h i ]

[0085] y type(i) =argmax(Softmax(W cls ·F det +b cls ))

[0086] Among them, Regressor(·) is a bounding box regression function, which outputs the corresponding multiple colonies containing the center point (x i ,y i ) and width and height (w i ,h i )’s predicted bounding box parameters B i , W cls 、b cls is the trainable parameter of the fully connected layer, Softmax(·) is the normalized probability function, and the output of the bacterial species recognition result y type(i) The bacterial species category with the highest probability in each bounding box.

[0087] S4. Input training data and optimize the model parameters in the strain identification process based on the loss function.

[0088] The loss function includes the vision-language model loss function Bounding Box Regression Loss Function and the micro-focus classification loss function

[0089] Loss function of large vision-language model is a cross-entropy loss function, which is used to constrain the consistency between the colony growth judgment results output by the vision-language large model and the true label, thereby improving the system's ability to discriminate the early growth status of the colony.

[0090] Bounding Box Regression Loss Function It is used to constrain the regression accuracy of the detection box coordinates and the true bounding box, thereby improving the accuracy of colony spatial positioning.

[0091] Micro-focused classification loss function The cross entropy loss function is used to constrain the matching relationship between the bacterial species prediction results output by the microscopic focused classification operation and the actual bacterial species labels, ensuring the accuracy of colony species identification.

[0092] The loss function of the visual-language model Bounding Box Regression Loss Function and the micro-focus classification loss function The average value will be used as the final loss value to optimize the model parameters in the process of strain identification.

[0093] During the specific implementation process, the colony time-series image sequences are automatically collected by the embedded camera module in an incubator under standardized conditions and transmitted to the central server via the intranet. There are more than 50 species of fungi, and each colony contains more than 200 sets of time-series image sequences. The training process is carried out on a server equipped with four NVIDIA H100 GPU graphics cards. The method proposed in this embodiment is trained using the AdamW optimizer, and the learning rate is set to 6e -4 , the batch size is set to 4. The window length Δt in the spatiotemporal growth discrimination operation is set to 3; the number of scales S is set to 3, and the dilation rate d of the multi-scale dilated convolution operation is set to 3. i ∈{1,2,3}; Growth judgment threshold τ grow Set to 0.7. After 120,000 iterations, the model finally completed training.

[0094] S5. Persist the trained model parameters and deploy them to the actual system to achieve rapid identification of pathogens.

[0095] On the other hand, reference Figure 2 The present invention also discloses a rapid pathogen identification system for implementing the above-mentioned rapid pathogen identification method, comprising:

[0096] The spatiotemporal growth discrimination module obtains a time-series image sequence of the colony, models the growth dynamics of the colony in the time dimension and the morphological distribution in the spatial dimension based on the colony time image sequence, and obtains the spatiotemporal characteristics;

[0097] The multi-scale structure perception module performs multi-scale dilated convolution operations on spatiotemporal features to obtain multi-scale response features and obtains a colony salient region feature map based on the multi-scale response features.

[0098] The growth interpretation module builds a large visual-linguistic model, inputs the colony significant region feature map into the large visual-linguistic model, determines whether there are growing colonies, and outputs the corresponding growth interpretation result; if the growth interpretation result is growth, the bacterial species is identified; if the growth interpretation result is no growth, the subsequent analysis is terminated;

[0099] Parameter training module, inputs training data and optimizes model parameters in the strain identification process based on the loss function;

[0100] The model deployment module persists the trained model parameters and deploys them to the actual system to achieve rapid identification of pathogens.

[0101] Preferably, the growth interpretation module includes a microscopic focus classification unit, which performs a microscopic focus classification operation on the significant area feature map and outputs a bacterial species identification result.

[0102] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0103] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for rapid identification of pathogens, characterized in that: The following steps are involved: Acquire a bacterial colony time-series image sequence, perform a spatiotemporal growth discrimination operation on the bacterial colony time-series image sequence, and obtain spatiotemporal features by modeling the growth dynamics of the bacterial colony in the time dimension and the morphological distribution in the space dimension; Performing a multi-scale structural perception operation on the spatiotemporal features to obtain response features at multiple scales, and obtaining a colony significant region feature map based on the response features at multiple scales; Constructing a large visual-language model, inputting the colony significant region feature map into the large visual-language model, determining whether there are growing colonies, and outputting the corresponding growth interpretation result; if the growth interpretation result is growth, performing bacterial species identification; if the growth interpretation result is no growth, terminating subsequent analysis; Input training data and optimize the model parameters in the strain identification process based on the loss function; The trained model parameters are persisted and deployed to the actual system to achieve rapid identification of pathogens.

