Intelligent analysis system and method for endocrine abnormality detection
Through the intelligent analysis system, the contrast enhancement and semantic feature extraction of the mass spectrum of the body fluid sample is solved, and the problem of difficulty in detecting low-abundance hormones in traditional endocrine detection methods is solved, achieving more accurate judgment of endocrine status and early change discovery, and improving endocrine detection Level of intelligence.
Patent Information
- Application Number
- CN202411231181.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-04
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-09-04
AI Technical Summary
Traditional endocrine detection methods are difficult to detect hormones with low abundance or special chemical properties, and their sensitivity is not enough to capture early or trace endocrine changes, affecting early diagnosis of the disease.
The mass spectra of the body fluid sample collected based on the liquid chromatography-mass spectrometer is used to perform contrast enhancement, semantic feature extraction and convolutional gating mechanism processing through an intelligent analysis system to generate key semantic significance feature maps for intelligent judgment of endocrine status .
It can detect lower abundance hormones, provide more accurate quantitative analysis, discover trace endocrine changes, and process more types of biomarkers, improving the intelligence level of endocrine detection.
Smart Images

Figure CN118731262B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of endocrine analysis, and more specifically, to an intelligent analysis system and method for endocrine abnormality detection. Background Art
[0002] The endocrine system is the body's hormone regulation system, responsible for regulating various physiological processes, including metabolism, growth, development, reproduction, and emotions. Endocrine abnormalities refer to dysfunction of the endocrine system, which leads to excessive or insufficient hormone secretion, thus causing a series of health problems. Therefore, the detection of endocrine abnormalities is crucial for the timely detection and treatment of endocrine diseases.
[0003] The detection of endocrine abnormalities mainly relies on the measurement of hormone levels in blood or urine samples. However, although traditional endocrine detection methods such as chemiluminescence and ELISA cover a wide range of hormone detection, they are limited when facing some special or emerging biomarkers. These traditional methods may not be able to detect hormones with low abundance or special chemical properties. In addition, traditional methods may lack sufficient sensitivity to detect early or trace endocrine changes, which are crucial for the early diagnosis of certain diseases.
[0004] Therefore, an intelligent analysis solution for endocrine abnormality detection is needed. Summary of the invention
[0005] In view of the shortcomings of the prior art, the present application provides an intelligent analysis system and method for endocrine abnormality detection.
[0006] According to one aspect of the present application, there is provided an intelligent analysis system for endocrine abnormality detection, comprising:
[0007] A patient subject body fluid sample mass spectrum acquisition module, used to acquire a mass spectrum of a body fluid sample of a target patient subject collected by a liquid chromatography-mass spectrometer;
[0008] A body fluid sample mass spectrum contrast enhancement module, used for performing contrast enhancement processing on the body fluid sample mass spectrum to obtain a contrast enhanced body fluid sample mass spectrum;
[0009] A body fluid sample mass spectrum semantic feature extraction module, used for passing the contrast enhanced body fluid sample mass spectrum through a body fluid sample mass spectrum feature extractor based on a pyramid structure to obtain a body fluid sample mass spectrum semantic feature map;
[0010] A body fluid sample mass spectrum key semantic saliency module, used for passing the body fluid sample mass spectrum semantic feature map through a body fluid sample key feature foreground saliency enhancer based on a convolution gating mechanism to obtain a body fluid sample mass spectrum key semantic saliency expression feature map as a body fluid sample mass spectrum key semantic saliency expression feature;
[0011] An endocrine state judgment module, used to obtain an endocrine state detection result based on the key semantically significant expression features of the mass spectrum of the body fluid sample, wherein the endocrine state detection result is used to indicate whether the endocrine state of the detected target patient object is abnormal;
[0012] The body fluid sample mass spectrum key semantic saliency module includes:
[0013] A body fluid sample mass spectrum semantic feature map layer normalization unit, used for performing layer normalization processing on the body fluid sample mass spectrum semantic feature map to obtain a normalized body fluid sample mass spectrum semantic feature map;
[0014] A normalized body fluid sample mass spectrum semantic feature map deep convolution encoding unit, used for performing point convolution-based channel expansion and deep convolution encoding based on a hole convolution layer on the normalized body fluid sample mass spectrum semantic feature map to obtain a body fluid sample mass spectrum semantic deep convolution original feature map and a body fluid sample mass spectrum semantic deep convolution backup feature map;
[0015] The body fluid sample mass spectrum semantic feature gated mask calculation unit is used to input the body fluid sample mass spectrum semantic deep convolution original feature map into the The foreground gated mask module of the function is used to obtain the semantic deep convolution gated mask weight feature map of the mass spectrum of the body fluid sample;
[0016] A body fluid sample mass spectrum semantic gated mask foreground highlighting feature calculation unit, used to calculate the position point multiplication between the body fluid sample mass spectrum semantic deep convolution gated mask weight feature map and the body fluid sample mass spectrum semantic deep convolution backup feature map to obtain a body fluid sample mass spectrum semantic gated mask foreground highlighting feature map;
[0017] The body fluid sample mass spectrum key semantic salient expression feature map generation unit is used to perform point convolution-based channel contraction on the body fluid sample mass spectrum semantic gated mask foreground highlighting feature map to obtain the body fluid sample mass spectrum key semantic salient expression feature map.
[0018] In the above-mentioned intelligent analysis system for endocrine abnormality detection, the body fluid sample mass spectrum contrast enhancement module is used to: perform piecewise linear transformation processing on the body fluid sample mass spectrum to obtain the contrast enhanced body fluid sample mass spectrum.
