A method and system for in vitro assisted assessment of liver fibrosis

Abstract

The present application relates to a kind of liver fibrosis in vitro auxiliary evaluation method and system, belong to image processing technical field, the method includes: obtaining based on plasma induced generation of uric acid crystal;Uric acid crystal is carried out image acquisition, and the crystal fluorescence image is obtained, and the crystal fluorescence image is preprocessed, and the target fluorescence image is obtained;Based on image segmentation model, the target fluorescence image is segmented, and a plurality of single crystal images are obtained;Based on target clinical task, the single crystal image of the same person to be measured is input into the deep learning model corresponding to target clinical task, and the evaluation result related to liver fibrosis is obtained.The method realizes the automated liver fibrosis in vitro auxiliary evaluation, improves the accuracy and efficiency of liver fibrosis in vitro auxiliary evaluation, avoids the trauma and risk of puncture biopsy.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, and in particular relates to an in vitro auxiliary assessment method and system for liver fibrosis. Background Technology

[0002] Liver fibrosis is a pathological process resulting from repeated chronic liver injury and an imbalance in repair, and it is a key step in the progression of many chronic liver diseases to cirrhosis and even liver cancer. Because liver fibrosis is reversible to a certain extent, its early detection, dynamic assessment, and stratified management are of significant clinical importance for delaying disease progression, guiding clinical intervention, and improving patient prognosis. Currently, liver biopsy is still considered the gold standard for assessing liver fibrosis; however, this method is invasive, carries a risk of infection, has low patient acceptance, and suffers from subjective judgment due to the limited sample volume, making it difficult to achieve large-sample repeat testing and continuous dynamic monitoring. Therefore, how to better achieve in vitro auxiliary assessment of liver fibrosis has become an urgent problem to be solved. Summary of the Invention

[0003] In view of the shortcomings of the prior art, the purpose of the invention is to provide an in vitro auxiliary assessment method and system for liver fibrosis.

[0004] In a first aspect, the present invention provides an in vitro auxiliary assessment method for liver fibrosis, comprising: acquiring uric acid crystals generated based on plasma; acquiring images of the uric acid crystals to obtain crystal fluorescence images, and preprocessing the crystal fluorescence images to obtain target fluorescence images; segmenting the target fluorescence images based on an image segmentation model to obtain multiple single crystal images; and inputting the single crystal images of the same subject into a deep learning model corresponding to the target clinical task based on a target clinical task to obtain assessment results related to liver fibrosis.

[0005] A second aspect of the present invention provides an in vitro auxiliary assessment system for liver fibrosis, comprising: an acquisition module for acquiring uric acid crystals generated based on plasma; an acquisition and preprocessing module for acquiring images of the uric acid crystals to obtain crystal fluorescence images, and preprocessing the crystal fluorescence images to obtain target fluorescence images; a segmentation module for segmenting the target fluorescence images based on an image segmentation model to obtain multiple single crystal images; and an acquisition module for inputting the single crystal images of the same subject into a deep learning model corresponding to the target clinical task based on a target clinical task to obtain assessment results related to liver fibrosis.

[0006] A third aspect of the present invention provides an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method described in any one aspect of the present invention.

[0007] A fourth aspect of the present invention provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to perform the method described in any one of the first aspects of the present invention.

[0008] The beneficial effects of this invention are as follows: The present invention describes an in vitro auxiliary assessment method and system for liver fibrosis, which involves acquiring uric acid crystals induced by plasma; acquiring images of the uric acid crystals to obtain fluorescence images, and preprocessing these fluorescence images to obtain target fluorescence images; segmenting the target fluorescence images using an image segmentation model to obtain multiple single-crystal images; and inputting the single-crystal images of the same subject into a deep learning model corresponding to the target clinical task to obtain assessment results related to liver fibrosis. This method achieves automated in vitro auxiliary assessment of liver fibrosis, improves the accuracy and efficiency of in vitro auxiliary assessment of liver fibrosis, and avoids the trauma and risks of puncture biopsy. Attached Figure Description

[0009] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. It is obvious that the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings.

[0010] Figure 1 This is a flowchart of an in vitro auxiliary assessment method for liver fibrosis according to an embodiment of the present invention; Figure 2 This is a flowchart of an in vitro auxiliary assessment method for liver fibrosis according to a specific embodiment of the present invention; Figure 3 This is an evaluation diagram of a deep learning model that assists in distinguishing between health and liver fibrosis according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the evaluation of a deep learning model corresponding to the auxiliary screening of liver fibrosis risk according to an embodiment of the present invention; Figure 5 This is an evaluation diagram of a deep learning model for assisting in the identification of early liver fibrosis according to an embodiment of the present invention; Figure 6This is a schematic diagram of an in vitro auxiliary assessment system for liver fibrosis according to an embodiment of the present invention; Figure 7 This is a structural block diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0011] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0012] Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts disclosed in this invention.

[0013] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. The terms "installed," "connected," and "linked" should be interpreted broadly; for example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0014] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of methods and systems consistent with some aspects of the invention as detailed in the appended claims.

[0015] This invention proposes an in vitro auxiliary assessment method, system, and related equipment for liver fibrosis. Specifically, the embodiments of the in vitro auxiliary assessment method, system, and related equipment for liver fibrosis of this invention are described below with reference to the accompanying drawings.

[0016] Figure 1 This is a flowchart of an in vitro auxiliary assessment method for liver fibrosis according to an embodiment of the present invention. It should be noted that the in vitro auxiliary assessment method for liver fibrosis according to this embodiment can be applied to the in vitro auxiliary assessment system for liver fibrosis according to this embodiment. This in vitro auxiliary assessment system for liver fibrosis can be configured on an electronic device or in a server. This application does not limit the scope of the application.

