Antarctic krill feeding habit identification and analysis method based on optical microscope

Abstract

The invention discloses an Antarctic krill feeding habit identification and analysis method based on an optical microscope, and relates to the technical field of marine biology, Antarctic krill samples in different time and space are collected in a target area, dissected and separated from digestive tract contents of the Antarctic krill samples, and microscopic observation samples are prepared; observing the microscopic observation sample by using an optical microscope, recording the morphological characteristics of the contents in the digestive tract, collecting a microscopic image of the contents in the digestive tract, and preliminarily classifying food particles; and performing molecular sequencing on the same batch of samples, extracting DNA sequence information of digestive tract contents, and obtaining molecular sequencing data. By combining the optical microscope image and the molecular sequencing data, the food particle type in the digestive tract of the euphausia superba can be more accurately identified, the morphological characteristics are provided by optical microscope observation, the species are confirmed from the molecular level by molecular sequencing, and the morphological characteristics and the molecular sequencing are complementary, so that the subjective judgment error caused by a single method is reduced; and the accuracy of feeding habit identification is improved.

Description

Technical Field

[0001] The present invention relates to the field of marine biology technology, and in particular to an optical microscope-based method for identifying and analyzing the feeding habits of Antarctic krill. Background Art

[0002] Antarctic krill is a strategic polar marine biological resource that plays a key role in the stability of the Southern Ocean ecosystem and the global material circulation and energy flow. Clarifying its feeding characteristics is a key link in clarifying its ecological function. In the context of global climate change, the feeding dynamics of Antarctic krill may change. Moreover, due to the limitations of polar survey sampling and the differences in the application of technical means, there is currently a lack of comprehensive and profound scientific understanding of Antarctic krill feeding habits, which limits the understanding of the ecological functions of Antarctic krill and the scientific management and sustainable utilization of Antarctic krill fisheries.

[0003] Optical microscopy, as a traditional means of observation, plays an important role in the identification of Antarctic krill diet. However, optical microscopy observation relies on the operator's experience and subjective judgment. When observing tiny organisms or species with similar morphology, it may be difficult to accurately distinguish them, resulting in inaccurate identification results. In addition, optical microscopy observation requires sample-by-sample analysis, which is difficult to meet the needs of large-scale research. Therefore, how to combine optical microscopy images and molecular sequencing data for comprehensive analysis, analyze the composition of krill digestive tract contents, clarify the regional and seasonal variation characteristics of Antarctic krill diet, and train a diet identification model through a large amount of labeled data to automatically identify and classify the types of food particles in the krill digestive tract is the problem to be solved by the present invention. To this end, a method for Antarctic krill diet identification and analysis based on optical microscopy is proposed. Summary of the Invention

[0004] The present invention aims to provide an optical microscope-based method for identifying and analyzing the feeding habits of Antarctic krill to solve the problems raised in the above-mentioned background technology.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: The method for identifying and analyzing the feeding habits of Antarctic krill based on an optical microscope includes the following steps: S1. Collect Antarctic krill samples at different times and spaces in the target area, dissect and separate their digestive tract contents, and prepare samples for microscopic observation through centrifugation and washing. S2. Observe the microscopic samples using an optical microscope, record the morphological characteristics of the digestive tract contents, collect microscopic images of the digestive tract contents, and preliminarily classify the food particles; S3. Perform molecular sequencing on the same batch of samples to extract DNA sequence information of the digestive tract contents and obtain molecular sequencing data; S4. Extract morphological features and species molecular features from microscopic images and molecular sequencing data, respectively. Then, fuse the optical microscope images with the molecular sequencing data to accurately annotate the food particles in the images and construct an annotated dataset. S5. Use the labeled dataset combined with the deep learning model to train the food habit identification model to automatically identify and classify food particle types; S6. Use the trained diet identification model to automatically identify and classify food particles in the krill digestive tract and analyze the regional and seasonal variations in Antarctic krill diet.

[0006] A further improvement of the technical solution of the present invention is that: S1 specifically includes: Based on the research objectives, a detailed sampling plan should be developed, including sampling time, location, and depth parameters. Key sea areas in the southwest Atlantic sector should be selected, and sampling points should be determined in different seasons and at different times of the day to ensure the spatial and temporal representativeness of the samples. Krill fishing vessels or scientific research vessels should be used to collect Antarctic krill samples using trawl nets according to the sampling plan. The collected krill samples should be immediately stored at low temperatures to prevent degradation of their internal substances. The collected Antarctic krill samples were transferred to a sterile operating table in the laboratory, and the Antarctic krill samples were preliminarily processed to remove impurities and excess tissues, and then the processed krill samples were fixed in formaldehyde solution; Dissect the fixed krill sample under a microscope, starting from the krill's mouth and following the digestive tract, gradually separating the digestive tract (stomach and intestines). During the dissection, use delicate dissecting tools and operate carefully to ensure that the digestive tract contents are not damaged or lost. Gently squeeze out the separated digestive tract contents and collect them in a 1.5ml centrifuge tube. The collected digestive tract contents are centrifuged, the supernatant is removed, the precipitated food particles are retained, and they are washed with buffer (physiological saline). The centrifugation and washing steps are repeated several times to remove impurities and excess liquid. The washed food particles are then made into glass slides. During the slide making process, the cleanliness and transparency of the glass slides are ensured. The food particles are evenly distributed on the glass slides, an appropriate amount of mounting medium is added, and a coverslip is applied to form a sample for microscopic observation.

