An invasive Saccharomyces cerevisiae detection method, device, and storage medium

Abstract

The present invention discloses a detection method, device and storage medium for invasive Saccharomyces cerevisiae, including using an image collector to collect multiple yeast images, using annotation software to annotate the positions and sizes of yeast cells in the multiple yeast images to obtain a training data set, using a feature extraction network model GhostNet to extract the feature information of the yeast images in the training data set to obtain a feature image training set, using an FPN feature pyramid network to enhance feature extraction of the feature image training set to obtain an enhanced feature image training set, and using the enhanced feature image training set to train a YOLOv3 model to obtain a trained yeast detection model. It realizes the reduction of the total number of feature extraction network parameters and the computational complexity, improves the recognition speed of the algorithm while ensuring the recognition accuracy; is conducive to extracting better image feature values, effectively reduces the problems of missed detection and false detection existing in the prior art, and improves the detection accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of cell detection, and particularly to a method and device for detecting invasive Saccharomyces cerevisiae and a computer-readable storage medium. Background Art

[0002] Saccharomyces cerevisiae is generally considered a safe non-pathogenic microorganism, which is often colonized in the skin, vaginal mucosa, digestive tract and respiratory tract. However, in the past two decades, the cases of invasive Saccharomyces cerevisiae infection diseases have been increasing continuously, especially in immunocompromised patient groups such as the elderly and critically ill patients. For example, taking yeast preparations can cause unexplained fever, pneumonia, fungemia, liver abscess, peritonitis, vaginitis, urinary tract infection or septic shock, etc. Therefore, Saccharomyces cerevisiae is also considered an emerging opportunistic pathogen, associated with various infections, such as antibiotic-associated diarrhea, adult acute diarrhea, HIV-associated diarrhea, urogenital tract infection, esophagitis, pneumonia, liver abscess or peritonitis infection. In addition, risk factors such as central venous catheter, antibiotic use, ICU admission and immunosuppression also promote the invasive infection of Saccharomyces cerevisiae.

[0003] Clinically, for the preliminary diagnosis and treatment of invasive Saccharomyces cerevisiae infection, it is necessary to detect and purify Saccharomyces cerevisiae in the patient's blood first. Because the prerequisite for Saccharomyces cerevisiae to cause fungal infection is that the ingested Saccharomyces cerevisiae penetrates the immune barrier in immunocompromised patients, enters the blood circulation system, spreads and invades the infection site. In recent years, with the development of the field of artificial intelligence, in medical imaging, object detection has gradually been used to detect yeast cells, tissues or organs in images. Since the feature extraction network of the existing YOLOv3 detection algorithm uses the DarkNet53 network, the accuracy of the extracted Saccharomyces cerevisiae image features is not high, the precision is poor, and there are large errors in complex medical image processing. Therefore, it is necessary to design a method for detecting invasive Saccharomyces cerevisiae to solve the problems of missed detection, false detection and low detection accuracy in the prior art. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and device for detecting invasive Saccharomyces cerevisiae and a storage medium, so as to solve the problems of low accuracy, poor precision, and missed detection and false detection in the existing Saccharomyces cerevisiae cell detection algorithm.

[0005] To solve the above technical problems, the present invention provides a method and device for detecting invasive Saccharomyces cerevisiae and a storage medium, including:

[0006] Construct an invasive Saccharomyces cerevisiae detection model based on the YOLOv3 network, and the yeast detection model includes: a feature extraction network GhostNet, an FPN feature pyramid network, and a prediction network;

[0007] Collect multiple yeast images using an image collector;

[0008] Use annotation software to annotate the positions and sizes of yeast cells in the multiple yeast images to obtain an annotated image dataset;

[0009] Use the feature extraction network model GhostNet to extract the feature information of the yeast images in the annotated image dataset to obtain a feature image training set;

[0010] Use the FPN feature pyramid network to enhance feature extraction for the feature image training set to obtain an enhanced feature image training set;

[0011] Use the prediction network to predict the enhanced feature image to obtain a yeast prediction image;

[0012] Train the yeast detection model until the loss function between the yeast prediction image and the annotated image converges to obtain a trained yeast detection model;

[0013] Use the yeast detection model to detect the yeast image to be detected and output the positions and sizes of the yeast in the image to be detected.

