Method for establishing depression diagnosis model based on infrared spectrum in trace blood
Through the mid-infrared spectroscopy of trace blood samples combined with multiple preprocessing and machine learning algorithms, a depression diagnosis model was established, which solved the subjective and empirical problems of depression diagnosis in the existing technology, and achieved rapid and accurate depression diagnosis , and continuously optimize model performance in practice.
Patent Information
- Application Number
- CN202411275576.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-12
- Publication Date
- 2024-12-13
AI Technical Summary
The existing diagnostic methods for depression are mainly based on clinical symptoms, subjective and empirical, lack of objective biomarkers, making it difficult to achieve early warning and precise diagnosis and treatment.
By collecting mid-infrared spectra of trace blood samples, combining multiple preprocessing methods and machine learning algorithms, a depression diagnosis model is established to achieve rapid and accurate diagnosis.
Without sample pretreatment, the diagnosis can be completed in a short time, with high accuracy and robustness, and in practice, parameter optimization and model upgrade can be carried out, with diagnostic accuracy of more than 99%.
Smart Images
Figure CN119132574A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of depression diagnosis, and in particular to a method for establishing a depression diagnosis model based on a trace amount of blood mid-infrared spectrum. Background Art
[0002] Depression, also known as depressive disorder, is a mood disorder characterized by significant and persistent low mood. This low mood is disproportionate to one's situation and can range from melancholy to grief. Due to the high heterogeneity and complex pathogenesis of depression, its diagnosis and treatment are currently mainly based on clinical symptoms, which are subjective and empirical. There is an urgent need to develop early warning and precise diagnosis and treatment methods based on objective biomarkers.
[0003] Mid-infrared spectroscopy is a fast, simple and efficient detection and analysis technology. Molecules selectively absorb 2.5-25μm (about 400-4000cm -1 ) wavelengths of infrared light cause the transition of vibrational energy levels and rotational energy levels in molecules, thereby obtaining a mid-infrared absorption spectrum. Mid-infrared spectroscopy has the characteristics of high maturity (mid-infrared spectroscopy instruments are the most mature and simple, with a long history of use and rich data accumulation) and strong analytical ability (fundamental frequency vibration is the strongest vibration absorbed among infrared active vibrations), and has been widely used in qualitative and quantitative analysis in many fields such as environmental monitoring, food safety, and drug analysis. The present invention obtains molecular structural information such as proteins, lipids, and nucleic acids by collecting mid-infrared spectra of trace blood samples, thereby realizing rapid diagnosis of depression. Summary of the invention
[0004] The invention provided by the present invention aims to provide a method for establishing a depression diagnosis model based on the mid-infrared spectrum of micro-blood. The method for establishing a depression diagnosis model based on the mid-infrared spectrum of micro-blood of the present invention does not require sample pre-treatment, and the mid-infrared spectrum of the patient to be diagnosed can be identified in a short time only by micro-blood, which has the advantages of being simple, fast, and accurate. It integrates multiple pre-processing methods and machine learning algorithms, has high accuracy and robustness in diagnostic applications, and can perform parameter optimization and model upgrades in practice.
[0005] In order to achieve the above effects, the present invention provides the following technical solution: a method for establishing a depression diagnosis model based on micro-blood mid-infrared spectroscopy, comprising the following steps:
[0006] S1. Collect and preserve trace blood samples from several clinical depression patients and healthy controls.
[0007] S2. Collect the mid-infrared spectrum of the blood sample obtained in step S1 to form a mid-infrared spectrum set, and pre-process the obtained spectrum set according to a variety of methods.
[0008] S3. Using the differently pre-processed mid-infrared spectra obtained in step S2, multiple basic models for mid-infrared spectrum recognition are established through a variety of machine learning algorithms, and a final mid-infrared spectrum recognition model is established through a certain model integration method.
[0009] S4. Collect the mid-infrared spectrum of the blood sample of the clinical patient to be diagnosed, and substitute it into the mid-infrared spectrum recognition model established in step S3 to obtain the spectrum recognition result.
[0010] S5. During the application of the mid-infrared spectral pattern recognition model, when the number of mid-infrared spectra outside the model that obtain real clinical diagnosis results accumulates to a certain amount, repeat steps S2-S3 to optimize and improve the model performance until the model diagnosis accuracy is higher than 99%.