[0019] In the above-mentioned intelligent analysis system for endocrine abnormality detection, the body fluid sample mass spectrum semantic feature extraction module includes: a contrast-enhanced body fluid sample mass spectrum shallow feature extraction unit, which is used to pass the contrast-enhanced body fluid sample mass spectrum through a convolution layer with a convolution kernel of 3×3 to obtain a contrast-enhanced body fluid sample mass spectrum shallow feature map; a contrast-enhanced body fluid sample mass spectrum shallow fusion unit, which is used to pass the contrast-enhanced body fluid sample mass spectrum shallow feature map through a body fluid sample mass spectrum shallow residual information fusion enhancement module to obtain a contrast-enhanced body fluid sample mass spectrum shallow residual fusion feature map; a contrast-enhanced body fluid sample mass spectrum deep feature extraction unit, which is used to pass the contrast-enhanced body fluid sample mass spectrum shallow residual fusion feature map through a convolution layer with a convolution kernel of 3×3 to obtain a contrast-enhanced body fluid sample mass spectrum deep feature map; a contrast-enhanced body fluid sample mass spectrum deep fusion unit, which is used to pass the contrast-enhanced body fluid sample mass spectrum deep feature map through a body fluid sample mass spectrum deep residual information fusion enhancement module to obtain a contrast-enhanced body fluid sample mass spectrum deep feature map. A mass spectrum deep residual fusion feature map; a contrast-enhanced body fluid sample mass spectrum feature modulation unit, used to input the contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map into a scale modulation module based on the SPPF layer to obtain a contrast-enhanced body fluid sample mass spectrum deep modulation feature map; a contrast-enhanced body fluid sample mass spectrum deep modulation feature upsampling unit, used to upsample the contrast-enhanced body fluid sample mass spectrum deep modulation feature map to obtain an upsampled contrast-enhanced body fluid sample mass spectrum deep modulation feature map; a contrast-enhanced body fluid sample mass spectrum multi-scale feature generation unit, used to fuse the upsampled contrast-enhanced body fluid sample mass spectrum deep modulation feature map with the contrast-enhanced body fluid sample mass spectrum shallow residual fusion feature map to obtain a contrast-enhanced body fluid sample mass spectrum multi-scale feature map; a body fluid sample mass spectrum semantic feature generation unit, used to pass the contrast-enhanced body fluid sample mass spectrum multi-scale feature map through a contrast-enhanced body fluid sample mass spectrum multi-scale residual information fusion enhancement module to obtain the body fluid sample mass spectrum semantic feature map.
[0020] In the above-mentioned intelligent analysis system for endocrine abnormality detection, the contrast-enhanced body fluid sample mass spectrum shallow fusion unit is used to: perform channel dimensionality reduction modulation on the contrast-enhanced body fluid sample mass spectrum shallow feature map through a point convolution layer to obtain a channel dimensionality reduction modulated contrast-enhanced body fluid sample mass spectrum shallow feature map; perform feature extraction on the channel dimensionality reduction modulated contrast-enhanced body fluid sample mass spectrum shallow feature map through a convolution layer with a convolution kernel of 3×3 to obtain a contrast-enhanced body fluid sample mass spectrum shallow implicit information feature map; perform channel dimensionality increase modulation on the contrast-enhanced body fluid sample mass spectrum shallow implicit information feature map through a point convolution layer to obtain a contrast-enhanced body fluid sample mass spectrum shallow feature map after channel modulation; fuse the contrast-enhanced body fluid sample mass spectrum shallow feature map after channel modulation and the channel dimensionality reduction modulated contrast-enhanced body fluid sample mass spectrum shallow feature map to obtain the contrast-enhanced body fluid sample mass spectrum shallow residual fusion feature map.
[0021] In the above-mentioned intelligent analysis system for endocrine abnormality detection, the contrast-enhanced body fluid sample mass spectrum feature modulation unit is used to: pass the contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map through a scale modulation module based on a point convolution layer to obtain a scale-adjusted contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map; perform maximum pooling processing on each feature matrix along the channel dimension in the scale-adjusted contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map to obtain a contrast-enhanced body fluid sample mass spectrum deep residual fusion feature vector; based on the contrast-enhanced body fluid sample mass spectrum deep residual fusion feature vector, perform channel-weighted enhancement on the scale-adjusted contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map to obtain the contrast-enhanced body fluid sample mass spectrum deep modulation feature map.
[0022] In the above-mentioned intelligent analysis system for endocrine abnormality detection, the body fluid sample mass spectrum semantic feature gated mask calculation unit is used to: set the feature values greater than the predetermined hyperparameter in each position of the body fluid sample mass spectrum semantic deep convolution original feature map to one, and set the rest to zero to obtain the body fluid sample mass spectrum semantic deep convolution gated mask weight feature map.
[0023] In the above-mentioned intelligent analysis system for endocrine abnormality detection, the endocrine status judgment module is used to: pass the key semantic significant expression feature map of the mass spectrum of the body fluid sample through the classifier-based endocrine abnormality detector to obtain the endocrine status detection result, and the endocrine status detection result is used to indicate whether the endocrine status of the target patient being detected is abnormal.
[0024] According to another aspect of the present application, there is provided an intelligent analysis method for endocrine abnormality detection, comprising:
[0025] Obtaining a mass spectrum of a body fluid sample of a target patient subject collected by a liquid chromatography-mass spectrometer;
[0026] performing contrast enhancement processing on the body fluid sample mass spectrum to obtain a contrast enhanced body fluid sample mass spectrum;
[0027] The contrast-enhanced body fluid sample mass spectrum is passed through a body fluid sample mass spectrum feature extractor based on a pyramid structure to obtain a body fluid sample mass spectrum semantic feature map;
[0028] The body fluid sample mass spectrum semantic feature map is passed through a body fluid sample key feature foreground saliency enhancer based on a convolution gating mechanism to obtain a body fluid sample mass spectrum key semantic saliency expression feature map as a body fluid sample mass spectrum key semantic saliency expression feature;
[0029] Based on the key semantically significant expression features of the mass spectrum of the body fluid sample, an endocrine status detection result is obtained, wherein the endocrine status detection result is used to indicate whether the endocrine status of the target patient being detected is abnormal;
[0030] The method of obtaining a body fluid sample mass spectrum key semantic salient expression feature map as a body fluid sample mass spectrum key semantic salient expression feature by passing the body fluid sample mass spectrum key feature foreground salient enhancer based on a convolution gating mechanism comprises:
[0031] Performing layer normalization processing on the body fluid sample mass spectrum semantic feature map to obtain a normalized body fluid sample mass spectrum semantic feature map;
[0032] Performing point convolution-based channel expansion and deep convolution coding based on a hole convolution layer on the normalized body fluid sample mass spectrum semantic feature map to obtain a body fluid sample mass spectrum semantic deep convolution original feature map and a body fluid sample mass spectrum semantic deep convolution backup feature map;
[0033] The body fluid sample mass spectrum semantic deep convolution original feature map is input based on The foreground gated mask module of the function is used to obtain the semantic deep convolution gated mask weight feature map of the mass spectrum of the body fluid sample;
[0034] Calculate the position point multiplication between the body fluid sample mass spectrum semantic deep convolution gated mask weight feature map and the body fluid sample mass spectrum semantic deep convolution backup feature map to obtain the body fluid sample mass spectrum semantic gated mask foreground highlight feature map;
[0035] The point convolution-based channel shrinkage is performed on the semantic gated mask foreground highlighting feature map of the body fluid sample mass spectrum to obtain the key semantic saliency expression feature map of the body fluid sample mass spectrum.