[0017] like Figure 1 As shown, the in vitro auxiliary assessment methods for liver fibrosis include: S110, Obtain uric acid crystals based on plasma-induced generation.

[0018] In an embodiment of the present invention, plasma from a subject is obtained; a supersaturated sodium urate solution is prepared based on uric acid powder and sodium hydroxide solution; a reaction system is constructed based on sodium acetate, the plasma sample, and the supersaturated sodium urate solution; and uric acid crystals are induced to form based on the reaction system, reaction plate, incubation time, and temperature. Thus, uric acid crystals induced from plasma can be obtained.

[0019] In this process, peripheral venous blood from the subjects was collected using EDTA anticoagulant tubes, centrifuged at 3500 rpm for 5 minutes, the supernatant plasma was separated and transferred to EP tubes, and then frozen at -80°C for later use to obtain the subjects' plasma.

[0020] In this process, 0.1 g of high-purity uric acid powder is weighed and added to 3 mL of 0.667 mol / L sodium hydroxide solution. The mixture is thoroughly mixed until completely dissolved to form a supersaturated sodium urate solution.

[0021] In this process, 60 μL of acetate-sodium acetate buffer (pH adjusted to 4.5) and 2.5 μL of plasma sample were added sequentially to a 384-well plate to acidify the system. Finally, 2 μL of saturated sodium urate solution was added and quickly mixed to complete the construction of the reaction system.

[0022] The reaction plate was placed in a 37°C incubator and incubated for 1 hour to induce the formation of uric acid crystals.

[0023] This invention constructs a plasma-induced uric acid crystal system in vitro, making the crystal itself a carrier of signal and information amplification. This approach goes beyond the direct measurement of existing indicators in samples; instead, it introduces new detectable targets through the crystal formation process, thus forming a detection strategy different from existing technologies.

[0024] S120 acquires images of uric acid crystals to obtain crystal fluorescence images, and preprocesses the crystal fluorescence images to obtain target fluorescence images.

[0025] In an embodiment of the present invention, a crystal fluorescence image can be acquired using an acquisition device.

[0026] For example, an inverted microscope and a high-resolution camera can be used to acquire crystal fluorescence images in the FITC channel. The image resolution should be no less than 1024×1024 pixels, the magnification should be controlled within the range of 2×-10×, and the LUT linear mapping range, gain parameters, and exposure time should be kept consistent to ensure image standardization and traceability.

[0027] Preprocessing of the crystal fluorescence image includes noise reduction and normalization.

[0028] S130, based on the image segmentation model, segments the target fluorescence image to obtain multiple single crystal images.

[0029] In an embodiment of the present invention, after obtaining a target fluorescence image, the target fluorescence image can be input into an image segmentation model to obtain multiple single-crystal images. Specific implementation details can be found in subsequent embodiments.

[0030] S140, based on the target clinical task, inputs the single crystal image of the same subject into the deep learning model corresponding to the target clinical task to obtain the assessment results related to liver fibrosis.

[0031] In embodiments of the present invention, the target clinical tasks include assisting in distinguishing between health and liver fibrosis, assisting in screening for liver fibrosis risk, or assisting in identifying early liver fibrosis.

[0032] In embodiments of the present invention, single-crystal images are input into deep learning models corresponding to different target clinical tasks to obtain assessment results related to liver fibrosis. Specific implementation methods can be found in subsequent embodiments.

[0033] According to an embodiment of the present invention, an in vitro auxiliary assessment method for liver fibrosis involves obtaining uric acid crystals generated based on plasma induction; acquiring images of the uric acid crystals to obtain crystal fluorescence images; preprocessing the crystal fluorescence images to obtain target fluorescence images; segmenting the target fluorescence images based on an image segmentation model to obtain multiple single crystal images; and inputting the single crystal images of the same subject into a deep learning model corresponding to the target clinical task to obtain assessment results related to liver fibrosis. This method achieves automated in vitro auxiliary assessment of liver fibrosis, improves the accuracy and efficiency of in vitro auxiliary assessment of liver fibrosis, and avoids the trauma and risks of puncture biopsy.

[0034] To enable those skilled in the art to more readily understand the present invention, Figure 2 This is an in vitro auxiliary assessment method for liver fibrosis according to a specific embodiment of the present invention, such as... Figure 2 As shown, this in vitro auxiliary assessment method for liver fibrosis includes: S210, Obtain uric acid crystals based on plasma-induced generation.

[0035] S220 acquires images of uric acid crystals to obtain crystal fluorescence images, and preprocesses the crystal fluorescence images to obtain target fluorescence images.

[0036] In the embodiments of the present invention, the implementation of steps S210-S220 can refer to the implementation of steps S110-S110 described above, and the present invention will not repeat the details.

[0037] S230 uses an image segmentation model to segment the target fluorescence image, resulting in multiple single-crystal images.

[0038] In embodiments of the present invention, the image segmentation model is trained as follows: a dataset is acquired and divided into a training set, a validation set, and a test set, wherein the dataset includes manually labeled crystal fluorescence image samples; a deep learning-based crystal instance segmentation model is constructed, which is used to identify and segment individual crystal instances in crystal fluorescence images and outputs instance segmentation results corresponding to the input image, wherein the instance segmentation results include multiple crystal instance regions and their corresponding instance numbers; in one embodiment, the crystal instance segmentation model uses Cellpose. An instance segmentation framework, or a deep learning segmentation network including an encoder and decoder, is used. Training parameters and loss functions and evaluation metrics are defined. Training sets are batch-input into the crystal instance segmentation model, forward propagation yields the prediction mask, loss values ​​are calculated, and backpropagation updates the network parameters. After each training epoch, the Dice coefficient and Intersection over Union (IoU) are calculated on the validation set. When the validation set performance no longer improves after several consecutive epochs, training stops, and the network weights with the best overall performance on the validation set are saved as the trained image segmentation model. The test set is input into the trained image segmentation model, and the prediction mask is output. If the Dice coefficient and / or IoU of the validation set reach a preset performance threshold, or if the improvement in the Dice coefficient and / or IoU of the validation set is less than a preset convergence threshold within K consecutive training epochs, then the trained image segmentation model is complete. The preset performance threshold can be set according to the annotation quality and application scenario. In one implementation, the preset thresholds for both the Dice coefficient and IoU are 0.9. The test set is used for the final generalization performance evaluation.