[0011] Further, according to the operating steps in S1, the volume of the micro-blood sample is 1-10 μl, the storage temperature is -80°C, and the test is completed within 24 hours after thawing at room temperature.
[0012] Furthermore, according to the operation steps in S2, the blood sample needs to be dripped or applied onto a ZnSe glass slide or a diamond ATR crystal, and after being naturally dried for 5-30 minutes, a mid-infrared spectrum is collected in a transmission or attenuated total reflection mode.
[0013] Further, according to the operation steps in S2, the acquisition range of the mid-infrared spectrum is 400-4000cm -1 , resolution range 1-16cm -1 , the scan times range is 2-64 times.
[0014] Further, according to the operating steps in S2, the preprocessing method of the mid-infrared spectrum includes at least one of SG smoothing (Savitzky-Golay, SG), wavelet transform (Wavelet Transform, WT), moving window average (Moving Window Average, MWA), standard normal variable transformation (Standard Normal Variable Transformation, SNV), multivariate scattering correction (Multiple Scattering Correction, MSC), normalization (Normalization, Norm), mean centering (Mean Centering, MC), standardization (Standardization, S), first-order derivative, and second-order derivative.
[0015] Further, according to the operating steps in S3, the multiple machine learning algorithms include principal component-linear discriminant analysis (PCA-LDA), principal component-quadratic discriminant analysis (PCA-QDA), partial least squares discriminant analysis (PLSDA), K-nearest neighbor (KNN) classification, logistic regression model, support vector machine (SVM), random forests (RF), extreme gradient boosting tree (XGBoost), multilayer perceptron (Multilayer Perceptron), etc. Perceptron, MLP) in one or more.
[0016] Further, according to the operation steps in S3, the model integration method includes one of voting method, averaging method, stacking method and non-cross stacking method.
[0017] Further, according to the operation steps in S4, the performance evaluation parameters of the model include accuracy, area under the receiver operating characteristic curve (AUC), sensitivity and specificity.
[0018] Further, according to the operation steps in S5, the number of accumulated mid-infrared spectra outside the model ranges from 200 to 1000.
[0019] Further, according to the operation steps in S5, the actual clinical diagnosis results include but are not limited to clinical manifestations, disease course requirements, laboratory tests, imaging spectroscopy, psychological tests and clinician evaluations.
[0020] The present invention provides a method for establishing a depression diagnosis model based on a trace amount of blood mid-infrared spectrum, which has the following beneficial effects:
[0021] (1) The present invention does not require sample pretreatment and can complete the mid-infrared spectrum recognition of the patient to be diagnosed in a short time using only a small amount of blood (as low as 1 microliter), which has the advantages of being simple, fast, and accurate.
[0022] (2) The present invention integrates multiple preprocessing methods and machine learning algorithms, and has high accuracy and robustness in diagnostic applications.
[0023] (3) The method provided by the present invention can perform parameter optimization and model upgrading in practice. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 The figure is a flow chart of the method for establishing a depression diagnosis model based on the infrared spectrum of trace blood according to the present invention.
[0025] Figure 2 This is the receiver operating characteristic curve of the final mid-infrared spectrum recognition model in Example 1 on the test set. DETAILED DESCRIPTION
[0026] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.