2. A method for rapid identification of pathogens according to claim 1, characterized in that: Obtain a colony time-series image sequence, perform a spatiotemporal growth discrimination operation on the colony time-series image sequence, and obtain spatiotemporal features by modeling the growth dynamics of the colony in the time dimension and the morphological distribution in the space dimension, including: The bacterial colony time series image sequence As input, I t ∈R H×W×C , T represents the number of image frames, H, W, and C represent the image height, width, and number of channels, respectively; the time series image is composed of bacterial colony images at different times during the culture process; The spatial features of each frame are obtained by modeling the change trend between adjacent frames in the time series image and the image spatial structure distribution and growth characteristics Splicing the spatial features and growth characteristics Get spatiotemporal characteristics 3. A method for rapid identification of pathogens according to claim 2, characterized in that: Performing a multi-scale structural perception operation on the spatiotemporal features to obtain response features at multiple scales, and obtaining a colony significant region feature map based on the response features at multiple scales, including: Multi-scale dilated convolution operations are introduced in parallel to process the spatiotemporal features. By setting different dilation rates d i The hole convolution obtains the response features at multiple scales Among them, Concat(·) represents the concatenation operation; Indicates that the convolution kernel size is 3×3 and the void rate is d i The dilated convolution operation; i = 1, 2, ..., S; S is the set scale number; Response characteristics at multiple scales Perform pixel-by-pixel significance scoring to obtain the colony significance region feature map S t : Among them, * is the convolution operation, σ is the Sigmoid activation function, W s 、b s are learnable parameters.

4. A method for rapid identification of pathogens according to claim 3, characterized in that: Input the colony significant region feature map into the visual-language model to determine whether there is a growing colony and output the corresponding growth interpretation result, including: The colony salient region feature map is input into the vision-language model, and the colony growth probability p is output. grow ; Set the growth judgment threshold τ grow , if p grow >τ grow , it is judged that there are growing colonies and the growth judgment result is output as growth; otherwise, it is judged that there are no growing colonies and the growth judgment result is output as no growth.

5. A method for rapid identification of pathogens according to claim 3, characterized in that: If the growth interpretation result is growth, the bacterial species identification is performed, including: Perform a microscopic focusing classification operation on the significant region feature map and output the bacterial species identification result. The process is as follows: By introducing a local perception mechanism, the edge morphology, texture, fuzzy boundary and tiny diffusion structure in the salient region feature map are fine-grained modeled to output micro features F det , which is expressed as follows: Among them, MicroFocus(·) represents the micro-focus classification operation; ⊙ represents element-wise multiplication; The microscopic features F det Input to the target detection head to realize the positioning of different colonies, predict the probability distribution of each colony belonging to each type of bacterial species, and output the recognition results: B i =Regressor(Softmax(F det ))=[x i ,y i ,w i ,h i ] y type(i) =argmax(Softmax(W cls ·F det +b cls )) Among them, Regressor(·) is a bounding box regression function, which outputs the corresponding multiple colonies containing the center point (x i ,y i ) and width and height (w i ,h i )’s predicted bounding box parameters B i , W cls 、b cls is the trainable parameter of the fully connected layer, Softmax(·) is the normalized probability function, and the output of the bacterial species recognition result y type(i) The bacterial species category with the highest probability in each bounding box.

6. A method for rapid identification of pathogens according to claim 5, characterized in that: The loss function includes the vision-language large model loss function Bounding Box Regression Loss Function and the micro-focus classification loss function The vision-language large model loss function is the cross entropy loss function, which is used to constrain the consistency between the colony growth judgment results output by the vision-language large model and the true label; The bounding box regression loss function Used to constrain the regression accuracy of the detection box coordinates and the true bounding box; The micro-focused classification loss function The cross entropy loss function is used to constrain the matching relationship between the strain category prediction results output by the micro-focused classification operation and the actual strain label; The visual-language model loss function The bounding box regression loss function and the microscopic focused classification loss function The average value will be used as the final loss value to optimize the model parameters in the process of strain identification.

7. A rapid pathogen identification system for implementing a rapid pathogen identification method according to any one of claims 1 to 6, characterized in that: include: A spatiotemporal growth discrimination module obtains a bacterial colony time-series image sequence, performs a spatiotemporal growth discrimination operation on the bacterial colony time-series image sequence, and obtains spatiotemporal features by modeling the growth dynamics of the bacterial colony in the time dimension and the morphological distribution in the space dimension; A multi-scale structure perception module performs a multi-scale structure perception operation on the spatiotemporal features to obtain response features at multiple scales, and obtains a colony significant region feature map based on the response features at multiple scales; The growth interpretation module constructs a visual-language model, inputs the colony significant region feature map into the visual-language model, determines whether there are growing colonies, and outputs the corresponding growth interpretation result; if the growth interpretation result is growth, the bacterial species is identified; if the growth interpretation result is no growth, the subsequent analysis is terminated; Parameter training module, inputs training data and optimizes model parameters in the strain identification process based on the loss function; The model deployment module persists the trained model parameters and deploys them to the actual system to achieve rapid identification of pathogens.

8. A rapid pathogen identification system according to claim 7, characterized in that: The growth interpretation module includes a microscopic focus classification unit, which performs a microscopic focus classification operation on the significant area feature map and outputs a bacterial species identification result.

Images