[0007] A further improvement of the technical solution of the present invention is that: S2 specifically includes: Place the optical microscope on a stable, vibration-free laboratory table, plug in the power supply and turn on the light source. Select the required objective lens (10×, 40×, 100× oil objective lens, etc.) according to observation needs, and adjust the coarse and fine focusing knobs to ensure that the stage is at an appropriate height. Then, place the prepared Antarctic krill digestive tract contents microscope observation sample on the stage and secure it with a specimen clamp to ensure that the sample is located in the center of the light hole. Adjust the reflector or light source brightness knob to ensure the field of view is at an appropriate brightness. Start observation with a low-magnification objective lens (10×), slowly move the stage, and scan the sample comprehensively to gain a preliminary understanding of the distribution of the digestive tract contents. After finding the food particles of interest, switch to a high-magnification objective lens (40× or 100× oil immersion lens) for detailed observation. Observe the morphological characteristics of the food particles, including shape (round, oval, irregular, etc.), size (measure the major and minor diameters), surface structure (smooth, rough, textured, etc.), and color, and record the observed characteristic information; When food particles with typical characteristics are observed, use the image acquisition system (digital camera) equipped with the microscope to capture microscopic images. Adjust the image acquisition parameters including resolution, exposure time, and white balance to ensure that the captured images are clear and accurate in color. For each typical food particle, collect multiple images at different angles and focal planes to more comprehensively display its morphological characteristics. After acquisition, save the images to a designated folder and name them according to the sample number and food particle characteristics. Based on the recorded morphological characteristics of food particles and the collected microscopic images, and with reference to existing literature and classification standards, the food particles in the digestive tract contents were preliminarily classified. Food particles with similar morphology and characteristics were grouped into the same category. Food particles with unique morphology and difficult to classify were recorded and labeled separately. After the classification was completed, the classification results were sorted out, and the number and proportion of each type of food particles were counted.

[0008] A further improvement of the technical solution of the present invention is that: S3 specifically includes: From a preserved sample of Antarctic krill digestive tract contents, take an appropriate amount and place it in a sterile centrifuge tube. Add lysis buffer and mix thoroughly using a vortex oscillator to promote complete cell lysis and release DNA. Remove impurities and cell debris by centrifugation, collect the supernatant, and extract DNA from the supernatant using an organic solvent extraction method. The extracted DNA should be tested for concentration and purity using a UV spectrophotometer and stored in a -20°C refrigerator until used. According to the requirements of the selected molecular sequencing platform (Illumina), the extracted DNA is subjected to library construction. The constructed library is then quality-tested using an Agilent 2100 bioanalyzer to ensure that the fragment size distribution and concentration of the library meet the standards. Once qualified, the library is loaded onto the sequencer for high-throughput sequencing, obtaining a large amount of raw sequencing data, namely, DNA sequence information of the digestive tract contents. Data quality control is performed on the raw sequencing data generated by sequencing to remove low-quality sequences, adapter sequences, and contaminating sequences containing nitrogenous bases. SPAdes splicing software is used to splice the filtered high-quality sequences, assembling short sequences into longer continuous sequences (contigs) or overlapping groups (scaffolds). At the same time, the CD-HIT tool is used to perform redundancy processing on complex samples and remove duplicate sequences to reduce the data volume and computational complexity of subsequent analysis. Using the bioinformatics database (NCBI's nt database or Silva database), the spliced ​​and de-redundant sequences were compared with the known sequences in the database. By setting the comparison parameters and thresholds, sequences with high similarity were screened out, the types of microorganisms or morphologically similar species in the digestive tract contents were determined, the number of sequences of different species or their relative abundance were counted, and their proportion in the sample was analyzed. Combined with the sample collection information, the diet composition of Antarctic krill was comprehensively judged.

[0009] A further improvement of the technical solution of the present invention is that: S4 specifically includes: Organize the collected microscopic images of the digestive tract contents of Antarctic krill, extract the morphological characteristics of the food particles, describe the shape, surface texture and color characteristics of the particles, and record the extracted morphological characteristics in detail in a table; Extract the molecular characteristics of species from molecular sequencing data, compare the sequenced DNA sequences with known bioinformatics databases, extract the species classification information corresponding to each sequence, including the species genus, species classification level and molecular characteristics of species abundance, and organize the molecular characteristic information into a data table; The food particles in the microscopic images were matched with their corresponding sample sources based on their number and location. For each food particle, the morphological features extracted from the microscopic images were correlated and integrated with the species molecular features obtained from molecular sequencing. The two feature data were then merged into a unified data table using a programming language (Python). This ensured that each food particle had complete morphological and molecular feature information, and established a correspondence between the morphological features of the microscopic images and the species molecular features obtained from molecular sequencing. Based on the fused data, the image annotation tool (LabelImg) is used to annotate the food particles in the microscopic image. The position and boundary of each food particle are marked on the image, and detailed annotation information is added to it, including the species name and morphological feature description. The annotated image and the corresponding feature data are then saved to construct a complete annotated dataset.