[0039] In one implementation, the training parameters include a learning rate, batch size, and number of training epochs. For example, the learning rate can be set to 0.0001 or a learning rate decay strategy can be used; the batch size can be set to 8; and the number of training epochs can be determined based on the convergence of the validation set performance. An early stopping strategy can be used during training, stopping training when the validation set performance meets a preset threshold or tends to stabilize.

[0040] The prediction mask can be understood as the decision map of the image segmentation model for each pixel in the image to belong to the crystal or the background. For any crystal instance, the instance segmentation mask can be further used to generate the corresponding binary mask, where white represents the crystal instance region and black represents the non-crystal instance region.

[0041] In an embodiment of the present invention, the target fluorescence image is input into an image segmentation model to obtain an instance segmentation mask containing multiple crystal instance numbers; based on the instance segmentation mask, contour extraction, bounding box localization, and ROI cropping are performed on each crystal instance to obtain multiple single crystal images.

[0042] S240, Define the target clinical task.

[0043] In embodiments of the present invention, the target clinical tasks include assisting in distinguishing between health and liver fibrosis, assisting in screening for liver fibrosis risk, or assisting in identifying early liver fibrosis.

[0044] S250, when the target clinical task is to help distinguish between healthy individuals and liver fibrosis, inputs single-crystal images into the deep learning model corresponding to the target clinical task to obtain assessment results related to liver fibrosis.

[0045] In an embodiment of the present invention, when the target clinical task is to assist in distinguishing between healthy individuals and liver fibrosis, the deep learning model corresponding to the target clinical task is trained in the following manner: A single-crystal image dataset for model training is obtained, comprising multiple single-crystal image samples from multiple subjects, and the single-crystal image samples are standardized. The standardized single-crystal image samples are labeled, and then divided into training, validation, and test sets using a hierarchical grouping cross-validation method (to prevent multiple single-crystal images of the same subject from appearing simultaneously in the training, validation, or test sets, thus avoiding data leakage). The images are labeled as either healthy or with liver fibrosis. A single single-crystal image from the training set is input into a deep learning model built on a ConvNeXt deep convolutional neural network, where the backbone network of the deep learning model is ConvNeXt_ The base network, a fully connected binary classification head, outputs the predicted probability of health or liver fibrosis. The predicted probabilities are compared with the true labels in the single-crystal image, and a cross-entropy loss function is used to determine the loss value between the predicted probabilities and the true labels. Based on the loss value, the gradients of the trainable parameters in the backbone network ConvNeXt_base are calculated, and the parameters of the backbone network and the classification head are updated until the preset number of training iterations is reached. After each training epoch, image-level and / or subject-level classification performance metrics are calculated on the validation set. These performance metrics include one or more of accuracy, AUC, F1 score, precision, recall, and validation loss. The optimal training epoch is selected based on the subject-level AUC, F1 score, or overall classification performance on the validation set, and the corresponding model weights are saved as the trained deep learning model.

[0046] In embodiments of the present invention, single-crystal images of the same subject are input into a deep learning model corresponding to the target clinical task, outputting the probability of health or liver fibrosis; the probabilities of health or liver fibrosis corresponding to multiple single-crystal images of the same subject are obtained; and the evaluation result is determined based on a probability averaging algorithm and the probabilities of health or liver fibrosis corresponding to multiple single-crystal images. For example, the average of the probabilities of health or liver fibrosis corresponding to multiple single-crystal images is obtained. When the average probability of liver fibrosis is greater than or equal to the average probability of health, the evaluation result is liver fibrosis; otherwise, the evaluation result is health.

[0047] S260, when the target clinical task is to assist in screening for the risk of liver fibrosis, inputs single-crystal images into the deep learning model corresponding to the target clinical task to obtain assessment results related to liver fibrosis.

[0048] In an embodiment of the present invention, when the target clinical task is to assist in screening for the risk of liver fibrosis, the deep learning model corresponding to the target clinical task includes an image depth feature extraction branch and a crystal morphology feature branch. The image depth feature extraction branch is used to automatically extract the depth features of a single crystal image, and the morphology feature branch is used to extract or input the morphology, grayscale, texture, and spatial distribution features of the single crystal.