[0036] In the above-mentioned intelligent analysis method for endocrine abnormality detection, the body fluid sample mass spectrum is subjected to contrast enhancement processing to obtain a contrast enhanced body fluid sample mass spectrum, including: performing piecewise linear transformation processing on the body fluid sample mass spectrum to obtain the contrast enhanced body fluid sample mass spectrum.
[0037] This application has significant technical effects due to the adoption of the above technical solutions:
[0038] The intelligent analysis system and method for endocrine abnormality detection provided by the present application first obtains the mass spectrogram of the body fluid sample of the target patient to be detected, and then uses the image recognition and processing technology based on machine learning to perform semantic analysis and feature enhancement processing of the mass spectrogram of the body fluid sample, so as to intelligently judge whether the endocrine state of the target patient to be detected is abnormal based on the key foreground significant feature information of the mass spectrogram of the body fluid sample. In this way, it is possible to detect hormones with lower abundance, thereby providing more accurate quantitative analysis, which is helpful to discover trace endocrine changes in the body of the target patient to be detected, and it is also possible to process and identify more types of biomarkers, thereby improving the intelligent level of endocrine detection and analysis. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:
[0040] Figure 1 4 is a system block diagram of an intelligent analysis system for endocrine abnormality detection according to an embodiment of the present application.
[0041] Figure 2 Schematic diagram of the architecture of an intelligent analysis system for endocrine abnormality detection according to an embodiment of the present application.
[0042] Figure 3 It is a block diagram of a body fluid sample mass spectrometry semantic feature extraction module in an intelligent analysis system for endocrine abnormality detection according to an embodiment of the present application.
[0043] Figure 4 It is a block diagram of a key semantic saliency module of a body fluid sample mass spectrum in an intelligent analysis system for endocrine abnormality detection according to an embodiment of the present application.
[0044] Figure 5 The figure is a flow chart of an intelligent analysis method for endocrine abnormality detection according to an embodiment of the present application. DETAILED DESCRIPTION
[0045] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.
[0046] The endocrine system is a fine regulation system in the human body. It controls and coordinates a variety of key physiological functions in the body, such as metabolism, growth and development, reproductive health, and emotional regulation, by secreting hormones. Although the concentration of these hormones in the blood is extremely low, their role is extremely important and extensive. Endocrine abnormalities refer to the dysfunction of this system, which may lead to abnormal increase or decrease in hormone levels, and then cause a variety of health problems, including but not limited to diabetes, thyroid disease, polycystic ovary syndrome, etc.
[0047] To ensure the balance and health of the endocrine system, timely and accurate detection of endocrine abnormalities is crucial. Currently, endocrine abnormalities are mainly detected by measuring hormone levels in blood or urine samples. These tests can provide us with important information about an individual's endocrine status and help doctors make accurate diagnosis and treatment decisions.
[0048] However, traditional endocrine detection methods, such as chemiluminescent immunoassay (CLIA) and enzyme-linked immunosorbent assay (ELISA), although they play an important role in a wide range of hormone detection, may be inadequate when faced with some special or emerging biomarkers. These methods may have limitations in detecting hormones of low abundance or with special chemical properties, and cannot capture those subtle but critical changes. In addition, traditional detection methods also have certain limitations in sensitivity. They may not provide sufficient detection accuracy in the early stages of the disease or when there are slight changes in hormone levels. These early or trace endocrine changes are of irreplaceable importance for the early diagnosis and intervention of certain diseases.
[0049] Therefore, in response to the above technical problems, the technical concept of the present application is to collect a mass spectrogram of a body fluid sample of a target patient object to be detected by a liquid chromatography-mass spectrometer, and use image recognition and processing technology based on machine learning to perform semantic analysis and feature enhancement processing of the body fluid sample mass spectrogram, so as to intelligently judge whether the endocrine state of the target patient object to be detected is abnormal based on the key prospect significant feature information of the body fluid sample mass spectrogram. In this way, hormones with lower abundance can be detected, so as to provide more accurate quantitative analysis, which is helpful to discover early or trace endocrine changes. And the intelligent analysis system for endocrine abnormality detection can process and identify more types of biomarkers, including special or emerging biomarkers that are difficult to detect by traditional methods, thereby improving the intelligence level of endocrine detection and analysis.
[0050] Figure 1 4 is a system block diagram of an intelligent analysis system for endocrine abnormality detection according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of an intelligent analysis system for endocrine abnormality detection according to an embodiment of the present application. Figure 1 and Figure 2 As shown, in the intelligent analysis system 100 for endocrine abnormality detection, it includes: a patient object body fluid sample mass spectrum acquisition module 110, which is used to acquire a body fluid sample mass spectrum of a target patient object to be detected collected by a liquid chromatography-mass spectrometer; a body fluid sample mass spectrum contrast enhancement module 120, which is used to perform contrast enhancement processing on the body fluid sample mass spectrum to obtain a contrast enhanced body fluid sample mass spectrum; a body fluid sample mass spectrum semantic feature extraction module 130, which is used to extract the body fluid sample mass spectrum semantic feature from the contrast enhanced body fluid sample through a body fluid sample mass spectrum feature extractor based on a pyramid structure to obtain a body fluid sample mass spectrum. A body fluid sample mass spectrum semantic feature map; a body fluid sample mass spectrum key semantic saliency module 140, used to pass the body fluid sample mass spectrum semantic feature map through a body fluid sample key feature foreground saliency enhancer based on a convolution gating mechanism to obtain a body fluid sample mass spectrum key semantic saliency expression feature map as a body fluid sample mass spectrum key semantic saliency expression feature; an endocrine state judgment module 150, used to obtain an endocrine state detection result based on the body fluid sample mass spectrum key semantic saliency expression feature, and the endocrine state detection result is used to indicate whether the endocrine state of the target patient being detected is abnormal.
[0051] In an embodiment of the present application, the patient object body fluid sample mass spectrometer acquisition module 110 is used to obtain a body fluid sample mass spectrometer of a detected target patient object collected by a liquid chromatography-mass spectrometer. It should be understood that the body fluid sample mass spectrometer is a graphical representation obtained after analyzing a biological sample (such as blood, urine, etc.) by a liquid chromatography-mass spectrometer, which displays the mass to charge ratio of different compounds in the sample and their relative abundance information. Specifically, the height and area of each peak in the body fluid sample mass spectrometer can directly reflect the concentration of hormones in the sample. By analyzing the concentration representation of hormones in the mass spectrometer, the pattern of endocrine abnormalities can be identified. That is, excessive or insufficient secretion of certain hormones will lead to specific mass spectrometric features, and machine learning algorithms can train models to identify these features, which are then used to identify whether there is an abnormal endocrine state.