[0010] A further improvement of the technical solution of the present invention is that: S5 specifically includes: Preprocess the constructed annotated dataset. This involves normalizing the microscopic images and resizing them uniformly to ensure that all images fed into the deep learning model are of the same size. The images are then grayscaled and divided into training, validation, and test sets. Based on the characteristics of the food habit identification task, a deep learning model architecture based on convolutional neural networks was selected to construct a food habit identification model. The model was built using the PyTorch deep learning framework, which includes multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layers are used to extract local features of the image, the pooling layers are used to reduce the feature dimension, and the fully connected layers are used to classify the features. During the construction process, the model's hyperparameters of the number of layers and neurons were set according to the characteristics of the annotated dataset and computing resources. The training set was then input into the constructed convolutional neural network model for training, and the model performance was regularly evaluated using the validation set data. The trained model is finally evaluated using the test set. The accuracy, recall, and F1 value indicators of the model on the test set are calculated to comprehensively evaluate the model's performance. If the model performance meets the requirements, it is deployed in the actual application environment. When a new microscopic image of the digestive tract contents of Antarctic krill is input, the food particle types are automatically identified and classified, and the species name corresponding to each particle is output.

[0011] A further improvement of the technical solution of the present invention is that the specific process of automatically identifying and classifying the types of food particles is as follows: Receive new microscopic images of Antarctic krill digestive tract contents and perform preprocessing operations. The microscopic images are adjusted to the same size as when the model was trained to ensure that the image size meets the model input requirements. The images are grayscaled to reduce data dimensions and improve processing efficiency. The images are then normalized to uniformly adjust pixel values ​​to between 0 and 1 to reduce the impact of factors such as lighting on the images. The preprocessed image is input into the trained food habit identification model. The convolutional layer extracts local features of the image, the pooling layer reduces the feature dimension, and the fully connected layer classifies the features. The food habit identification model identifies and classifies each food particle in the image based on the learned features and parameters, outputs the probability distribution of the species name corresponding to each particle, and selects the species name with the highest probability as the classification result for the particle. Based on the inference results of the food habit identification model, the species name corresponding to each food particle is output. At the same time, the image annotation tool (LabelImg) is used to mark the position and boundary of each food particle on the original microscopic image and add the corresponding species name annotation information.

[0012] A further improvement of the technical solution of the present invention is that the probability distribution expression of the species name corresponding to each particle is: ; Where, is the probability that the i-th food particle belongs to the k-th species, is the original output of the model for the i-th food particle belonging to the k-th species, is the exponential sum of the raw outputs of all species for the ith food particle, used to normalize the probability distribution, The value range of is (0,1), which represents the probability value, and n is the number of species names; Select the species name with the highest probability as the classification result of the particle, that is: ; Where, is the species name of the i-th food particle predicted by the model, Indicates taking The maximum k value.

[0013] A further improvement of the technical solution of the present invention is that: S6 specifically includes: Antarctic krill samples were collected in different regions and seasons in Antarctica. Samples of their digestive tract contents were obtained, dissected, and microscopic images of the digestive tract contents were prepared. The collected microscopic images were preprocessed, including image resizing, grayscale conversion, and normalization, to ensure that the images were consistent with the input used for model training. The preprocessed images were used as input to the model. Combined with the trained food habit identification model, each food particle in the microscopic image is identified and classified, and the probability distribution of the species name corresponding to each particle is output. The one with the highest probability is selected as the classification result to obtain preliminary food particle classification data; Based on the food particle classification results derived from the feeding habit identification model, we conducted a statistical analysis of krill feeding habits data from different regions and seasons, calculated the frequency and proportion of each type of food particle in each region and season, and analyzed the dominant food types; Compare the differences in feeding habits between different regions, analyze the impact of regional environmental factors on krill feeding habits, and at the same time, analyze the changing patterns of feeding habits between different seasons in the same region, study the relationship between seasonal changes and krill feeding habits adjustments, and based on the data analysis results, explain the regional and seasonal changes in Antarctic krill feeding habits, generate an identification and analysis report, including classification results and statistical analysis, to intuitively display the characteristics of feeding habits changes.

[0014] Due to the adoption of the above technical solution, the present invention has the following technical advancements compared to the prior art: The present invention provides an optical microscope-based method for identifying and analyzing the diet of Antarctic krill. By combining optical microscope images and molecular sequencing data, the types of food particles in the digestive tract of Antarctic krill can be identified more accurately. Optical microscope observation provides morphological characteristics, while molecular sequencing confirms species at the molecular level. The two complement each other, reducing the subjective judgment errors caused by a single method and improving the accuracy of diet identification.