[0049] The deep learning model corresponding to the target clinical task is trained in the following way: A binary classification dataset for model training is obtained, comprising multiple single-crystal image samples from multiple subjects, and these single-crystal image samples are standardized. The standardized single-crystal image samples are then labeled, and the standardized binary classification dataset is divided into training, validation, and test sets using a hierarchical grouping cross-validation method (ensuring that multiple single-crystal images corresponding to the same subject do not appear simultaneously in the training, validation, or test sets to avoid data leakage). The images are labeled as either non-fibrillated or exhibiting varying degrees of fibrillation. A single single-crystal image from the training set is then input into an image depth-specific... In the feature extraction branch, an N-dimensional depth image feature vector is output after feature extraction and pooling processing. The pooling process uses Global Average Pooling (GAP). The depth feature branch is constructed based on a ConvNeXt deep convolutional neural network, with ConvNeXt_base as the backbone network. A single crystal image from the training set is input to the crystal morphology feature branch, outputting an M-dimensional crystal morphology feature vector. The N-dimensional depth feature vector and the M-dimensional crystal morphology feature vector are fused to obtain a fused feature, which is then input to the classification layer to output the predicted positive category value corresponding to the target clinical task. This predicted value is then converted to a predicted positive category probability using a Sigmoid function. Specifically, in the task of assisting in screening for liver fibrosis risk, the positive categories are F1-F4; in the task of assisting in identifying early liver fibrosis, the positive category is F1. Based on the predicted positive category value and the true label, a binary cross-entropy loss function is used to calculate the loss value. In another implementation, the classification layer outputs predicted values ​​for two mutually exclusive categories, which are then converted into predicted probabilities for the two categories using a Softmax function and trained using a cross-entropy loss function. Based on the loss values, the gradients of the trainable parameters in the backbone network ConvNeXt_base are calculated, and the trainable parameters in the image depth feature extraction branch, crystal morphology feature branch, fusion layer, and classification layer are updated until the number of training iterations reaches a preset number. The classification accuracy is calculated on the validation set, and the deep learning model training is complete when the classification accuracy is greater than a preset threshold.

[0050] Among them, crystal morphological characteristics include one or more of the following: crystal area, perimeter, aspect ratio, roundness, convexity, gray intensity, edge gradient, center-edge intensity ratio, and area ratio. In an embodiment of the present invention, a single-crystal image of the same subject is input into a deep learning model corresponding to the target clinical task, and the predicted probability of the positive category corresponding to the single-crystal image is output. The predicted probabilities of the positive categories corresponding to multiple single-crystal images are averaged to obtain the subject-level positive prediction probability. Based on the subject-level positive prediction probability and a preset threshold, the evaluation result is determined. For example, when the subject-level positive prediction probability is greater than or equal to the preset threshold, a positive risk of liver fibrosis is output; otherwise, a no-fibrosis-related evaluation result is output.

[0051] S270, when the target clinical task is to assist in the identification of early liver fibrosis, inputs the single crystal image into the deep learning model corresponding to the target clinical task to obtain the evaluation results related to liver fibrosis.

[0052] In an embodiment of the present invention, the target clinical task is to assist in the identification of early liver fibrosis. The method of inputting single-crystal images into the deep learning model corresponding to the target clinical task to obtain assessment results related to liver fibrosis can refer to the method described above where the target clinical task is to assist in screening for liver fibrosis risk, and the method of inputting single-crystal images into the deep learning model corresponding to the target clinical task to obtain assessment results related to liver fibrosis is also described. This invention will not elaborate further on this aspect.

[0053] The deep learning models for assisting in the identification of early liver fibrosis and those for assisting in the screening of liver fibrosis risk share the same technical approach, differing only in their training, testing, and validation sets, as well as their labeling. In other words, the deep learning models for assisting in the identification of early liver fibrosis and those for assisting in the screening of liver fibrosis risk can employ the same or similar network structures and training processes, but their sample inclusion scope, category labeling, positive category definition, and model parameters can differ. Specifically, the deep learning model for assisting in the screening of liver fibrosis risk labels single-crystal image samples as either non-fibrotic or exhibiting varying degrees of fibrosis risk, while the deep learning model for assisting in the identification of early liver fibrosis labels single-crystal image samples as either non-fibrotic or early fibrosis.

[0054] In the first embodiment of this invention, peripheral venous blood was collected from the subject, and plasma was separated into EP tubes and stored at -80°C for later use. During detection, the plasma sample was reconstituted and added to a 384-well plate with acetate-sodium acetate buffer, followed by the addition of supersaturated sodium urate solution. The plate was then incubated at 37°C for 1 hour to induce the formation of uric acid crystals. Within a certain observation period, the uric acid crystals induced by plasma showed no significant changes in morphology or fluorescence, indicating that the detection system constructed in this invention has good stability. This characteristic helps reduce random errors during the detection process and provides a technical basis for repeated detection, delayed interpretation, and standardized experimental procedures. Furthermore, plasma itself has weak autofluorescence, requiring specialized instruments for detection. However, uric acid crystals induced by plasma significantly enhance the fluorescence signal, and the difference between liver fibrosis samples and healthy samples is more pronounced at the crystal level. Therefore, this invention solves the problems of weak signal and insufficient inter-group differences when directly detecting plasma, improving the clarity and discriminative ability of the detection results.

[0055] After crystallization, inverted fluorescence microscopy was used to acquire crystal fluorescence images, preferably using the FITC channel, and maintaining consistent exposure time, gain parameters, and LUT linear mapping range to ensure standardized acquisition of crystal fluorescence images. Subsequently, background correction and crystal region identification were performed on the crystal fluorescence images. Then, the preprocessed crystal fluorescence image (target fluorescence image) was segmented into individual crystal images. These individual crystal images were input into a deep learning model based on a ConvNeXt deep convolutional neural network, which outputs a prediction result indicating whether the single crystal is healthy or has liver fibrosis. The predicted probabilities of multiple individual crystal images from the same sample source were statistically fused, preferably using a probability averaging algorithm, to obtain the evaluation result. Figure 3 As shown, in the internal test data, the deep learning model based on single-crystal fluorescence images can simultaneously capture crystal fluorescence intensity, fluorescence spatial distribution, and morphological information, demonstrating good ability to distinguish between healthy and liver fibrosis. Its classification performance is better than schemes that use only single fluorescence features or single morphological features, indicating that combined feature analysis helps improve the accuracy and robustness of sample-assisted assessment. (AUC of crystal fluorescence combined with morphology = 1.0; AUC of fluorescence intensity only = 0.881; AUC of morphology only = 0.991).