[0052] In an embodiment of the present application, the body fluid sample mass spectrum contrast enhancement module 120 is used to perform contrast enhancement processing on the body fluid sample mass spectrum to obtain a contrast enhanced body fluid sample mass spectrum. Specifically, in an embodiment of the present application, the body fluid sample mass spectrum contrast enhancement module is used to: perform piecewise linear transformation processing on the body fluid sample mass spectrum to obtain the contrast enhanced body fluid sample mass spectrum. It should be understood that the signals in the body fluid sample mass spectrum may not differ much in grayscale or intensity, so that the grayscale distribution is concentrated, resulting in some details being difficult to identify, such as some proteins or peptides may be difficult to detect in conventional mass spectra due to their low abundance. Therefore, in the technical solution of the present application, the body fluid sample mass spectrum is subjected to piecewise linear transformation processing to obtain a contrast enhanced body fluid sample mass spectrum. Specifically, firstly, by analyzing the grayscale distribution of the body fluid sample mass spectrum, it is understood which grayscale value pixels are more concentrated, as well as the contrast distribution of the image, to obtain a grayscale histogram. Next, based on the grayscale histogram, the grayscale intervals of the body fluid sample that need to enhance the contrast are identified, which are usually areas where pixels are concentrated. Then, a piecewise linear transformation function is designed, which performs nonlinear mapping of grayscale values within the selected grayscale interval, and maps the original grayscale values in the mass spectrum of the body fluid sample to a wider grayscale range, thereby improving the overall contrast and readability of the mass spectrum of the body fluid sample.
[0053] In an embodiment of the present application, the body fluid sample mass spectrum semantic feature extraction module 130 is used to obtain a body fluid sample mass spectrum semantic feature map by passing the contrast-enhanced body fluid sample mass spectrum through a body fluid sample mass spectrum feature extractor based on a pyramid structure. Accordingly, considering that the contrast-enhanced body fluid sample mass spectrum contains shallow feature information such as grayscale, edge, contrast, etc. of the body fluid sample, it also contains semantic deep information identified in the body fluid, such as protein expression patterns, correlation states between different biological molecules, etc., and these shallow features can affect the expressiveness of deep features. The pyramid network can capture and extract deep and shallow feature information of different scales in the image, but the shallow features have a weak ability to influence the information between deep features, that is, the information flow from shallow to deep may be limited, resulting in a lack of sufficient position information for deep features, and a large amount of computational effort will be accumulated during the information processing process. Therefore, in order to strengthen the information exchange and influence between shallow features and deep features and reduce the amount of model parameters at the same time, in the technical solution of the present application, the contrast enhanced body fluid sample mass spectrum is passed through a pyramid-structured body fluid sample mass spectrum feature extractor to obtain a body fluid sample mass spectrum semantic feature map.
[0054] Specifically, the pyramid-structured body fluid sample mass spectrometry feature extractor consists of two parts: a feature extraction network and a feature fusion network. The feature fusion network uses residual fusion and jump connection mechanisms to optimize the flow of information, which can efficiently combine the primary (shallow) and advanced (deep) features in the contrast-enhanced body fluid sample mass spectrometry. In this way, the precise location information captured from the bottom layer of the image can be effectively uploaded to a higher level, making the deep features more accurate when integrating these location information, further improving the quality and accuracy of the overall feature expression. At the same time, the introduction of the pyramid structure allows residual extraction of features and information transmission at multiple levels, which helps to reduce the computational burden in the entire feature transmission process, avoids additional computational waste, thereby speeding up the network operation speed and improving the overall computing performance.
[0055] Specifically, Figure 3 FIG. 4 is a block diagram of a body fluid sample mass spectrometry semantic feature extraction module in an intelligent analysis system for endocrine abnormality detection according to an embodiment of the present application. Figure 3As shown, the body fluid sample mass spectrum semantic feature extraction module 130 includes: a contrast enhanced body fluid sample mass spectrum shallow feature extraction unit 131, which is used to pass the contrast enhanced body fluid sample mass spectrum through a convolution layer with a convolution kernel of 3×3 to obtain a contrast enhanced body fluid sample mass spectrum shallow feature map; a contrast enhanced body fluid sample mass spectrum shallow fusion unit 132, which is used to pass the contrast enhanced body fluid sample mass spectrum shallow feature map through a body fluid sample mass spectrum shallow residual information fusion enhancement module to obtain a contrast enhanced body fluid sample mass spectrum. The shallow residual fusion feature map of the mass spectrum of the contrast-enhanced body fluid sample is obtained by a deep feature extraction unit 133 for extracting the shallow residual fusion feature map of the mass spectrum of the contrast-enhanced body fluid sample through a convolution layer with a convolution kernel of 3×3 to obtain a deep feature map of the mass spectrum of the contrast-enhanced body fluid sample; the deep fusion unit 134 for extracting the deep residual fusion feature map of the mass spectrum of the contrast-enhanced body fluid sample is used to obtain a deep residual fusion feature map of the mass spectrum of the contrast-enhanced body fluid sample through a deep residual information fusion enhancement module of the mass spectrum of the body fluid sample. feature map; a contrast-enhanced body fluid sample mass spectrum feature modulation unit 135, used to input the contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map into a scale modulation module based on the SPPF layer to obtain a contrast-enhanced body fluid sample mass spectrum deep modulation feature map; a contrast-enhanced body fluid sample mass spectrum deep modulation feature upsampling unit 136, used to upsample the contrast-enhanced body fluid sample mass spectrum deep modulation feature map to obtain an upsampled contrast-enhanced body fluid sample mass spectrum deep modulation feature map; a contrast-enhanced body fluid sample mass spectrum multi-scale feature generation unit 137, used to fuse the upsampled contrast-enhanced body fluid sample mass spectrum deep modulation feature map with the contrast-enhanced body fluid sample mass spectrum shallow residual fusion feature map to obtain a contrast-enhanced body fluid sample mass spectrum multi-scale feature map; a body fluid sample mass spectrum semantic feature generation unit 138, used to pass the contrast-enhanced body fluid sample mass spectrum multi-scale feature map through the contrast-enhanced body fluid sample mass spectrum multi-scale residual information fusion enhancement module to obtain the body fluid sample mass spectrum semantic feature map.