[0015] The present invention provides an optical microscope-based method for identifying and analyzing the diet of Antarctic krill. Combined with a deep learning model, it can automatically identify and classify food particles, greatly improving analysis efficiency and meeting the needs of large-scale research. It makes it possible to conduct a comprehensive study of the diet of Antarctic krill, analyze the regional and seasonal variation characteristics of the diet of Antarctic krill, and reveal its diet dynamics. By collecting samples in different regions and seasons, we can understand the food selection preferences of Antarctic krill under different environmental conditions, providing an important basis for understanding its ecological functions and adaptation mechanisms. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0017] Figure 1 It is a schematic diagram of the workflow of the present invention; Figure 2 Schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0019] Example 1, as Figure 1 、 Figure 2 As shown, the present invention provides an analysis method for identifying the feeding habits of Antarctic krill based on an optical microscope, comprising the following steps: S1. Collect Antarctic krill samples at different times and spaces in the target area (Southwest Atlantic sector). Dissect and separate the digestive tract contents. Prepare samples for microscopic observation through centrifugation and washing. Develop a detailed sampling plan based on the research objectives, including sampling time, location, and depth parameters. Select key sea areas in the Southwest Atlantic sector and identify sampling points in different seasons and during different time periods to ensure the temporal and spatial representativeness of the samples. Use krill fishing vessels or scientific research vessels to collect Antarctic krill samples using trawl nets according to the sampling plan. Collected krill samples must be immediately stored at low temperatures to prevent degradation of their internal substances. During each sampling voyage, sampling points were set up during the day (6:00-18:00) and at night (18:00-6:00 the following day) to study the influence of circadian rhythms on the distribution and feeding of Antarctic krill. Key sea areas in the southwest Atlantic sector, including the waters near the South Orkney Islands, the southern waters of the South Shetland Islands, and the Bransfield Strait, were selected as sampling areas. Within each key sea area, 5-8 sampling points were evenly distributed based on environmental parameters such as ocean currents, water temperature, and salinity using Geographic Information System (GIS) technology, combined with historical data and real-time monitoring information, to ensure spatial representativeness of the samples. Based on previous research and relevant literature, the sampling depth ranges were determined to be surface (0-50 meters), mid-layer (50-200 meters), and bottom layer (200 meters to the seabed). At each sampling point, sampling is carried out according to different depth layers, and at least three samples are collected in each depth layer to obtain the distribution and ecological information of Antarctic krill in different water layers. Antarctic krill samples are collected using mid-water trawl or bottom trawl. The width of the trawl mouth is not less than 10 meters. Each trawl operation time is controlled to be about 30 minutes, and the trawl speed is maintained at 2-3 knots. The collected Antarctic krill samples must be immediately placed in an insulated box filled with ice and stored at a low temperature of 0-4°C to prevent the degradation of their body substances. The collected Antarctic krill samples are transferred to the laboratory sterile operating table for preliminary processing to remove impurities and excess tissues. The processed krill samples are then fixed in formaldehyde solution to prevent the degradation of biological tissues and food particles in the samples. After fixation, the samples were properly stored at low temperatures to prevent deterioration. The krill samples were gently rinsed with deionized water to remove impurities and seawater attached to the surface. Dissecting scissors and tweezers were used to remove excess tissue such as the shell and appendages of the krill samples, leaving only the intact body parts. The preliminarily processed Antarctic krill samples were placed in a specimen bottle containing a 10% formaldehyde solution, ensuring that the samples were completely immersed in the fixative. The formaldehyde solution can effectively prevent the degradation of biological tissues and food particles in the samples, preserving their original morphology and structure. Label the specimen bottle with the sampling time, location, depth, sample number, and other information in detail. Store the fixed sample in a 4°C refrigerator to prevent sample deterioration and microbial growth. Dissect the fixed krill sample under a microscope, starting from the krill's mouth and following the digestive tract to gradually separate the digestive tract (stomach and intestines). During the dissection, use delicate dissecting tools and operate carefully to ensure that the digestive tract contents are not damaged or lost. Gently squeeze out the separated digestive tract contents and collect them into a 1.5ml centrifuge tube. During the extraction process, try to avoid mixing with other tissues or impurities to ensure the purity of the contents. Centrifuge the collected digestive tract contents, remove the supernatant, and retain the precipitated food particles. Wash with buffer (physiological saline). Repeat the centrifugation and washing steps several times to remove impurities and excess liquid. Then, prepare the washed food particles into a slide. During the slide preparation process, ensure the cleanliness and transparency of the slide. Distribute the food particles evenly on the slide, add an appropriate amount of mounting medium, and cover with a coverslip to prepare the sample for microscopic observation. Among them, the centrifuge tube containing the contents of the digestive tract is placed in a centrifuge and centrifuged at 3000-4000 rpm for 5-10 minutes to allow the food particles to settle to the bottom of the centrifuge tube, remove the supernatant, add an appropriate amount of physiological saline as a buffer to the centrifuge tube, gently shake the centrifuge tube to resuspend the food particles, centrifuge again, repeat the washing steps 3-5 times to remove impurities and excess liquid until the supernatant is clear and transparent. After the last centrifugation, carefully pour out the supernatant and retain the precipitated food particles. Use a micropipette to draw a small amount of food particle suspension, drop it on a clean glass slide, and use a coverslip to gently cover the suspension to avoid bubbles. If the food particles are unevenly distributed, use a dissecting needle to gently adjust their position so that they are evenly distributed on the glass slide. Add an appropriate amount of neutral gum as a sealing agent to the edge of the coverslip. After sealing, place the glass slide in a ventilated place to dry to form a sample for microscopic observation; S2. Observe the sample using an optical microscope, record the morphological characteristics of the digestive tract contents, collect microscopic images of the digestive tract contents, and preliminarily classify the food particles. Place the optical microscope on a stable, vibration-free laboratory table, plug in the power cord and turn on the light source. Select the required objective lens (10×, 40×, 100× oil objective lens, etc.) according to your observation needs, and adjust the coarse and fine focusing knobs to ensure that the stage is at an appropriate height. Then, place the prepared Antarctic krill digestive tract contents microscope observation sample on the stage and secure it with a specimen clamp, ensuring that the sample is located in the center of the light aperture. Adjust