[0014] Preferably, the first convolutional layer of the feature extraction network model GhostNet is a dilated convolution.

[0015] Preferably, the FPN feature pyramid network includes a conventional convolution and a depthwise separable convolution.

[0016] Preferably, the depthwise separable convolution consists of two parts: a depthwise convolution and a pointwise convolution.

[0017] Preferably, the FPN feature pyramid network uses the HardSwish activation function, and its formula is as follows

[0018]

[0019] where x is the neuron variable of the convolutional neural network.

[0020] Preferably, the calculation formula of the loss function of the yeast detection model is:

[0021] Loss yolov3 =∑(Coord loss +Conf loss +Class loss )

[0022] where Coord loss is the position loss function, Conf lossis the classification loss function, Class loss is the confidence loss function.

[0023] Preferably, the position loss function uses the CIoU function as the loss function for Saccharomyces cerevisiae detection, and its calculation formula is:

[0024]

[0025] where, b, b gt respectively represent the center points of the predicted bounding box and the ground truth box, ρ represents the Euclidean distance, and c represents the diagonal length of the smallest box covering the two bounding boxes;

[0026] α is a parameter used to balance the ratio, and its calculation formula is:

[0027]

[0028] v is used to measure the consistency of the width-to-height ratio between the predicted bounding box and the ground truth box, and its calculation formula is:

[0029]

[0030] where, and respectively represent the width-to-height ratios of the ground truth box and the predicted bounding box.

[0031] The present invention also provides a device for an invasive Saccharomyces cerevisiae detection method, including:

[0032] A labeling module for labeling the position and size of yeast cells in the yeast image;

[0033] A feature extraction module that uses the feature extraction network model GhostNet to extract the feature information of the yeast image in the labeled image dataset;

[0034] A strengthened feature extraction module that uses the FPN feature pyramid network to strengthen the feature extraction of the feature image dataset to obtain a strengthened feature image dataset;

[0035] A prediction module that predicts the strengthened feature image to obtain a yeast prediction image;

[0036] A training module that trains the yeast detection model until the loss function between the yeast prediction image and the labeled image converges to obtain a trained yeast detection model;

[0037] A yeast detection module for detecting the yeast image to be detected and outputting the position and size of the yeast in the image to be detected.

[0038] The present invention also provides a device for an invasive Saccharomyces cerevisiae detection method, comprising:

[0039] An image collector for collecting yeast images;

[0040] A memory for storing computer programs;

[0041] A processor for implementing the steps of an invasive Saccharomyces cerevisiae detection method as described in any one of claims 1 to 7 when executing the computer program.

[0042] The present invention also provides a computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, and the steps of the above-mentioned invasive Saccharomyces cerevisiae detection method are implemented when the computer program is executed by a processor.

[0043] For the invasive Saccharomyces cerevisiae detection method provided by the present invention, a feature extraction network model GhostNet is used to extract the feature information of yeast images in the training dataset. By using the Ghost module, more feature maps can be generated with fewer parameters, achieving a reduction in the total number of parameters and computational complexity of the feature extraction network. While ensuring the recognition accuracy, the recognition speed of the algorithm is improved; the Feature Pyramid Network (FPN) is used to enhance the feature extraction of the feature image training set. By using the FPN, it is beneficial to extract better image feature values, effectively reducing the problems of missed detection and false detection in the prior art and improving the detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0045] Figure 1 It is a flowchart of the first specific embodiment of the invasive Saccharomyces cerevisiae detection method provided by the present invention;

[0046] Figure 2 It is a structural diagram of the GhostNet network;

[0047] Figure 3 It is a normal convolution diagram;

[0048] Figure 4 It is a dilated convolution diagram;

[0049] Figure 5 It is a schematic diagram of a depthwise separable convolution;

[0050] Figure 6 This is the network structure diagram of the present invention;

[0051] Figure 7 This is the flowchart of the method of the present invention;

[0052] Figure 8 This is the diagram of the target object in the sample;

[0053] Figure 9 This is the diagram of the yeast detection result;

[0054] Figure 10 This is the structural block diagram of an invasive Saccharomyces cerevisiae detection device provided by the present invention. Detailed implementation manners

[0055] The core of the present invention is to provide an invasive Saccharomyces cerevisiae detection method, device and storage medium, which improve the accuracy and efficiency of Saccharomyces cerevisiae detection and effectively avoid the situations of missed detection and false detection.