[0056] More specifically, in an embodiment of the present application, the contrast-enhanced body fluid sample mass spectrum shallow fusion unit 132 is used to: perform channel dimensionality reduction modulation on the contrast-enhanced body fluid sample mass spectrum shallow feature map through a point convolution layer to obtain a channel dimensionality reduction modulated contrast-enhanced body fluid sample mass spectrum shallow feature map; perform feature extraction on the channel dimensionality reduction modulated contrast-enhanced body fluid sample mass spectrum shallow feature map through a convolution layer with a convolution kernel of 3×3 to obtain a contrast-enhanced body fluid sample mass spectrum shallow implicit information feature map; perform channel dimensionality increase modulation on the contrast-enhanced body fluid sample mass spectrum shallow implicit information feature map through a point convolution layer to obtain a contrast-enhanced body fluid sample mass spectrum shallow feature map after channel modulation; fuse the contrast-enhanced body fluid sample mass spectrum shallow feature map after channel modulation and the channel dimensionality reduction modulated contrast-enhanced body fluid sample mass spectrum shallow feature map to obtain the contrast-enhanced body fluid sample mass spectrum shallow residual fusion feature map.
[0057] More specifically, in an embodiment of the present application, the contrast-enhanced body fluid sample mass spectrum feature modulation unit is used to: pass the contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map through a scale modulation module based on a point convolution layer to obtain a scale-adjusted contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map; perform maximum pooling processing on each feature matrix along the channel dimension in the scale-adjusted contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map to obtain a contrast-enhanced body fluid sample mass spectrum deep residual fusion feature vector; perform channel-weighted enhancement on the scale-adjusted contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map based on the contrast-enhanced body fluid sample mass spectrum deep residual fusion feature vector to obtain the contrast-enhanced body fluid sample mass spectrum deep modulation feature map.
[0058] In an embodiment of the present application, the body fluid sample mass spectrum key semantic saliency module 140 is used to pass the body fluid sample mass spectrum semantic feature map through a body fluid sample key feature foreground saliency enhancer based on a convolution gating mechanism to obtain a body fluid sample mass spectrum key semantic saliency expression feature map as a body fluid sample mass spectrum key semantic saliency expression feature. Accordingly, considering that the body fluid sample mass spectrum semantic feature map contains key body fluid sample mass spectrum semantic features, such as the abundance of biomarkers and peptide sequence information, and these features are crucial to understanding the biological characteristics of body fluid samples, they can reveal the endocrine health status of individuals, the presence of disease markers, etc. Therefore, in order to focus on and significantly express the key sample mass spectrum semantic features in the body fluid sample mass spectrum semantic feature map, in the technical solution of the present application, the body fluid sample mass spectrum semantic feature map is passed through a body fluid sample key feature foreground saliency enhancer based on a convolution gating mechanism to obtain a body fluid sample mass spectrum key semantic saliency expression feature map.
[0059] In particular, the foreground saliency enhancer of key features of body fluid samples based on the convolutional gating mechanism captures the information of the mass spectrometry semantic features of body fluid samples in the spatial and depth dimensions through layer normalization and convolutional encoding, and uses a gated mask module to dynamically select foreground features, and then highlights these features through point-by-point multiplication operations, and finally adjusts the features through channel compression to enhance the model's ability to recognize important information. Specifically, layer normalization is performed on the mass spectrometry semantic feature map of body fluid samples to unify the feature scales of different layers and obtain a normalized feature map. Subsequently, these normalized feature maps are subjected to point convolution and dilated convolution processing to enhance the expressiveness of the feature map in the channel dimension, and allow for deeper feature extraction, generating a backup and original feature map of the deep convolution. Next, the convolutional gated feedforward mechanism is used to adaptively enhance the foreground salient key features in the mass spectrometry semantic feature map of body fluid samples to generate a feature map with gated mask weights. Then, these gated mask weights are applied to the original deep convolution feature map through point-by-point multiplication operations to highlight the significant areas in the semantic feature map of the mass spectrometry of the body fluid sample. Finally, these prominent features are further optimized by performing channel shrinkage processing through point convolution to obtain a significant expression feature map.
[0060] Specifically, Figure 4 FIG. 4 is a block diagram of a key semantic saliency module for body fluid sample mass spectra in an intelligent analysis system for endocrine abnormality detection according to an embodiment of the present application. Figure 4 As shown, the body fluid sample mass spectrum key semantic saliency module 140 includes: a body fluid sample mass spectrum semantic feature layer normalization unit 141, which is used to perform layer normalization processing on the body fluid sample mass spectrum semantic feature map to obtain a normalized body fluid sample mass spectrum semantic feature map; a normalized body fluid sample mass spectrum semantic feature map deep convolution encoding unit 142, which is used to perform point convolution-based channel expansion and deep convolution encoding based on a hole convolution layer on the normalized body fluid sample mass spectrum semantic feature map to obtain a body fluid sample mass spectrum semantic deep convolution original feature map and a body fluid sample mass spectrum semantic deep convolution backup feature map; a body fluid sample mass spectrum semantic feature gated mask calculation unit 143, which is used to input the body fluid sample mass spectrum semantic deep convolution original feature map into a matrix based on A foreground gated mask module of the function is used to obtain a body fluid sample mass spectrum semantic deep convolution gated mask weight feature map; a body fluid sample mass spectrum semantic gated mask foreground highlighting feature calculation unit 144 is used to calculate the position point multiplication between the body fluid sample mass spectrum semantic deep convolution gated mask weight feature map and the body fluid sample mass spectrum semantic deep convolution backup feature map to obtain a body fluid sample mass spectrum semantic gated mask foreground highlighting feature map; a body fluid sample mass spectrum key semantic significant expression feature map generation unit 145 is used to perform point convolution-based channel contraction on the body fluid sample mass spectrum semantic gated mask foreground highlighting feature map to obtain the body fluid sample mass spectrum key semantic significant expression feature map.
[0061] More specifically, in an embodiment of the present application, the body fluid sample mass spectrometry semantic feature gated mask calculation unit 143 is used to: set the feature values greater than a predetermined hyperparameter in each position of the body fluid sample mass spectrometry semantic deep convolution original feature map to one, and set the rest to zero to obtain the body fluid sample mass spectrometry semantic deep convolution gated mask weight feature map.