[0027] Embodiment 1
[0028] See also Figure 1 The method for establishing a depression diagnosis model based on micro-blood mid-infrared spectroscopy includes the following steps: collect serum from 200 clinical depression patients and 200 healthy people, 5 microliters of each sample, and store them in a -80℃ refrigerator for later use. After the frozen blood sample is thawed at room temperature, 1 microliter is taken with a pipette and dripped onto a diamond ATR crystal. After 5 minutes of natural drying, the mid-infrared spectrum is collected in the attenuated total reflection mode. The spectrum collection range is 400-4000cm -1 , resolution range 1cm -1, the scanning number range is 16 times. All 200 samples were collected to obtain 200 spectra to form the spectrum set of the mid-infrared spectrum recognition model. The original infrared spectrum set was subjected to 12 preprocessing methods, including baseline correction, SG smoothing, MSC, SNV, standardization, MWA, first-order derivative, second-order derivative, principal component analysis, SG smoothing + first-order derivative, SG smoothing + second-order derivative, and MSC + first-order derivative, to obtain 12 preprocessed spectrum sets. The preprocessed spectrum set was randomly divided into training set and test set, where the training set was modeled using 9 machine learning algorithms, including PCA-LDA, PCA-QDA, PLSDA, KNN classification, logistic regression model, SVM, RF, XGBoost, and MLP. The models were ranked from high to low according to their accuracy (number of correctly predicted samples / total number of samples) on the test set, namely: SG smoothing + first-order derivative + MLP, SG smoothing + second-order derivative + KNN, baseline correction + RF, and baseline correction + SVM. The above five optimal models were integrated using the Blending method. The specific principle is that the original spectral training set is predicted by the five optimal models to obtain the probability of the classification result, and the probability is used as the data set and the true label of the sample is used as the label for secondary model training to obtain the final mid-infrared spectral recognition model. Figure 2 As shown in the receiver operating characteristic curve, the prediction accuracy of the final mid-infrared spectral recognition model on the test set is 89%, the AUC is 0.93, and the sensitivity and specificity are 96% and 75%, respectively. Then, the mid-infrared spectra of the serum samples of clinical patients to be diagnosed are collected and substituted into the mid-infrared spectral recognition model established in step S3 to obtain the spectrum recognition results. In the actual application of the model, when the number of mid-infrared spectra outside the model with real clinical diagnosis results accumulates to 400 cases, these data are merged with the initially modeled data set, and the model is retrained to improve performance until the model diagnosis accuracy is higher than 99%.
[0029] Embodiment 2
[0030] See also Figure 1 The method for establishing a depression diagnosis model based on micro-blood mid-infrared spectroscopy includes the following steps: 100 blood samples from clinical depression patients and 100 plasma samples from healthy people are collected, 10 microliters of each sample is stored in a -80℃ refrigerator for later use. After the frozen blood sample is thawed at room temperature, 1 microliter is taken with a pipette and dropped onto a diamond ATR crystal. After 20 minutes of natural drying, the mid-infrared spectrum is collected in the attenuated total reflection mode. The spectrum collection range is 400-4000cm -1 , resolution range 4cm -1, the scanning number range is 32 times. All 400 samples were collected to obtain 400 spectra to form the spectrum set of the mid-infrared spectrum recognition model. The original infrared spectrum set was subjected to 12 preprocessing methods, including baseline correction, SG smoothing, MSC, SNV, standardization, MWA, first-order derivative, second-order derivative, principal component analysis, SG smoothing + first-order derivative, SG smoothing + second-order differential, and MSC + first-order derivative, to obtain 12 preprocessed spectrum sets. For each spectral set, the Kennard-Stone algorithm was used to divide it into a training set and a test set. The training set was modeled using nine machine learning algorithms, including PCA-LDA, PCA-QD A, PLSDA, KNN classification, logistic regression model, SVM, RF, XGBoost, and MLP. The models were ranked from high to low according to their accuracy on the test set (number of correctly predicted samples / total number of samples), namely: SG smoothing + first-order derivative + XGBoost (partial parameters: 'n_est imators': 74, 'l earn i ng_rate': 0.67, 'max_depth': 5, 'reg_a l pha': 3.5), SG smoothing + second-order derivative + PLSDA (number of factors 11), RF (partial parameters: {'n_est imators': 95, 'max_features': 25, 'max_depth': 7, 'oob_score': true), baseline correction + SVM (RBF kerne l, C = 1285, gamma = 0.85), SNV + PCA-QDA (factor number 23). The above five optimal models are integrated by the Stacking method. The specific principle is that the original spectrum training set is predicted by the five optimal models to obtain the probability of the classification result, and the probability is used as the data set and the real label of the sample is used as the label for secondary model training to obtain the final mid-infrared spectrum recognition model. Its prediction accuracy on the test set is 90%, AUC is 0.92, and the sensitivity and specificity are 97% and 78%, respectively. Then, the mid-infrared spectrum of the plasma of the clinical patient to be diagnosed is collected and substituted into the mid-infrared spectrum recognition model established in step S3 to obtain the spectrum recognition result. In the actual application of the model, when the number of mid-infrared spectra outside the model with real clinical diagnosis results accumulates to 400 cases, these data are merged with the initially modeled data set, and the model is retrained to improve performance until the model diagnosis accuracy is higher than 99%.