[0056] In the second embodiment of the present invention, the sample collection, plasma preparation, uric acid crystal induction, and crystal fluorescence image acquisition methods can be performed in accordance with the first embodiment described above. The difference is that in this embodiment, the data is labeled according to the liver fibrosis stage information of the samples, wherein the F0 sample is set as the negative class, and the F1-F4 samples are set as the positive class, so as to construct a binary classification dataset of no liver fibrosis and different degrees of liver fibrosis.

[0057] In this embodiment, single-crystal images can be directly input into a deep learning model for end-to-end training; alternatively, manual features such as crystal area, perimeter, Feret diameter, aspect ratio, roundness, convex hull solidity, grayscale statistics, gradient features, center / edge grayscale ratio, and area proportion can be extracted and fused with image depth features before being input into a classifier to output single-image-level risk screening results. Subsequently, multiple single-crystal results from the same sample source are summarized to obtain sample-level liver fibrosis risk screening results. This embodiment demonstrates that the present invention can not only assist in distinguishing between healthy individuals and those with liver fibrosis but also serve as a liver fibrosis risk screening tool for in vitro auxiliary determination of F0 (i.e., no fibrosis) and F1-F4 (i.e., different degrees of liver fibrosis) samples. Figure 4 As shown, in the internal test data, the deep learning model of this invention has a good ability to distinguish between F0 and F1-F4 samples (AUC=0.955), indicating that the technical solution can be used for in vitro auxiliary judgment in the scenario of liver fibrosis risk screening.

[0058] In the third embodiment of the present invention, the sample collection, plasma preparation, uric acid crystal induction, and crystal fluorescence image acquisition methods can also be performed in accordance with the first embodiment described above. The difference is that this embodiment only screens samples with liver fibrosis stages F0 and F1, where F0 is set as the negative class and F1 as the positive class, in order to construct a binary classification dataset for early identification.

[0059] After uniform preprocessing of the acquired single-crystal images, classification can be performed directly using a ConvNeXt-based deep learning model; alternatively, a fusion of deep features and handcrafted crystal features can be employed to enhance the model's ability to identify subtle differences between F0 and F1. Handcrafted features may include crystal area, perimeter, aspect ratio, roundness, grayscale statistics, and fluorescence distribution parameters. Finally, the prediction results from multiple single-crystal images are summarized into early identification results at the sample level to determine whether the sample exhibits early liver fibrosis-related changes. This embodiment demonstrates that the present invention can be further applied to distinguish between F0 (i.e., no fibrosis) and F1 (i.e., early fibrosis), thereby providing an in vitro detection method based on ex vivo plasma-induced uric acid crystal fluorescence images for early auxiliary identification of liver fibrosis. Figure 5 As shown, in a set of internal test data, the deep learning model of this invention can effectively distinguish between F0 and F1 samples (AUC=0.950), indicating that the technical solution has the potential to be used for early auxiliary identification of liver fibrosis.

[0060] In embodiments of the present invention, other crystallizable systems can be used to replace the uric acid crystal system, provided that the crystallization process can be regulated by plasma components, an imageable crystal structure can be formed, and detectable fluorescence differences can be observed. The composition and parameters of the reaction system can be adjusted while ensuring that crystal formation and fluorescence signals can be induced. The buffer solution can be an acetate-sodium acetate system or other buffer systems that provide an acidic environment; the pH range can be adjusted within a weakly acidic range; the plasma volume, uric acid solution concentration, and ratio can be optimized; the reaction temperature can be room temperature or physiologically relevant temperatures such as 37°C; the incubation time can be extended or shortened depending on the crystal formation. Different data processing paths can be used for extracting and determining crystal features. Parameters such as fluorescence intensity, signal-to-noise ratio, and coefficient of variation can be extracted manually; morphological features such as crystal area, perimeter, and roundness can be extracted; traditional statistical models (such as logistic regression and SVM) can be used for determination; deep learning models can be used to automatically extract latent features and perform classification; and multi-feature fusion or weighted scoring methods can be used for result output.

[0061] An in vitro auxiliary assessment method for liver fibrosis according to an embodiment of the present invention: 1. Using ex vivo plasma for in vitro testing offers the advantages of being non-invasive, repeatable, and providing standardized, image-based output.

[0062] Using ex vivo plasma as the testing subject eliminates the need for invasive procedures such as liver biopsy, making it more acceptable to patients and more suitable for repeated testing, dynamic follow-up, and long-term monitoring. Furthermore, it can output crystal fluorescence images, crystal morphology images, characteristic parameters, and model analysis results, rather than just single numerical values ​​or scores. Therefore, the test results are more intuitive, easier to save and compare, and more conducive to establishing standardized databases, batch management, and quality control. Compared to invasive examination methods or technical solutions that only output a single indicator, this invention offers better compliance, visualization, and standardization potential in practical applications.

[0063] 2. More suitable for auxiliary screening, stratification and dynamic assessment of liver fibrosis.

[0064] As the disease progresses, the induced fluorescence of the crystals gradually increases, suggesting that it can not only be used to assist in distinguishing between healthy samples and liver fibrosis samples, but also has the potential for further application in the stratification, grading, and risk assessment of liver fibrosis. Therefore, the clinical application of this invention is not limited to a simple "presence or absence" judgment, but can be extended to more refined disease management scenarios.

[0065] 3. It is easy to integrate with automated equipment, software systems, and product implementation methods, and has good prospects for clinical translation and industrialization.

[0066] The technical process includes standardized steps such as sample processing, induced crystallization, image acquisition, feature extraction, and model analysis. Therefore, it can be implemented as a detection method, or further developed into a system, device, reagent kit, or computer program product. This feature gives the invention good industrial adaptability and patent layout extensibility. Simultaneously, by constructing a novel crystal fluorescence detection system, signal enhancement, multidimensional feature analysis, and intelligent output are achieved, which is expected to provide a simple, low-invasive, repeatable, and automated in vitro auxiliary assessment method for liver fibrosis. Therefore, it has good application prospects in clinical auxiliary detection, standardized laboratory testing, and subsequent product transformation.