[0062] In the embodiment of the present application, specifically, the body fluid sample mass spectrum key semantic saliency module 140 is used to: process the body fluid sample mass spectrum semantic feature map through a body fluid sample key feature foreground saliency enhancer based on a convolution gating mechanism, using the following foreground saliency formula to obtain the body fluid sample mass spectrum key semantic saliency expression feature map; wherein the foreground saliency formula is:
[0063] ;
[0064] ;
[0065] ;
[0066] ;
[0067] ;
[0068] ;
[0069] ;
[0070] in, is the mass spectrum semantic feature map of the body fluid sample, Indicates the layer normalization operation on the feature map. is the normalized body fluid sample mass spectrum semantic feature map, is the point convolution operation, The convolution kernel is The dilated convolution operation, is the original feature map of the semantic deep convolution of the mass spectrum of the body fluid sample, A semantic deep convolution backup feature map of the mass spectrum of the body fluid sample, for Activation function, For masking, is the semantic deep convolution gated mask weight feature map of the body fluid sample mass spectrum, is the feature value of each position in the feature map, To predetermine the hyperparameters, For point-by-position multiplication, is a semantic gated mask foreground highlighting feature map of the mass spectrum of the body fluid sample, It is a key semantic saliency expression feature map of the mass spectrum of the body fluid sample.
[0071] In an embodiment of the present application, the endocrine state judgment module 150 is used to obtain an endocrine state detection result based on the key semantic significant expression feature of the body fluid sample mass spectrum, and the endocrine state detection result is used to indicate whether the endocrine state of the detected target patient object is abnormal. Specifically, in an embodiment of the present application, the endocrine state judgment module 150 is used to: pass the key semantic significant expression feature map of the body fluid sample mass spectrum through an endocrine abnormality detector based on a classifier to obtain the endocrine state detection result, and the endocrine state detection result is used to indicate whether the endocrine state of the detected target patient object is abnormal. That is, the body fluid sample mass spectrum key semantic significant expression feature obtained by foreground significant enhancement using the body fluid sample mass spectrum semantic feature map is classified and processed, so as to intelligently determine whether the endocrine state of the detected target patient object is abnormal. In this way, it is possible to detect hormones with lower abundance, provide more accurate quantitative analysis, and help discover early or trace endocrine changes. And the intelligent analysis system for endocrine abnormality detection can process and identify more types of biomarkers, including those special or emerging biomarkers that are difficult to detect by traditional methods.
[0072] In particular, the present application takes into account the multi-scale and multi-depth image semantic features of the contrast-enhanced body fluid sample mass spectrum expressed by the body fluid sample mass spectrometry semantic feature map. After performing key feature foreground saliency based on the convolution gating mechanism, the obtained body fluid sample mass spectrometry key semantic saliency expression feature map will also have convolution gating differences between local feature semantic features and key feature foregrounds, resulting in insufficient feature saliency coverage and outlier class inference mapping deviation, thereby affecting the accuracy of the endocrine status detection result obtained by the classifier-based endocrine abnormality detector using the body fluid sample mass spectrometry key semantic saliency expression feature map.
[0073] In a preferred example, the key semantic significant expression feature map of the body fluid sample mass spectrum is passed through a classifier-based endocrine abnormality detector to obtain an endocrine state detection result, including: calculating the feature mean of the key semantic significant expression feature map of the body fluid sample mass spectrum, and dividing the feature mean by the difference between the maximum feature value and the minimum feature value of the body fluid sample mass spectrum key semantic significant expression feature map to obtain a body fluid sample mass spectrum key semantic significant expression distribution representation value; subtracting the body fluid sample mass spectrum key semantic significant expression distribution representation value from the body fluid sample mass spectrum key semantic significant expression distribution representation value and dividing it by the body fluid sample mass spectrum key semantic significant expression distribution representation value to obtain a body fluid sample mass spectrum key semantic significant expression distribution modulation value; activating the body fluid sample mass spectrum key semantic significant expression feature map through a probabilistic function to obtain a probabilistic body fluid sample mass spectrum key semantic significant expression feature map; performing point subtraction between the probabilistic body fluid sample mass spectrum key semantic significant expression feature map and the body fluid sample mass spectrum key semantic significant expression distribution modulation value, taking the absolute value and calculating the value. The negative of the logarithmic value with base 2 is obtained to obtain a probabilistic body fluid sample mass spectrum key semantic significant expression distribution modulation information feature map; after dividing the body fluid sample mass spectrum key semantic significant expression distribution representation value by one minus the difference of each eigenvalue of the probabilistic body fluid sample mass spectrum key semantic significant expression feature map, all eigenvalues of the probabilistic body fluid sample mass spectrum key semantic significant expression feature map are summed up, and divided by the scale of the body fluid sample mass spectrum key semantic significant expression feature map to obtain a probabilistic body fluid sample mass spectrum key semantic significant expression distribution modulation bias value; the probabilistic body fluid sample mass spectrum key semantic significant expression distribution modulation information feature map and the probabilistic body fluid sample mass spectrum key semantic significant expression distribution modulation bias value and the weight as a hyperparameter are multiplied to obtain an optimized body fluid sample mass spectrum key semantic significant expression feature map; the optimized body fluid sample mass spectrum key semantic significant expression feature map is passed through the classifier-based endocrine abnormality detector to obtain the endocrine state detection result.
[0074] The optimized key semantic significant expression feature graph of the body fluid sample mass spectrum is expressed as:
[0075] ;
[0076] ;
[0077] in, is the probabilistic body fluid sample mass spectrum key semantic significant expression feature map, is the key semantically significant expression distribution representation value of the mass spectrum of the body fluid sample, Subtract by position point, Click on the location to add. is the weight as a hyperparameter, is the scale of the body fluid sample mass spectrum key semantic significant expression feature map, that is, the width of each feature matrix along the channel dimension of the body fluid sample mass spectrum key semantic significant expression feature map multiplied by the height and then multiplied by the number of channels of the body fluid sample mass spectrum key semantic significant expression feature map, is the feature value of each position in the probabilistic body fluid sample mass spectrum key semantic significant expression feature map, is the feature mean of the key semantic significant expression feature map of the mass spectrum of the body fluid sample, and are respectively the maximum eigenvalue and the minimum eigenvalue of the key semantic significant expression feature map of the mass spectrum of the body fluid sample, It is the optimized key semantically significant expression feature map of the mass spectrometry of the body fluid sample.