the reflector or light source brightness knob to ensure appropriate field of view brightness. Start observation with a low-magnification objective lens (10×), slowly move the stage, and fully scan the sample to gain a preliminary understanding of the distribution of digestive tract contents. After finding the food particles of interest, switch to a high-magnification objective lens (40× or 100× oil immersion lens) for detailed observation. Observe the morphological characteristics of the food particles, including shape (round, oval, irregular, etc.), size (measure the major and minor diameters), surface structure (smooth, rough, with texture, etc.), and color, and record the observed characteristic information. When food particles with typical characteristics are observed, use the image acquisition system (digital camera) equipped with the microscope to capture microscopic images. Adjust the image acquisition parameters including resolution, exposure time, and white balance to ensure that the captured images are clear and accurate in color. For each typical food particle, collect multiple images at different angles and focal planes to more comprehensively display its morphological characteristics. After the collection is completed, the images are saved to a designated folder and named according to the sample number and food particle characteristics. Based on the recorded morphological characteristics of the food particles and the collected microscopic images, the food particles in the digestive tract contents are preliminarily classified with reference to existing literature and classification standards. Food particles with similar morphology and similar characteristics are grouped into the same category. Food particles with unique morphology and difficult to classify are recorded and labeled separately. After the classification is completed, the classification results are sorted and the number and proportion of each type of food particles are counted; S3. Perform molecular sequencing on the same batch of samples, extract DNA sequence information from the digestive tract contents, and obtain molecular sequencing data for identifying the species of microorganisms or morphologically similar species. Through bioinformatics analysis, identify the species source of food particles and provide molecular evidence for food habit identification. Take an appropriate amount of the preserved Antarctic krill digestive tract content sample and place it in a sterile centrifuge tube. Add lysis buffer and mix thoroughly using a vortex oscillator to promote cell lysis and release DNA. Remove impurities and cell debris by centrifugation, collect the supernatant, and extract DNA from the supernatant using an organic solvent extraction method. The extracted DNA needs to be tested for concentration and purity using a UV spectrophotometer to ensure OD260 / OD28 0 ratio is between 1.8-2.0, which meets the requirements of subsequent Illumina sequencing and is stored in a -20°C refrigerator for future use. The extracted DNA is used for library construction according to the requirements of the selected molecular sequencing platform (Illumina). For the short-read sequencing platform, the DNA is fragmented by ultrasonic fragmentation, and specific sequencing adapters are connected to both ends of the fragments using T4 DNA ligase. At the same time, PCR amplification technology is used to amplify the DNA fragments containing adapters as templates to increase the amount of DNA. The constructed library is then quality tested using an Agilent 2100 bioanalyzer to ensure that the fragment size distribution and concentration of the library meet the standards; After passing the test, the library is loaded onto a sequencer for high-throughput sequencing to obtain a large amount of raw sequencing data, namely the DNA sequence information of the digestive tract contents. The raw sequencing data generated by sequencing is subjected to data quality control to remove low-quality sequences, adapter sequences, and contaminating sequences containing nitrogenous bases. The filtered high-quality sequences are then spliced ​​using SPAdes splicing software to assemble short sequences into longer continuous sequences (contigs) or overlapping groups (scaffolds). At the same time, the CD-HIT tool is used to deduplicate complex samples and remove duplicate sequences to reduce the data volume and computational complexity of subsequent analysis. The spliced ​​and de-duplicated sequences are compared with known sequences in the bioinformatics database (NCBI's nt database or Silva database) using bioinformatics databases. By setting the alignment parameters and thresholds, sequences with high similarity are screened out to determine the types of microorganisms or morphologically similar species in the digestive tract contents. The number or relative abundance of sequences of different species is counted, and their proportion in the sample is analyzed. Combined with the sample collection information, the diet composition of Antarctic krill is comprehensively judged. S4. Extract morphological characteristics and species molecular characteristics from microscopic images and molecular sequencing data, respectively. Then, fuse the optical microscope images with the molecular sequencing data, accurately annotate the food particles in the images, construct an annotated dataset, organize the collected microscopic images of the digestive tract contents of Antarctic krill, and extract the morphological characteristics of the food particles. Use image processing software (ImageJ or OpenCV) to open the images, observe the morphology of different food particles in the images, and measure the major diameter, minor diameter, perimeter, and area parameters of the particles to quantify their size characteristics. At the same time, the shape, surface texture and color characteristics of the particles are described, and the extracted morphological characteristic information is recorded in detail in a table. The molecular characteristics of the species are extracted from the molecular sequencing data. The sequenced DNA sequences are compared with known bioinformatics databases. The species classification information corresponding to each sequence is extracted, including the genus and species classification level of the species and the molecular characteristics of the species abundance. The molecular characteristic information is organized into a data table. According to the number and position information of the food particles in the microscopic image and the sample source corresponding to each sequence in the molecular sequencing, the two are matched. For each food particle, the morphological characteristics extracted from its microscopic image are associated and integrated with the species molecular characteristics obtained by molecular sequencing. The two characteristic data are merged into a unified data table using the programming language (Python) to ensure that each food particle has complete morphological and molecular characteristic information, and to establish a correspondence between the morphological characteristics of the microscopic image and the species molecular characteristics obtained by molecular sequencing. Based on the fused data, the image annotation tool (LabelImg) is used to annotate the food particles in the microscopic image. The location and boundaries of each food particle are marked on the image, and detailed annotation information is added to it, including the species name and morphological feature description. The annotated image and the corresponding feature data are then saved to construct a complete annotated dataset. S5. Use the labeled dataset combined with the deep learning model to train the food habit identification model to automatically identify and classify food particle types; S6. Use the trained diet identification model to automatically identify and classify food particles in the krill digestive tract and analyze the regional and seasonal variations in Antarctic krill diet.