[0056] In order to enable those skilled in the art to better understand the solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific implementation manners. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0057] Please refer to Figure 1 , Figure 1 This is the flowchart of the first specific embodiment of an invasive Saccharomyces cerevisiae detection method provided by the present invention; the specific operation steps are as follows:

[0058] Based on the YOLOv3 network, a yeast detection model is constructed. The yeast detection model includes: a feature extraction network GhostNet, an FPN feature pyramid network, and a prediction network;

[0059] Step S101: Use an image collector to collect multiple yeast images;

[0060] The image collector adopts the medical image acquisition microscope (Nikon ECLIPSE Ti2) model. The image collector is not limited to the Nikon ECLIPSE Ti2 model and can be existing products that meet the requirements, and is not limited in this embodiment of the present invention.

[0061] Step S102: Use annotation software to annotate the positions and sizes of yeast cells in the multiple yeast images to obtain an annotated image data set;

[0062] Step S103: Use the feature extraction network model GhostNet to extract the feature information of yeast images in the labeled image dataset, and obtain a feature image training set;

[0063] As Figure 2 Shown in the GhostNet network structure diagram, GhostNet is a lightweight network proposed by Huawei Noah's Ark Lab in 2020. It proposes a novel Ghost module that can generate more feature maps with fewer parameters. Specifically, the ordinary convolutional layer in the deep neural network is divided into two parts. The first part involves ordinary convolutions, but the total number of them will be strictly controlled. Given the inherent feature maps of the first part, a series of simple linear operations are then applied to generate more feature maps. Compared with ordinary convolutional neural networks, without changing the size of the output feature maps, the total number of parameters and the computational complexity required in this Ghost module are both reduced;

[0064] As Figure 3 Ordinary convolution diagram, Figure 4 Shown in the dilated convolution diagram, the first convolutional layer of the feature extraction network model GhostNet is a dilated convolution. Dilated convolution is a convolutional idea proposed for the problem of image semantic segmentation where downsampling will reduce the image resolution and lose information. It can increase the receptive field without losing information through pooling and under the same computational conditions, enabling each convolutional output to contain information in a larger range.

[0065] Step S104: Use the FPN feature pyramid network to enhance the feature extraction of the feature image training set, and obtain an enhanced feature image training set;

[0066] The FPN feature pyramid network includes ordinary convolution and depthwise separable convolution;

[0067] As Figure 5 Shown in the depthwise separable convolution schematic diagram, the depthwise separable convolution technology can greatly reduce the number of parameters, thereby greatly reducing the computational time and the size of the model, which is beneficial for deployment to portable devices. It is composed of two parts: depthwise (channel-wise convolution) and pointwise (pointwise convolution).

[0068] The FPN feature pyramid network uses the HardSwish activation function. The HardSwish activation function is an activation function of the neural network. It was proposed in the MobileNetV3 network and has advantages such as good numerical stability and fast computational speed. Its formula is as follows

[0069]

[0070] Among them, x is the neuron variable of the convolutional neural network.

[0071] Step S105: Use the prediction network to predict the enhanced feature image to obtain a yeast prediction image.

[0072] Step S106: Train the yeast detection model until the loss function between the yeast prediction image and the labeled image converges to obtain a trained yeast detection model;

[0073] The calculation method of the loss function of the yeast detection model is:

[0074] Loss yolov3 = ∑(Coord loss + Conf loss + Class loss )

[0075] Among them, Coord loss is the position loss function, Conf loss is the classification loss function, and Class loss is the confidence loss function.

[0076] The position loss function Coord loss uses the CIoU function as the loss function for Saccharomyces cerevisiae detection, and its calculation method is:

[0077]

[0078] Among them, b, b gt represent the center points of the predicted bounding box and the ground truth box respectively, ρ represents the Euclidean distance, and c represents the diagonal length of the smallest box covering the two bounding boxes;

[0079] α is a parameter used to balance the ratio, and its calculation method is:

[0080]

[0081] v is used to measure the consistency of the width-to-height ratio between the predicted bounding box and the ground truth box, and its calculation formula is:

[0082]

[0083] Among them, and represent the width-to-height ratios of the ground truth box and the predicted bounding box respectively.