[0078] Based on this, the probability information distribution planning based on the feature value of the body fluid sample mass spectrum key semantic significant expression feature map is carried out through the Bernoulli probability modulation distribution of the body fluid sample mass spectrum key semantic significant expression feature map relative to the feature value distribution, and the probability reverse mapping of the probability feature of the body fluid sample mass spectrum key semantic significant expression feature map as a whole is used as the extended coverage of the set mapping space of the body fluid sample mass spectrum key semantic significant expression feature map, so as to independently understand the interactive path between the intuitive probability information distribution and the abstract probability space mapping of the body fluid sample mass spectrum key semantic significant expression feature map, so as to improve the accuracy of the endocrine state detection result obtained by the classifier-based endocrine abnormality detector of the optimized body fluid sample mass spectrum key semantic significant expression feature map by avoiding the counterfactual reasoning mapping of the outlier feature distribution of the body fluid sample mass spectrum key semantic significant expression feature map to the class probability. In this way, it is possible to detect hormones with lower abundance, provide more accurate quantitative analysis, and help discover early or trace endocrine changes. Moreover, the intelligent analysis system for endocrine abnormality detection is able to process and identify a wider variety of biomarkers, including those special or emerging biomarkers that are difficult to detect with traditional methods.
[0079] In summary, the intelligent analysis system 100 for endocrine abnormality detection based on the embodiment of the present application is explained, which first obtains the mass spectrogram of the body fluid sample of the target patient to be detected, and then uses the image recognition and processing technology based on machine learning to perform semantic analysis and feature enhancement processing of the body fluid sample mass spectrogram, so as to intelligently judge whether the endocrine state of the target patient to be detected is abnormal based on the key prospect significant feature information of the body fluid sample mass spectrogram. In this way, it is possible to detect hormones with lower abundance, thereby providing more accurate quantitative analysis, which is helpful to discover trace endocrine changes in the body of the target patient to be detected, and it is also possible to process and identify more types of biomarkers, thereby improving the intelligence level of endocrine detection and analysis.
[0080] As described above, the intelligent analysis system 100 for endocrine abnormality detection according to the embodiment of the present application can be implemented in various terminal devices, such as a server for intelligent analysis of endocrine abnormality detection. In one example, the intelligent analysis system 100 for endocrine abnormality detection according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the intelligent analysis system 100 for endocrine abnormality detection can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the intelligent analysis system 100 for endocrine abnormality detection can also be one of the many hardware modules of the terminal device.
[0081] Alternatively, in another example, the intelligent analysis system 100 for endocrine abnormality detection and the terminal device may also be separate devices, and the intelligent analysis system 100 for endocrine abnormality detection may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.
[0082] Figure 5 FIG. 1 is a flow chart of an intelligent analysis method for endocrine abnormality detection according to an embodiment of the present application. Figure 5As shown, in the intelligent analysis method for endocrine abnormality detection, it includes: S110, obtaining a body fluid sample mass spectrum of a target patient object to be detected collected by a liquid chromatography-mass spectrometer; S120, performing contrast enhancement processing on the body fluid sample mass spectrum to obtain a contrast enhanced body fluid sample mass spectrum; S130, passing the contrast enhanced body fluid sample mass spectrum through a body fluid sample mass spectrum feature extractor based on a pyramid structure to obtain a body fluid sample mass spectrum semantic feature map; S140, passing the body fluid sample mass spectrum semantic feature map through a body fluid sample key feature foreground saliency enhancer based on a convolution gating mechanism to obtain a body fluid sample mass spectrum key semantic saliency expression feature map as a body fluid sample mass spectrum key semantic saliency expression feature; S150, based on the body fluid sample mass spectrum key semantic saliency expression feature, obtaining an endocrine status detection result, and the endocrine status detection result is used to indicate whether the endocrine state of the target patient object to be detected is abnormal.
[0083] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned intelligent analysis method for endocrine abnormality detection have been described in the above reference. Figures 1 to 4 The invention has been described in detail in the description of the intelligent analysis system for endocrine abnormality detection, and therefore, its repeated description will be omitted.
[0084] In summary, the intelligent analysis method for endocrine abnormality detection based on the embodiment of the present application is explained, which first obtains the mass spectrogram of the body fluid sample of the target patient to be detected, and then uses the image recognition and processing technology based on machine learning to perform semantic analysis and feature enhancement processing of the mass spectrogram of the body fluid sample, so as to intelligently judge whether the endocrine state of the target patient to be detected is abnormal based on the key prospect significant feature information of the mass spectrogram of the body fluid sample. In this way, hormones with lower abundance can be detected to provide more accurate quantitative analysis, which is helpful to discover trace endocrine changes in the body of the target patient to be detected, and at the same time, it can also process and identify more types of biomarkers, thereby improving the intelligence level of endocrine detection and analysis.
[0085] The above is only a preferred embodiment of the present application and does not constitute any form of limitation to the present application. Although the present application has been disclosed as a preferred embodiment as above, it is not intended to limit the present application. Any technician familiar with the profession can make some changes or modifications to equivalent embodiments of the technical contents disclosed above without departing from the scope of the technical solution of the present application. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present application without departing from the content of the technical solution of the present application still falls within the scope of the technical solution of the present application.
Claims
1. An intelligent analysis system for endocrine abnormality detection, characterized in that: include: A patient subject body fluid sample mass spectrum acquisition module, used to acquire a mass spectrum of a body fluid sample of a target patient subject collected by a liquid chromatography-mass spectrometer; A body fluid sample mass spectrum contrast enhancement module, used for performing contrast enhancement processing on the body fluid sample mass spectrum to obtain a contrast enhanced body fluid sample mass spectrum; A body fluid sample mass spectrum semantic feature extraction module, used for passing the contrast enhanced body fluid sample mass spectrum through a body fluid sample mass spectrum feature extractor based on a pyramid structure to obtain a body fluid sample mass spectrum semantic feature map; A body fluid sample mass spectrum key semantic saliency module, used for passing the body fluid sample mass spectrum semantic feature map through a body fluid sample key feature foreground saliency enhancer based on a convolution gating mechanism to obtain a body fluid sample mass spectrum key semantic saliency expression feature map as a body fluid sample mass spectrum key semantic saliency expression feature; An endocrine state judgment module, used to obtain an endocrine state detection result based on the key semantically significant expression features of the mass spectrum of the body fluid sample, wherein the endocrine state detection result is used to indicate whether the endocrine state of the detected target patient object is abnormal; The body fluid sample mass spectrum key semantic saliency module includes: A body fluid sample mass spectrum semantic feature map layer normalization unit, used for performing layer normalization processing on the body fluid sample mass spectrum semantic feature map to obtain a normalized body fluid sample mass spectrum semantic feature map; A normalized body fluid sample mass spectrum semantic feature map deep convolution encoding unit, used for performing point convolution-based channel expansion and deep convolution encoding based on a hole convolution layer on the normalized body fluid sample mass spectrum semantic feature map to obtain a body fluid sample mass spectrum semantic deep convolution original feature map and a body fluid sample mass spectrum semantic deep convolution backup feature map; The body fluid sample mass spectrum semantic feature gated mask calculation unit is used to input the body fluid sample mass spectrum semantic deep convolution original feature map into the The foreground gated mask module of the function is used to obtain the semantic deep convolution gated mask weight feature map of the mass spectrum of the body fluid sample; A body fluid sample mass spectrum semantic gated mask foreground highlighting feature calculation unit, used to calculate the position point multiplication between the body fluid sample mass spectrum semantic deep convolution gated mask weight feature map and the body fluid sample mass spectrum semantic deep convolution backup feature map to obtain a body fluid sample mass spectrum semantic gated mask foreground highlighting feature map; The body fluid sample mass spectrum key semantic salient expression feature map generation unit is used to perform point convolution-based channel contraction on the body fluid sample mass spectrum semantic gated mask foreground highlighting feature map to obtain the body fluid sample mass spectrum key semantic salient expression feature map.