[0020] Example 2, as Figure 1 、 Figure 2 As shown, based on Example 1, the present invention provides a technical solution: preferably, S5 specifically includes: The constructed annotated dataset was preprocessed. Microscopic images were normalized and resized uniformly to ensure that all images input to the deep learning model were of the same size. The images were also grayscaled and divided into training, validation, and test sets in a ratio of 7:1.5:1.5. The training set was used for model feature learning and parameter adjustment, the validation set was used to evaluate model performance and prevent overfitting during training, and the test set was used to ultimately evaluate the generalization ability of the model. Based on the characteristics of the food habit identification task, a deep learning model architecture based on convolutional neural networks was selected to construct a food habit identification model. The model was built using the PyTorch deep learning framework and contained multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layers were used to extract local features of the image, the pooling layers were used to reduce feature dimensions, and the fully connected layers were used to classify features. During the construction process, the model's number of layers and neuron hyperparameters are set according to the characteristics of the labeled dataset and computing resources, and then the training set is input into the built convolutional neural network model for training. During the training process, the backpropagation algorithm and the stochastic gradient descent optimization algorithm are used to adjust the model's weight parameters to minimize the loss function, so that the error between the model's prediction results and the true label gradually decreases. At the same time, the validation set data is used to regularly evaluate the model performance, and the accuracy and loss function curves are plotted to observe whether the model is overfitting or underfitting. If overfitting occurs, regularization technology is used for optimization to improve the model's generalization ability. The trained model is finally evaluated using the test set, and the accuracy, recall rate, and F1 value indicators of the model on the test set are calculated to comprehensively evaluate the model's performance. If the model performance meets the requirements, it will be deployed in the actual application environment. When a new microscopic image of the digestive tract contents of Antarctic krill is input, the food particle types are automatically identified and classified, and the species name corresponding to each particle is output; In addition, the specific process of automatically identifying and classifying food particle types is as follows: Receive new microscopic images of Antarctic krill digestive tract contents and perform preprocessing operations. The microscopic images are adjusted to the same size as when the model was trained to ensure that the image size meets the model input requirements. The images are then grayscaled to reduce data dimensions and improve processing efficiency. The images are then normalized and pixel values ​​are uniformly adjusted to between 0 and 1 to reduce the impact of factors such as light on the images. The preprocessed images are input into the trained food habit identification model. The local features of the image are extracted through the convolution layer, the feature dimension is reduced through the pooling layer, and the features are classified by the fully connected layer. The food habit identification model identifies and classifies each food particle in the image based on the learned features and parameters, and outputs the probability distribution of the species name corresponding to each particle. The species name with the highest probability is selected as the classification result for the particle. Based on the inference results of the food habit identification model, the species name corresponding to each food particle is output. At the same time, the image annotation tool (LabelImg) is used to annotate the position and boundaries of each food particle on the original microscopic image and add the corresponding species name annotation information. The probability distribution expression of the species name corresponding to each particle is: ; Where, is the probability that the i-th food particle belongs to the k-th species, is the original output of the model for the i-th food particle belonging to the k-th species, is the exponential sum of the raw outputs of all species for the ith food particle, used to normalize the probability distribution, The value range of is (0,1), which represents the probability value, and n is the number of species names; Select the species name with the highest probability as the classification result of the particle, that is: ; Where, is the species name of the i-th food particle predicted by the model, Indicates taking The maximum k value; S6 specifically includes: Antarctic krill samples were collected in different regions and seasons in Antarctica, and samples of their contents were obtained from their digestive tracts. These were dissected and microscopic images of the digestive tract contents were prepared. The collected microscopic images were preprocessed, including adjusting the image size, grayscale conversion, and normalization to ensure that the images were consistent with the input when the model was trained. The preprocessed images were used as input to the model. Combined with the trained food habit identification model, each food particle in the microscopic image was identified and classified, and the probability distribution of the species name corresponding to each particle was output. The one with the highest probability was selected as the classification result, obtaining preliminary food particle classification data. The food particle classification results were derived based on the food habit identification model. Statistical analysis of krill feeding habits data from different regions and seasons was conducted. The frequency and proportion of each type of food particle in each region and season were calculated. The dominant food species were analyzed, dietary differences between different regions were compared, and the impact of regional environmental factors on krill feeding habits were analyzed. Furthermore, the changing patterns of feeding habits between different seasons in the same region were analyzed, and the association between seasonal changes and krill feeding adjustments was studied. Based on the data analysis results, the regional and seasonal variations in Antarctic krill feeding habits were explained. An identification and analysis report was generated, including classification results and statistical analysis, to visually demonstrate the characteristics of feeding habits changes. The feeding habits classification data of Antarctic krill samples collected from different regions were summarized, and the frequency and number of each type of food particle in each region were recorded. The feeding habits data for each region were grouped according to the number of study areas. The frequency (proportion of the number of samples with each type of food particle to the total number of samples) and proportion (the proportion of the number of food particles of this type to the total number of food particles) of each type of food particle in each region were calculated. Visualization tools including bar charts and pie charts were used to display the distribution of each type of food in different regions and compare the feeding habits differences between different regions. Through statistical analysis methods, we compared the significant differences in the frequency and proportion of various food particles in different regions, analyzed the impact of regional environmental factors on krill diet, conducted time series analysis on data from different seasons in the same region, observed the seasonal trends in the frequency and proportion of various food particles, determined the dominant food types in each season, analyzed the seasonal changes in dominant foods, and studied the relationship between seasonal changes and krill diet adjustments through correlation analysis. Based on the comparative analysis results, we then comprehensively explained the regional and seasonal variation characteristics of Antarctic krill diet.