[0084] Such as Figure 6As shown in the network structure diagram of the present invention, the yeast detection model extracts a total of three feature layers for object detection, uses these three feature layers to construct an FPN feature pyramid for enhanced feature extraction, and performs feature fusion on feature layers of different shapes, which is beneficial to extracting better features. The feature layer of 13x13x160 is subjected to 5 convolutional processes, and after the processing, the prediction result is obtained using YoloHead. A part of it is used for upsampling UmSampling2D and then combined with the 26x26x112 feature layer, and the shape of the combined feature layer is 26x26x368. The combined feature layer is subjected to 5 convolutional processes again, and after the processing, the prediction result is obtained using YoloHead. A part of it is used for upsampling UmSampling2D and then combined with the 52x52x40 feature layer, and the shape of the combined feature layer is 52x52x168. The combined feature layer is subjected to 5 convolutional processes again, and after the processing, the prediction result is obtained using YoloHead.

[0085] As Figure 7 As shown in the method structure flow chart of the present invention, an invasive Saccharomyces cerevisiae detection method provided in this embodiment replaces the feature extraction network in the original YOLOv3 network with a GhostNet network, replaces the first convolutional layer of the GhostNet network with a dilated convolution, changes some convolutional layers in the FPN feature pyramid network to depthwise separable convolutions, where some activation functions of the FPN feature pyramid network are changed to hardswish, and the position loss function is changed to a CIoU loss function. An invasive Saccharomyces cerevisiae detection model is constructed based on the improved YOLOv3 network, and the invasive Saccharomyces cerevisiae detection model is tested.

[0086] Step S107: Use the yeast detection model to detect the yeast image to be detected, and output the position and size of the yeast in the image to be detected.

[0087] An invasive Saccharomyces cerevisiae detection method provided by the present invention uses the feature extraction network model GhostNet to extract the feature information of the yeast images in the training dataset, which reduces the number of model parameters, decreases both the system calculation time and the model size, and is more conducive to being deployed to embedded devices, improving the recognition speed of the algorithm while ensuring the recognition accuracy; uses the FPN feature pyramid network to enhance the feature extraction of the feature image training set, performs feature fusion on feature layers of different scales, which is beneficial to extracting better features and improves the accuracy of object detection.

[0088] Based on the above embodiments, this embodiment further details the specific operations of yeast sample preparation and image acquisition of the present invention, as follows:

[0089] Step S201: Add 95 g of deionized water to a beaker and preheat it to 30 °C in a water bath heating box;

[0090] Step S202: Add 5 g of sucrose (0.2 M, Sinopharm Chemical Reagent Co., Ltd.) and 1 g of dry Saccharomyces cerevisiae to the beaker and gently stir with a glass rod;

[0091] Step S203: Place the beaker in a water bath heating box at 30 °C and continue to incubate for half an hour. After the Saccharomyces cerevisiae is fully activated, set it aside. Use a hemocytometer to count the Saccharomyces cerevisiae solution, and the concentration is about 7.3×107 cells / ml;

[0092] The blood samples are from volunteer donations, collected in anticoagulant tubes, and stored in a 4 °C refrigerator. Drop an appropriate amount of blood sample into a 1 ml centrifuge tube and centrifuge it at a speed of 1000 revolutions per minute for 5 minutes in a high-speed centrifuge. Then, remove the supernatant and add 1 ml of isotonic solution again and set it aside.

[0093] Step S204: Add 1 ml of the blood sample to 1 ml of the Saccharomyces cerevisiae solution, and place the mixture in a 4 °C refrigerator for later use. Before collecting images, ultrasonically process the mixed solution for 5 minutes to make the cells evenly distributed in the solution.