2. The intelligent analysis system for endocrine abnormality detection according to claim 1, characterized in that: The body fluid sample mass spectrum contrast enhancement module is used to perform piecewise linear transformation processing on the body fluid sample mass spectrum to obtain the contrast enhanced body fluid sample mass spectrum.
3. The intelligent analysis system for endocrine abnormality detection according to claim 2, characterized in that: The body fluid sample mass spectrum semantic feature extraction module includes: A contrast-enhanced body fluid sample mass spectrum shallow feature extraction unit, used for passing the contrast-enhanced body fluid sample mass spectrum through a convolution layer with a convolution kernel of 3×3 to obtain a contrast-enhanced body fluid sample mass spectrum shallow feature map; A contrast-enhanced body fluid sample mass spectrum shallow fusion unit, used for passing the contrast-enhanced body fluid sample mass spectrum shallow feature map through the body fluid sample mass spectrum shallow residual information fusion enhancement module to obtain a contrast-enhanced body fluid sample mass spectrum shallow residual fusion feature map; A contrast-enhanced body fluid sample mass spectrum deep feature extraction unit, used for passing the contrast-enhanced body fluid sample mass spectrum shallow residual fusion feature map through a convolution layer with a convolution kernel of 3×3 to obtain a contrast-enhanced body fluid sample mass spectrum deep feature map; A contrast-enhanced body fluid sample mass spectrum deep fusion unit, used for passing the contrast-enhanced body fluid sample mass spectrum deep feature map through a body fluid sample mass spectrum deep residual information fusion enhancement module to obtain a contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map; A contrast-enhanced body fluid sample mass spectrum feature modulation unit, used for inputting the contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map into a scale modulation module based on an SPPF layer to obtain a contrast-enhanced body fluid sample mass spectrum deep modulation feature map; A contrast-enhanced body fluid sample mass spectrum deep modulation feature upsampling unit, used for upsampling the contrast-enhanced body fluid sample mass spectrum deep modulation feature map to obtain an upsampled contrast-enhanced body fluid sample mass spectrum deep modulation feature map; A contrast-enhanced body fluid sample mass spectrum multi-scale feature generation unit, used for fusing the upsampled contrast-enhanced body fluid sample mass spectrum deep modulation feature map with the contrast-enhanced body fluid sample mass spectrum shallow residual fusion feature map to obtain a contrast-enhanced body fluid sample mass spectrum multi-scale feature map; The body fluid sample mass spectrum semantic feature generation unit is used to obtain the body fluid sample mass spectrum semantic feature map by fusing the contrast enhanced body fluid sample mass spectrum multi-scale feature map through the contrast enhanced body fluid sample mass spectrum multi-scale residual information enhancement module.
4. The intelligent analysis system for endocrine abnormality detection according to claim 3, characterized in that: The contrast-enhanced body fluid sample mass spectrometry shallow fusion unit is used to: The contrast-enhanced body fluid sample mass spectrum shallow feature map is subjected to channel dimensionality reduction modulation through a point convolution layer to obtain a channel dimensionality reduction modulated contrast-enhanced body fluid sample mass spectrum shallow feature map; The shallow feature map of the mass spectrum of the body fluid sample with the channel dimension reduction modulation contrast enhancement is extracted through a convolution layer with a convolution kernel of 3×3 to obtain a shallow implicit information feature map of the mass spectrum of the body fluid sample with contrast enhancement; The shallow implicit information feature map of the contrast-enhanced body fluid sample mass spectrum is modulated by a point convolution layer to obtain a shallow feature map of the contrast-enhanced body fluid sample mass spectrum after channel modulation; The shallow feature map of the mass spectrum of the contrast-enhanced body fluid sample after channel modulation and the shallow feature map of the mass spectrum of the contrast-enhanced body fluid sample after channel dimension reduction modulation are fused to obtain the shallow residual fusion feature map of the mass spectrum of the contrast-enhanced body fluid sample.
5. The intelligent analysis system for endocrine abnormality detection according to claim 4, characterized in that: The contrast enhancement body fluid sample mass spectrum characteristic modulation unit is used to: The contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map is passed through a scale modulation module based on a point convolution layer to obtain a scale-adjusted contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map; Performing maximum pooling processing on each feature matrix along the channel dimension in the scale-adjusted contrast-enhanced body fluid sample mass spectrum deep residual fusion feature map to obtain a contrast-enhanced body fluid sample mass spectrum deep residual fusion feature vector; Based on the deep residual fusion feature vector of the contrast enhanced body fluid sample mass spectrum, the scale-adjusted contrast enhanced body fluid sample mass spectrum deep residual fusion feature map is channel-weighted enhanced to obtain the contrast enhanced body fluid sample mass spectrum deep modulation feature map.
6. The intelligent analysis system for endocrine abnormality detection according to claim 5, characterized in that: The body fluid sample mass spectrum semantic feature gated mask calculation unit is used to: set the feature values greater than the predetermined hyperparameter in each position of the body fluid sample mass spectrum semantic deep convolution original feature map to one, and set the rest to zero to obtain the body fluid sample mass spectrum semantic deep convolution gated mask weight feature map.
7. The intelligent analysis system for endocrine abnormality detection according to claim 6, characterized in that: The endocrine status judgment module is used to: pass the key semantic significant expression feature map of the body fluid sample mass spectrum through a classifier-based endocrine abnormality detector to obtain the endocrine status detection result, and the endocrine status detection result is used to indicate whether the endocrine status of the target patient being detected is abnormal.