[0021] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. An optical microscope-based method for identifying and analyzing the feeding habits of Antarctic krill, characterized in that: The following steps are involved: S1. Collect Antarctic krill samples at different times and spaces in the target area, dissect them, separate their digestive tract contents, and prepare samples for microscopic observation; S2. Observe the microscopic samples using an optical microscope, record the morphological characteristics of the digestive tract contents, collect microscopic images of the digestive tract contents, and preliminarily classify the food particles; S3. Perform molecular sequencing on the same batch of samples to extract DNA sequence information of the digestive tract contents and obtain molecular sequencing data; S4. Extract morphological features and species molecular features from microscopic images and molecular sequencing data respectively, then fuse the optical microscope images with the molecular sequencing data, annotate the food particles in the images, and construct an annotated dataset. S5. Use the labeled dataset combined with the deep learning model to train the food habit identification model to automatically identify and classify food particle types; S6. Use the trained diet identification model to automatically identify and classify food particles in the krill digestive tract and analyze the regional and seasonal variations in Antarctic krill diet.

2. The method for identifying and analyzing the feeding habits of Antarctic krill based on an optical microscope according to claim 1, characterized in that: Said S1 specifically includes: Develop a detailed sampling plan based on the research objectives, including sampling time, location, and depth parameters. Use krill fishing vessels or scientific research vessels to collect Antarctic krill samples using trawl nets according to the sampling plan, and immediately cryopreserve the collected krill samples. The collected Antarctic krill samples were transferred to a sterile operating table in the laboratory, and the Antarctic krill samples were preliminarily processed to remove impurities and excess tissues, and then the processed krill samples were fixed in formaldehyde solution; The fixed krill samples were dissected under a microscope, their digestive tracts were separated, and the contents of the digestive tracts were squeezed out and collected into 1.5 ml centrifuge tubes; The collected digestive tract contents are centrifuged, the supernatant is removed, the precipitated food particles are retained, and washed with buffer. The centrifugation and washing steps are repeated several times to remove impurities and excess liquid. The washed food particles are then made into glass slides, an appropriate amount of mounting medium is added, and a coverslip is applied to form a sample for microscopic observation.

3. The method for identifying and analyzing the feeding habits of Antarctic krill based on an optical microscope according to claim 1, characterized in that: The S2 specifically includes: Place the optical microscope on a stable, vibration-free laboratory table, connect the power supply and turn on the light source. Select the required objective lens according to observation needs, and adjust the coarse and fine focusing knobs to make the stage at an appropriate height. Then place the prepared Antarctic krill digestive tract contents microscope observation sample on the stage and secure it with a specimen clamp. Start with a low-magnification objective lens, slowly move the stage, and scan the sample thoroughly to gain a preliminary understanding of the distribution of the digestive tract contents. Once you find the food particles of interest, switch to a high-magnification objective lens for detailed observation. Observe the morphological characteristics of the food particles, including shape, size, surface structure, and color, and record the observed characteristic information. When food particles with typical characteristics are observed, use the image acquisition system equipped with the microscope to collect microscopic images, adjust the image acquisition parameters including resolution, exposure time, and white balance, and collect multiple images of each typical food particle at different angles and focal planes. After the acquisition is completed, save the images to a designated folder and name them according to the sample number and food particle characteristics; Based on the recorded morphological characteristics of food particles and the collected microscopic images, and with reference to existing literature and classification standards, the food particles in the digestive tract contents were preliminarily classified. Food particles with similar morphology and characteristics were grouped into the same category. Food particles with unique morphology and difficult to classify were recorded and labeled separately. After the classification was completed, the classification results were sorted out, and the number and proportion of each type of food particles were counted.

4. The method for identifying and analyzing the feeding habits of Antarctic krill based on an optical microscope according to claim 1, characterized in that: The S3 specifically includes: From a preserved sample of Antarctic krill digestive tract contents, take an appropriate amount and place it in a sterile centrifuge tube. Add lysis buffer and mix thoroughly using a vortex oscillator to release the DNA. Remove impurities and cell debris by centrifugation, collect the supernatant, and extract DNA from the supernatant using an organic solvent extraction method. The extracted DNA needs to be tested for concentration and purity using a UV spectrophotometer and stored in a -20°C refrigerator until used. According to the requirements of the selected molecular sequencing platform, the extracted DNA is subjected to library construction, and the constructed library is then quality tested. Once qualified, the library is loaded onto a sequencer for high-throughput sequencing to obtain raw sequencing data, i.e., DNA sequence information of the digestive tract contents. Data quality control was performed on the raw sequencing data generated by sequencing to remove low-quality sequences, adapter sequences, and contaminating sequences containing nitrogenous bases. SPAdes splicing software was used to splice the filtered high-quality sequences. At the same time, the CD-HIT tool was used to perform redundancy processing on complex samples and remove duplicate sequences. Using the bioinformatics database, the spliced ​​and de-redundant sequences were compared with the known sequences in the database. By setting the comparison parameters and thresholds, sequences with high similarity were screened out, the number of sequences or relative abundance of different species were counted, and their proportion in the sample was analyzed. The diet composition of Antarctic krill was comprehensively judged in combination with the sample collection information.