[0094] Step S205: Drop 10 μl of the cell mixture sample on a glass slide, cover it with a coverslip and press it to make a sample, and place it on the stage of a microscope (Nikon ECLIPSE Ti2, 400-fold magnification);

[0095] As Figure 8 shown by the target object in the sample, move the stage and randomly collect 500 sample images. The image resolution is 2454×1632. Manually annotate the Saccharomyces cerevisiae in each image and record the position and size of each Saccharomyces cerevisiae.

[0096] Figure 9 These are the yeast detection results for this example;

[0097] Divide the collected images into a training set, a validation set, and a test set according to a ratio of 8:1:1. Improve the original YOLOv3 network, replace the feature extraction network with the GhostNet network, and set the first convolutional layer of the GhostNet network as a dilated convolution. Replace the convolutional layers in the detection part with depthwise separable convolutions, change the activation function to the HardSwish activation function, and finally use the CIoU function as the loss function of the algorithm. Then train the improved YOLOv3 on the training set and verify it on the validation set. After training is completed, test the effect on the test set.

[0098] Using the data of this embodiment for detection, the present invention introduces the detection method of convolutional neural network into the detection of invasive Saccharomyces cerevisiae infection diseases, realizing the rapid detection of Saccharomyces cerevisiae cells in blood, and the AP (average precision) reaches 97.24%. At the same time, it also provides a new idea and method for the detection of pathogenic fungi such as Candida and Cryptococcus.

[0099] In summary, the present invention uses depthwise separable convolution technology to reduce the number of model parameters, greatly reducing the calculation time and the size of the model; uses the HardSwish activation function to replace the Swish function, enhancing the digital stability of the system and improving the calculation speed while ensuring the system accuracy; uses the CIoU function to replace the original IoU loss function, reducing the impact on network learning and training and improving network performance. In summary, the invasive Saccharomyces cerevisiae detection method provided by the present invention uses fewer parameters to generate more feature maps, realizes the reduction of the total number of feature extraction network parameters and the calculation complexity, improves the recognition speed of the algorithm while ensuring the recognition accuracy, is conducive to extracting better image feature values, effectively reduces the problems of missed detection and false detection in the prior art, and improves the detection accuracy.

[0100] Please refer to Figure 10 , Figure 10 which is the structural block diagram of an invasive Saccharomyces cerevisiae detection device provided by the present invention. The specific device may include:

[0101] The annotation module 100 is used to annotate the position and size of yeast cells in the yeast image;

[0102] The feature extraction module 200 is used to extract the feature information of the yeast image in the annotated image dataset by using the feature extraction network model GhostNet;

[0103] The enhanced feature extraction module 300 is used to enhance the feature extraction of the feature image training set by using the FPN feature pyramid network to obtain an enhanced feature image training set;

[0104] The prediction module 400 predicts the enhanced feature image to obtain a yeast prediction image;

[0105] The training module 500 is used to train the yeast detection model until the loss function between the yeast prediction image and the annotated image converges, and obtain a trained yeast detection model;

[0106] The yeast detection module 600 is used to detect the yeast image to be detected and output the position and size of the yeast in the image to be detected.

[0107] An invasive Saccharomyces cerevisiae detection device according to this embodiment is used to implement the aforementioned invasive Saccharomyces cerevisiae detection method. Therefore, the specific implementation manners in an invasive Saccharomyces cerevisiae detection device can be seen in the embodiment part of the aforementioned invasive Saccharomyces cerevisiae detection method. For example, the annotation module 100, the feature extraction module 200, the enhanced feature extraction module 300, the prediction module 400, the training module 500, and the yeast detection module 600 are respectively used to implement steps S101, S102, S103, S104, S105, S106, and S107 in the aforementioned invasive Saccharomyces cerevisiae detection method. Therefore, the specific implementation manners can refer to the descriptions of the corresponding parts of each embodiment and will not be elaborated herein.

[0108] A specific embodiment of the present invention further provides a device for an invasive Saccharomyces cerevisiae detection method, including: an image collector for collecting yeast images; a memory for storing computer programs; and a processor for implementing the steps of the aforementioned invasive Saccharomyces cerevisiae detection method when executing the computer programs.

[0109] A specific embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and the computer program implements the steps of the aforementioned invasive Saccharomyces cerevisiae detection method when executed by a processor.