[0067] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of this application.

[0068] According to one aspect of the present invention, an in vitro auxiliary assessment system for liver fibrosis is also provided. Figure 6 This is a schematic diagram of an in vitro auxiliary assessment system for liver fibrosis according to an embodiment of the present invention; as shown. Figure 6 As shown, it includes: Acquisition module 610 is used to acquire uric acid crystals generated based on plasma induction; The acquisition and preprocessing module 620 is used to acquire images of the uric acid crystals to obtain crystal fluorescence images, and to preprocess the crystal fluorescence images to obtain target fluorescence images; The segmentation module 630 is used to segment the target fluorescence image based on an image segmentation model to obtain multiple single crystal images; The module 640 is used to input the single crystal image of the same subject into the deep learning model corresponding to the target clinical task based on the target clinical task, so as to obtain the assessment results related to liver fibrosis.

[0069] The in vitro auxiliary assessment system for liver fibrosis according to embodiments of the present invention acquires uric acid crystals generated based on plasma induction; images of the uric acid crystals are acquired to obtain crystal fluorescence images, and the crystal fluorescence images are preprocessed to obtain target fluorescence images; the target fluorescence images are segmented based on an image segmentation model to obtain multiple single crystal images; based on a target clinical task, the single crystal images of the same subject are input into a deep learning model corresponding to the target clinical task to obtain assessment results related to liver fibrosis. This achieves automated in vitro auxiliary assessment of liver fibrosis, improves the accuracy and efficiency of in vitro auxiliary assessment of liver fibrosis, and avoids the trauma and risks of puncture biopsy.

[0070] Optionally, the acquisition module 610 is specifically used to acquire the plasma of the subject; prepare a supersaturated sodium urate solution based on uric acid powder and sodium hydroxide solution; construct a reaction system based on sodium acetate, the plasma sample, and the supersaturated sodium urate solution; and induce the formation of uric acid crystals based on the reaction system, reaction plate, incubation time, and temperature.

[0071] Optionally, the image segmentation model is trained as follows: A dataset is acquired and divided into a training set, a validation set, and a test set, wherein the dataset includes manually labeled crystal fluorescence image samples; a deep learning-based crystal instance segmentation model is constructed, which is used to identify and segment individual crystal instances in crystal fluorescence images and outputs instance segmentation results corresponding to the input image, wherein the instance segmentation results include multiple crystal instance regions and their corresponding instance numbers; training parameters are set and a loss function and evaluation metrics are defined; the training set is batch-input into the crystal instance segmentation model, forward propagation is used to obtain a prediction mask, the loss value is calculated, and backward propagation is performed. The network parameters are updated. After each training epoch, the Dice coefficient and Intersection over Union (IoU) are calculated on the validation set. When the performance of the validation set no longer improves after several consecutive epochs, training stops, and the network weights with the best overall performance on the validation set are saved as the trained image segmentation model. The test set is input into the trained image segmentation model, and a prediction mask is output. If the Dice coefficient and / or IoU of the validation set reach a preset performance threshold, or if the improvement of the Dice coefficient and / or IoU of the validation set in K consecutive training epochs is less than a preset convergence threshold, then the trained image segmentation model is complete.

[0072] Optionally, the segmentation module 630 is specifically used to input the target fluorescence image into the image segmentation model, and the image segmentation model outputs multiple single crystal images.

[0073] Optionally, the target clinical task includes assisting in distinguishing between health and liver fibrosis, assisting in screening for liver fibrosis risk, or assisting in identifying early liver fibrosis.

[0074] Optionally, when the target clinical task is to assist in distinguishing between healthy individuals and liver fibrosis, the deep learning model corresponding to the target clinical task is trained in the following manner: A single-crystal image dataset for model training is obtained, the single-crystal image dataset including multiple single-crystal image samples from multiple subjects, and the single-crystal image samples are standardized; the standardized single-crystal image samples are labeled, and the standardized single-crystal image samples are divided into a training set, a validation set, and a test set using a hierarchical grouping cross-validation method, wherein the label is either healthy or liver fibrosis; a single single-crystal image from the training set is input into a deep learning model constructed based on a ConvNeXt deep convolutional neural network, wherein the backbone network of the deep learning model constructed based on the ConvNeXt deep convolutional neural network is ConvNeXt_ The base network, a binary fully connected layer, outputs the predicted probability of health or liver fibrosis. The predicted probability is compared with the true label of the single-crystal image, and a cross-entropy loss function is used to determine the loss value between the predicted probability and the true label. Based on the loss value, the gradient of the trainable parameters in the backbone network ConvNeXt_base is calculated, and the parameters of the backbone network and the classification head are updated until the training iterations reach a preset number. After each training epoch, image-level and / or subject-level classification performance metrics are calculated on the validation set. These performance metrics include one or more of accuracy, AUC, F1 score, precision, recall, and validation loss. The optimal training epoch is selected based on the subject-level AUC, F1 score, or overall classification performance on the validation set, and the corresponding model weights are saved as the trained deep learning model. Optionally, module 640 is specifically used to input the single-crystal image of the same subject into the deep learning model corresponding to the target clinical task, and output the probability of health or liver fibrosis; obtain the probability of health or liver fibrosis corresponding to multiple single-crystal images of the same subject; and determine the evaluation result based on the probability averaging algorithm and the probability of health or liver fibrosis corresponding to multiple single-crystal images.