5. The method for identifying and analyzing the feeding habits of Antarctic krill based on an optical microscope according to claim 1, characterized in that: The S4 specifically includes: Organize the collected microscopic images of the digestive tract contents of Antarctic krill, extract the morphological characteristics of the food particles, describe the shape, surface texture and color characteristics of the particles, and record the extracted morphological characteristics in detail in a table; Extract the molecular characteristics of species from molecular sequencing data, compare the sequenced DNA sequences with known bioinformatics databases, extract the species classification information corresponding to each sequence, including the species genus, species classification level and molecular characteristics of species abundance, and organize the molecular characteristic information into a data table; Based on the number and location information of the food particles in the microscopic image and the sample source corresponding to each sequence in the molecular sequencing, the two are matched. For each food particle, the morphological characteristics extracted from the microscopic image are correlated and integrated with the species molecular characteristics obtained by molecular sequencing. The two feature data are merged into a unified data table to establish the correspondence between the morphological characteristics of the microscopic image and the species molecular characteristics obtained by molecular sequencing; Based on the fused data, image annotation tools are used to annotate the food particles in the microscopic image, marking the position and boundaries of each food particle on the image, and adding detailed annotation information for it, including species name and morphological feature description. The annotated image and the corresponding feature data are then saved to construct a complete annotated dataset.

6. The method for identifying and analyzing the feeding habits of Antarctic krill based on an optical microscope according to claim 1, characterized in that: The S5 specifically includes: The constructed annotated dataset is preprocessed, wherein the microscopic images are normalized, the image size is uniformly adjusted, the images are grayscaled, and the annotated dataset is divided into a training set, a validation set, and a test set; Based on the characteristics of the food habit identification task, we selected a deep learning model architecture based on a convolutional neural network to construct a food habit identification model. We used the PyTorch deep learning framework to build the model, which includes multiple convolutional layers, pooling layers, and fully connected layers. We set the model's hyperparameters, such as the number of layers and neurons, based on the characteristics of the annotated dataset and computing resources. We then input the training set into the constructed convolutional neural network model for training, and regularly evaluated the model's performance using the validation set data. The trained model is finally evaluated using the test set to comprehensively assess its performance. If the model performance meets the requirements, it is deployed in a real-world application environment. When a new microscopic image of the digestive tract contents of Antarctic krill is input, the model automatically identifies and classifies the types of food particles and outputs the species name corresponding to each particle.

7. The method for identifying and analyzing the feeding habits of Antarctic krill based on an optical microscope according to claim 6, characterized in that: The specific process of automatically identifying and classifying food particle types is as follows: Receive new microscopic images of the digestive tract contents of Antarctic krill, perform preprocessing operations, adjust the microscopic images to the same size as used when training the model, grayscale the images, and then normalize the images; The preprocessed image is input into the trained food habit identification model to identify and classify each food particle in the image, output the probability distribution of the species name corresponding to each particle, and select the species name with the highest probability as the classification result of the particle; Based on the inference results of the food habit identification model, the species name corresponding to each food particle is output. At the same time, the image annotation tool is used to mark the position and boundary of each food particle on the original microscopic image, and the corresponding species name annotation information is added.

8. The method for identifying and analyzing the feeding habits of Antarctic krill based on an optical microscope according to claim 7, characterized in that: The probability distribution expression of the species name corresponding to each particle is: ； Where, is the probability that the i-th food particle belongs to the k-th species, is the original output of the model for the i-th food particle belonging to the k-th species, is the exponential sum of the raw outputs of all species for the ith food particle, The value range of is (0,1), which represents the probability value, and n is the number of species names; Select the species name with the highest probability as the classification result of the particle, that is: ； Where, is the species name of the i-th food particle predicted by the model, Indicates taking The maximum k value.

9. The method for identifying and analyzing the feeding habits of Antarctic krill based on an optical microscope according to claim 1, characterized in that: The S6 specifically includes: Antarctic krill samples were collected in different regions and seasons in Antarctica, and samples of their digestive tract contents were obtained. These samples were dissected and microscopic images of the digestive tract contents were prepared. The collected microscopic images were preprocessed and used as input for the model. Combined with the trained food habit identification model, each food particle in the microscopic image is identified and classified, and the probability distribution of the species name corresponding to each particle is output. The one with the highest probability is selected as the classification result to obtain preliminary food particle classification data; Based on the food particle classification results derived from the feeding habit identification model, we conducted a statistical analysis of krill feeding habits data from different regions and seasons, calculated the frequency and proportion of each type of food particle in each region and season, and analyzed the dominant food types; Compare the differences in feeding habits between different regions, analyze the impact of regional environmental factors on krill feeding habits, and at the same time, analyze the changing patterns of feeding habits between different seasons in the same region, study the relationship between seasonal changes and krill feeding habits adjustments, and based on the data analysis results, explain the regional and seasonal changes in Antarctic krill feeding habits, generate an identification and analysis report, including classification results and statistical analysis, to intuitively display the characteristics of feeding habits changes.

Images