[0031] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A method for establishing a depression diagnosis model based on micro-blood mid-infrared spectroscopy, characterized in that: The following steps are involved: S1. Collect and preserve micro blood samples from several clinical depression patients and healthy controls; S2, collecting the mid-infrared spectrum of the blood sample obtained in step S1 to form a mid-infrared spectrum set, and preprocessing the obtained spectrum set according to a variety of methods; S3, using the mid-infrared spectra obtained in step S2 after different preprocessing, establishing multiple basic models for mid-infrared spectrum recognition through multiple machine learning algorithms, and establishing a final mid-infrared spectrum recognition model through a certain model integration method; S4, collecting the mid-infrared spectrum of the blood sample of the clinical patient to be diagnosed, and substituting it into the mid-infrared spectrum recognition model established in step S3 to obtain the spectrum recognition result; S5. During the application of the mid-infrared spectral pattern recognition model, when the number of mid-infrared spectra outside the model that obtain real clinical diagnosis results accumulates to a certain amount, repeat steps S2-S3 to optimize and improve the model performance until the model diagnosis accuracy is higher than 99%.
2. The method for establishing a depression diagnosis model based on micro-blood mid-infrared spectroscopy according to claim 1, characterized in that: The method comprises the following steps: according to the operation steps in S1, the volume of the micro blood sample is 1-10 microliters, the storage temperature is -80°C, and the test is completed within 24 hours after thawing at room temperature.
3. The method for establishing a depression diagnosis model based on micro-blood mid-infrared spectroscopy according to claim 1, characterized in that: The method comprises the following steps: according to the operation steps in S2, the blood sample needs to be dripped or coated on a ZnSe glass slide or a diamond ATR crystal, and after being naturally dried for 5-30 minutes, a mid-infrared spectrum is collected in a transmission or attenuated total reflection mode.
4. The method for establishing a depression diagnosis model based on micro-blood mid-infrared spectroscopy according to claim 1, characterized in that: The following steps are included: according to the operation steps in S2, the mid-infrared spectrum is collected in the range of 400-4000cm -1 , resolution range 1-16cm -1 , the scan times range is 2-64 times.
5. The method for establishing a depression diagnosis model based on micro-blood mid-infrared spectroscopy according to claim 1, characterized in that: The following steps are involved: According to the operation steps in S2, the preprocessing method of the mid-infrared spectrum includes at least one of SG smoothing, wavelet transform, moving window averaging, standard normal variable transformation, multivariate scattering correction, normalization, mean centering, standardization, first-order derivative, and second-order derivative.
6. The method for establishing a depression diagnosis model based on micro-blood mid-infrared spectroscopy according to claim 1, characterized in that: The following steps are involved: According to the operation steps in S3, the multiple machine learning algorithms include principal component-linear discriminant analysis, principal component-quadratic discriminant analysis, partial least squares discriminant analysis, K nearest neighbor classification, logistic regression model, support vector machine, random forest, extreme gradient boosting tree, and one or more of multilayer perceptron.
7. The method for establishing a depression diagnosis model based on micro-blood mid-infrared spectroscopy according to claim 1, characterized in that: The following steps are involved: According to the operation steps in S3, the model integration method includes one of a voting method, an averaging method, a stacking method, and a non-cross stacking method.
8. The method for establishing a depression diagnosis model based on micro-blood mid-infrared spectroscopy according to claim 1, characterized in that: The method comprises the following steps: according to the operation steps in S4, the performance evaluation parameters of the model include accuracy, area under the receiver operating characteristic curve, sensitivity and specificity.
9. The method for establishing a depression diagnosis model based on micro-blood mid-infrared spectroscopy according to claim 1, characterized in that: The method comprises the following steps: according to the operation steps in S5, the number of accumulated mid-infrared spectra outside the model ranges from 200 to 1000.
10. The method for establishing a depression diagnosis model based on micro-blood mid-infrared spectroscopy according to claim 1, characterized in that: The following steps are included: according to the operation steps in S5, the actual clinical diagnosis results include but are not limited to clinical manifestations, disease course requirements, laboratory tests, imaging spectroscopy, psychological tests and clinician evaluations.