[0075] Optionally, when the target clinical task is to assist in screening for liver fibrosis risk, the deep learning model corresponding to the target clinical task includes an image depth feature extraction branch and a crystal morphology feature branch. The image depth feature extraction branch is used to automatically extract depth features from single-crystal images, and the morphology feature branch is used to extract or input the morphology, grayscale, texture, and spatial distribution features of the single crystal. The deep learning model corresponding to the target clinical task is trained as follows: a binary classification dataset is obtained for model training, comprising multiple single-crystal image samples from multiple subjects, and the single-crystal image samples are standardized; the standardized single-crystal image samples are labeled, and the standardized binary classification dataset is divided into a training set, a validation set, and a test set using a hierarchical grouping cross-validation method, wherein the label indicates no fibrosis or the presence of different degrees of fibrosis; a single single-crystal image from the training set is input into the image depth feature extraction branch, and after feature extraction and pooling, an N-dimensional depth image feature vector is output, wherein the pooling... The image processing is performed using Global Average Pooling (GAP). The deep feature branch is constructed based on a ConvNeXt deep convolutional neural network, with the backbone network being ConvNeXt_base. A single single-crystal image from the training set is input to the crystal morphology feature branch, outputting an M-dimensional crystal morphology feature vector. The N-dimensional deep feature vector and the M-dimensional crystal morphology feature vector are fused to obtain a fused feature, which is then input to the classification layer, outputting a predicted positive category value corresponding to the target clinical task. This predicted value is then converted to a predicted positive category probability using a Sigmoid function. Based on the predicted positive category value and the true label, a binary cross-entropy loss function is used to calculate the loss value. According to the loss value, the gradient of the trainable parameters in the backbone network ConvNeXt_base is calculated, and the trainable parameters in the image deep feature extraction branch, the crystal morphology feature branch, the fusion layer, and the classification layer are updated until the training iterations reach a preset number. The classification accuracy is calculated on the validation set. If the classification accuracy is greater than a preset threshold, the deep learning model training is complete.

[0076] Optionally, module 640 is specifically used to input the single-crystal image of the same subject into the deep learning model corresponding to the target clinical task, and output the positive category prediction probability corresponding to the single-crystal image; average the positive category prediction probabilities corresponding to multiple single-crystal images to obtain the subject-level positive prediction probability; and determine the evaluation result based on the subject-level positive prediction probability and a preset threshold.

[0077] According to one aspect of the present invention, an electronic device is provided.

[0078] Figure 7 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Figure 7 As shown, an electronic device may include one or more ( Figure 7 Only one is shown in the image. A processor 102 (which may include, but is not limited to, a microprocessor unit (MPU) or a programmable logic device (PLD)) and a memory 104 for storing data are also shown. In one exemplary embodiment, the electronic device may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 7 The structure shown is for illustrative purposes only and does not limit the structure of the terminal device described above. For example, the terminal device may also include components that are more... Figure 7 The more or fewer components shown, or having the same Figure 7 Equivalent functions or ratios shown Figure 7 The functions shown have more different configurations.

[0079] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the in vitro auxiliary assessment method for liver fibrosis in this embodiment of the invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to terminal devices via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0080] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the switching device. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module used for wireless communication with the Internet.

[0081] This invention proposes a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform an in vitro auxiliary assessment method for liver fibrosis.

[0082] The applicant of this invention has provided a detailed description of the embodiments of the invention in conjunction with the accompanying drawings. However, those skilled in the art should understand that the above embodiments are merely preferred embodiments of the invention. The detailed description is only intended to help readers better understand the spirit of the invention and is not intended to limit the scope of protection of the invention. On the contrary, any improvements or modifications made based on the inventive spirit of the invention should fall within the scope of protection of the invention.

[0083] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0084] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the protection scope of the present invention.

Claims

1. An in vitro auxiliary assessment method for liver fibrosis, characterized in that, include: Obtain uric acid crystals based on plasma-induced formation; The uric acid crystals are imaged to obtain a crystal fluorescence image, and the crystal fluorescence image is preprocessed to obtain a target fluorescence image; The target fluorescence image is segmented based on an image segmentation model to obtain multiple single-crystal images; Based on the target clinical task, the single-crystal image of the same subject is input into the deep learning model corresponding to the target clinical task to obtain the assessment results related to liver fibrosis.

2. The in vitro auxiliary assessment method for liver fibrosis according to claim 1, characterized in that, The step of inducing the formation of uric acid crystals based on the plasma includes: Obtain the plasma from the test subject; A supersaturated sodium urate solution was prepared based on uric acid powder and sodium hydroxide solution. A reaction system was constructed based on sodium acetate, the plasma sample, and the supersaturated sodium urate solution; Based on the aforementioned reaction system, reaction plate, incubation time, and temperature, the uric acid crystals are induced to form.

3. The in vitro auxiliary assessment method for liver fibrosis according to claim 1, characterized in that, The image segmentation model is trained in the following way: Obtain a dataset and divide the dataset into a training set, a validation set, and a test set, wherein the dataset includes manually labeled crystal fluorescence image samples; A deep learning-based crystal instance segmentation model is constructed. The crystal instance segmentation model is used to identify and segment a single crystal instance in a crystal fluorescence image and output the instance segmentation result corresponding to the input image. The instance segmentation result includes multiple crystal instance regions and their corresponding instance numbers. Set training parameters and define loss function and evaluation metrics; The training set is batch-input into the crystal instance segmentation model. Forward propagation is used to obtain the prediction mask, the loss value is calculated, and backpropagation is used to update the network parameters. After each training round, the Dice coefficient and intersection-union ratio (IoU) are calculated on the validation set. When the performance of the validation set no longer improves for several consecutive rounds, training is stopped, and the network weights with the best overall performance on the validation set are saved as the trained image segmentation model. The test set is input into the trained image segmentation model, which outputs a predicted mask. If the Dice coefficients and / or the Intersection over Union (IoU) of the validation set reach a preset performance threshold, or if the improvement of the Dice coefficients and / or IoU in the validation set is less than a preset convergence threshold within K consecutive training rounds, then the training of the image segmentation model is complete. The target fluorescence image is segmented based on an image segmentation model to obtain multiple single-crystal images, including: The target fluorescence image is input into the image segmentation model, and the image segmentation model outputs multiple single crystal images.