[0110] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0111] Those skilled in the art can further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0112] The steps of the methods or algorithms described in connection with the embodiments disclosed herein may be implemented directly in hardware, in software modules executed by a processor, or in a combination thereof. The software modules may be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium well known in the art.

[0113] The above has introduced in detail a method, apparatus, and storage medium for detecting invasive Saccharomyces cerevisiae provided by the present invention. Specific examples are used herein to illustrate the principles and implementation manners of the present invention. The description of the above embodiments is only for helping to understand the method and its core idea of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle of the present invention, several improvements and modifications can still be made to the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. An invasive Saccharomyces cerevisiae detection method, characterized in that, including: Construct an invasive Saccharomyces cerevisiae detection model based on the YOLOv3 network. The yeast detection model includes: a feature extraction network GhostNet, an FPN feature pyramid network, and a prediction network; Use an image collector to collect multiple yeast images; Use annotation software to annotate the positions and sizes of yeast cells in the multiple yeast images to obtain an annotated image dataset; Use the feature extraction network model GhostNet to extract the feature information of the yeast images in the annotated image dataset to obtain a feature image training set; the first convolutional layer of the feature extraction network model GhostNet is a dilated convolution; Use the FPN feature pyramid network to enhance feature extraction for the feature image training set to obtain an enhanced feature image training set; the FPN feature pyramid network includes a conventional convolution and a depthwise separable convolution; the depthwise separable convolution consists of two parts: a depthwise convolution and a pointwise convolution; Use the prediction network to predict the enhanced feature image to obtain a yeast prediction image; Train the yeast detection model until the loss function between the yeast prediction image and the annotated image converges to obtain a trained yeast detection model; Use the yeast detection model to detect the yeast image to be detected and output the positions and sizes of the yeast in the image to be detected.

2. The invasive Saccharomyces cerevisiae detection method according to claim 1, characterized in that, The FPN feature pyramid network uses the HardSwish activation function, and its formula is as follows ； Among them, is the neuron variable of the convolutional neural network.

3. The invasive Saccharomyces cerevisiae detection method according to claim 1, characterized in that, The calculation formula of the loss function of the yeast detection model is: ； Among them, is the position loss function, is the classification loss function, is the confidence loss function.

4. The invasive Saccharomyces cerevisiae detection method according to claim 3, characterized in that, The position loss function uses the CIoU function as the loss function for Saccharomyces cerevisiae detection, and its calculation formula is: ； Among them, respectively represent the center points of the predicted bounding box and the ground truth box, represents the Euclidean distance, and c represents the diagonal length of the smallest box covering the two bounding boxes; is a parameter for balancing the ratio, and its calculation formula is: ； To measure the consistency of the width-to-height ratio between the predicted bounding box and the ground truth box, the calculation formula is as follows: ； Among them, and respectively represent the aspect ratios of the target ground truth box and the predicted bounding box.

5. An apparatus for an invasive Saccharomyces cerevisiae detection method, characterized in that, including: An annotation module for annotating the positions and sizes of yeast cells in the yeast image; A feature extraction module that uses the feature extraction network model GhostNet to extract the feature information of the yeast images in the annotated image dataset; the first convolutional layer of the feature extraction network model GhostNet is a dilated convolution; An enhanced feature extraction module that uses the FPN feature pyramid network to enhance feature extraction for the feature image dataset to obtain an enhanced feature image dataset; the FPN feature pyramid network includes a conventional convolution and a depthwise separable convolution; the depthwise separable convolution consists of two parts: a depthwise convolution and a pointwise convolution; A prediction module that predicts the enhanced feature image to obtain a yeast prediction image; A training module that trains the yeast detection model until the loss function between the yeast prediction image and the annotated image converges to obtain a trained yeast detection model; A yeast detection module for detecting the yeast image to be detected and outputting the positions and sizes of the yeast in the image to be detected.

6. A device for an invasive Saccharomyces cerevisiae detection method, characterized in that, including: An image collector for collecting yeast images; A memory for storing computer programs; A processor for implementing the steps of an invasive Saccharomyces cerevisiae detection method as described in any one of claims 1 to 4 when executing the computer program.

7. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by the processor, it implements the steps of an invasive Saccharomyces cerevisiae detection method as described in any one of claims 1 to 4.

Images