4. The in vitro auxiliary assessment method for liver fibrosis according to claim 1, characterized in that, The target clinical tasks include assisting in the differentiation between healthy individuals and liver fibrosis, assisting in the screening of liver fibrosis risk, or assisting in the identification of early liver fibrosis.

5. The in vitro auxiliary assessment method for liver fibrosis according to claim 4, characterized in that, When the target clinical task is to assist in distinguishing between healthy individuals and liver fibrosis, the deep learning model corresponding to the target clinical task is trained in the following manner: A single-crystal image dataset for model training is obtained, the single-crystal image dataset comprising multiple single-crystal image samples from multiple subjects, and the single-crystal image samples are standardized. The standardized single-crystal image samples are labeled, and the standardized single-crystal image samples are divided into training set, validation set and test set by hierarchical grouping cross-validation, wherein the label is healthy or liver fibrosis. The single crystal image in the training set is input into a deep learning model based on ConvNeXt deep convolutional neural network, wherein the backbone network of the deep learning model based on ConvNeXt deep convolutional neural network is ConvNeXt_base, the classification head is a binary classification fully connected layer, and the predicted probability of health or liver fibrosis is output. The predicted probability is compared with the true label of the single crystal image, and the loss value between the predicted probability and the true label is determined by using the cross-entropy loss function. Based on the loss value, calculate the gradient of the trainable parameters in the backbone network ConvNeXt_base network, and update the parameters of the backbone network and the classification head until the number of training iterations reaches the preset number; After each training round, image-level and / or subject-level classification performance metrics are calculated on the validation set. The classification performance metrics include one or more of accuracy, AUC, F1 score, precision, recall, and validation loss. The optimal training rounds are selected based on the AUC, F1 score, or overall classification performance of the validation set subjects, and the corresponding model weights are saved as the trained deep learning model. Specifically, the single-crystal image of the same subject is input into the deep learning model corresponding to the target clinical task to obtain assessment results related to liver fibrosis, including: The single-crystal image of the same subject is input into the deep learning model corresponding to the target clinical task, and the probability of being healthy or having liver fibrosis is output. Obtain the probability of health or liver fibrosis corresponding to multiple single-crystal images of the same subject; The evaluation result is determined based on a probability averaging algorithm and the probability of health or liver fibrosis corresponding to multiple single-crystal images.

6. The in vitro auxiliary assessment method for liver fibrosis according to claim 4, characterized in that, When the target clinical task is to assist in screening for liver fibrosis risk, the deep learning model corresponding to the target clinical task includes an image depth feature extraction branch and a crystal morphology feature branch. The image depth feature extraction branch is used to automatically extract the depth features of a single crystal image, and the morphology feature branch is used to extract or input the morphology, grayscale, texture, and spatial distribution features of the single crystal. The deep learning model corresponding to the target clinical task is trained in the following way: A binary classification dataset for model training is obtained, the binary classification dataset comprising multiple single-crystal image samples from multiple subjects, and the single-crystal image samples are standardized. The standardized single-crystal image samples are labeled, and the standardized binary classification dataset is divided into training set, validation set and test set using a hierarchical grouping cross-validation method. The labels are either non-fiber or have different degrees of fiber. A single single-crystal image from the training set is input into the image depth feature extraction branch. After feature extraction and pooling, an N-dimensional depth image feature vector is output. The pooling is global average pooling (GAP). The depth feature branch is constructed based on a ConvNeXt deep convolutional neural network, and the backbone network is ConvNeXt_base. A single crystal image from the training set is input into the crystal morphology feature branch, and an M-dimensional crystal morphology feature vector is output. The N-dimensional depth feature vector and the M-dimensional crystal morphology feature vector are fused to obtain a fused feature, which is then input into the classification layer to output the positive category prediction value corresponding to the target clinical task, and converted into the positive category prediction probability by the Sigmoid function. Based on the predicted positive category value and the true label, the loss value is calculated using the binary cross-entropy loss function; Based on the loss value, the gradient of the trainable parameters in the backbone network ConvNeXt_base is calculated, and the trainable parameters in the image depth feature extraction branch, the crystal morphology feature branch, the fusion layer and the classification layer are updated until the training iterations reach the preset number of times. The classification accuracy is calculated on the validation set, and the deep learning model is trained when the classification accuracy is greater than a preset threshold.

7. The in vitro auxiliary assessment method for liver fibrosis according to claim 6, characterized in that, The single-crystal image of the same subject is input into the deep learning model corresponding to the target clinical task to obtain assessment results related to liver fibrosis, including: The single crystal image of the same subject is input into the deep learning model corresponding to the target clinical task, and the positive category prediction probability corresponding to the single crystal image is output. The positive category prediction probability corresponding to multiple single crystal images is averaged to obtain the positive prediction probability at the subject level. The evaluation result is determined based on the positive prediction probability of the test subject and the preset threshold.

8. An in vitro auxiliary assessment system for liver fibrosis, characterized in that, include: The acquisition module is used to acquire uric acid crystals generated based on plasma induction. The acquisition and preprocessing module is used to acquire images of the uric acid crystals to obtain crystal fluorescence images, and to preprocess the crystal fluorescence images to obtain target fluorescence images; The segmentation module is used to segment the target fluorescence image based on an image segmentation model to obtain multiple single-crystal images; The module is used to input the single-crystal image of the same subject into the deep learning model corresponding to the target clinical task, based on the target clinical task, to obtain the assessment results related to liver fibrosis.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 7.

Images