Newborn congenital heart disease screening method based on tandem mass spectrometry

Abstract

The invention discloses a newborn congenital heart disease screening method based on tandem mass spectrometry, and the method comprises the following steps: 1, collecting heel blood tandem mass spectrometry data of newborns subjected to echocardiography examination after birth, and preprocessing the data; 2, performing feature extraction on the data collected and preprocessed in the step 1, and dividing the data into a training set and a test set according to the extracted features; and 3, constructing a model by using a sklearn packet in Python, training the model on the training set divided in the step 2, evaluating the model on the test set, inputting data to be screened into the model, and screening data of the congenital heart through the model, wherein the evaluation index is the area under the ROC curve. According to the newborn congenital heart disease screening method based on tandem mass spectrometry, ventricular septal defect with the maximum proportion in congenital heart disease is focused on, and good economic cost benefits are achieved.

Description

Technical Field

[0001] The invention relates to a data processing method, and more particularly to a method for screening congenital heart disease in newborns based on tandem mass spectrometry. Background Art

[0002] In recent years, some research on screening models for CHD and other birth defects has made progress. Yin et al. 16 used proteomic analysis and machine learning models to study early pregnancy plasma samples to identify biomarkers for early detection of CHD, but the study was limited by the small sample size and has not yet been validated in a large-scale clinical setting. Chen et al. 17 developed a CHDdECG model to screen for CHD using children’s ECG data, achieving a ROC-AUC of more than 0.9 on the test set and outperforming cardiologists in CHD diagnosis, but the difficulty in obtaining ECG data in newborns limits its application in actual clinical practice. Tang et al. 18 A two-stage deep transfer learning model, DDCHD-DenseNet, was developed, which showed high sensitivity and specificity in screening critical ductus arteriosus-dependent CHD. However, this model is only targeted at ductus arteriosus-dependent CHD and requires echocardiography, which limits its application in resource-limited areas. Summary of the invention

[0003] In view of the shortcomings of the prior art, the object of the present invention is to provide a method for screening congenital heart disease in newborns based on tandem mass spectrometry, which saves a lot of manpower and material resources.

[0004] To achieve the above object, the present invention provides the following technical solution: a method for screening congenital heart disease in newborns based on tandem mass spectrometry, characterized in that it comprises the following steps:

[0005] Step 1: collect heel blood tandem mass spectrometry data of newborns who have undergone echocardiography after birth and preprocess the data;

[0006] Step 2: extract features from the data collected and preprocessed in step 1, and divide the data into a training set and a test set according to the extracted features;

[0007] Step three, use the sklearn package in Python to build a model. The model is trained on the training set divided in step two and evaluated on the test set. The evaluation indicator is the area under the ROC curve. The data to be screened is input into the model, and the data of congenital heart is screened out through the model.

[0008] As a further improvement of the present invention, the specific method of preprocessing the data in step one is: matching the echocardiogram results and the heel blood tandem mass spectrometry data according to personal information, establishing a database, and dividing it into a congenital heart disease group, a physiological change group, and a control group.

[0009] As a further improvement of the present invention, the specific method of feature extraction in step 2 is: the normal group and the physiological change group are defined as negative samples, the CHD group is defined as a positive sample, 46 differentially expressed metabolites and clinical characteristics such as maternal age, birth weight and gestational age are extracted, 60% of the total samples are used as a training set, and 40% are used as a test set.

[0010] As a further improvement of the present invention, in step three, after the model is constructed, a 5-fold cross-validation method is used to validate the model to ensure that the sample ratio of each category in each subset is consistent with the original data.

[0011] The beneficial effects of the present invention are that the present invention shows reasonable predictive ability in identifying VSD cases (AUC reaches 0.71 after optimization), especially cases with defects >3 mm. It still provides a cost-effective and efficient new method for screening congenital heart disease (CHD) in newborns. The present invention is based on existing genetic and metabolic screening data of newborns, does not require additional testing, and saves a lot of manpower and material resources. Although the "dual indicator method" covers almost all severe congenital heart diseases (CCHD), the invention can effectively identify VSD, the largest type of CHD subtype, showing high cost-effectiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A diagram showing the accuracy score of a machine learning algorithm. DETAILED DESCRIPTION

[0013] The present invention will be further described below in detail with reference to the embodiments shown in the accompanying drawings.

[0014] Reference Figure 1 As shown, a method for screening congenital heart disease in newborns based on tandem mass spectrometry in this embodiment specifically includes the following steps:

[0015] 1. Data processing: Newborns who underwent echocardiography after birth in Zhejiang Province from April 2020 to December 2021 were collected, and heel blood tandem mass spectrometry data were collected. The echocardiography results and heel blood tandem mass spectrometry data were matched according to personal information to establish a database (n=22539) and divided them into congenital heart disease group, physiological change group, and control group.

[0016] 2. Feature extraction: The normal group and the physiological change group were defined as negative samples, and the CHD group was defined as positive samples. 46 differentially expressed metabolites and clinical characteristics such as maternal age, birth weight, and gestational age were extracted. 60% of the total samples were used as training sets and 40% as test sets.

[0017] 3. Model construction and validation: The sklearn package (version 1.2.2) in Python (version 3.9.12) was used. The algorithm was trained on the training set and evaluated on the test set, with the evaluation indicator being the area under the ROC curve (AUC). To increase the robustness of the model and avoid overfitting, a 5-fold cross-validation was used to ensure that the proportion of samples in each category in each subset was consistent with the original data. The model was further optimized through strict inclusion criteria, hyperparameter adjustment, and feature selection to improve its accuracy. Eighteen machine learning methods were tried to identify CHD in the early neonatal period. To improve the model performance, we subsequently developed a model specifically for VSD screening, considering that VSD is the most common subtype in clinical practice. VSD patients were considered as positive samples, and the normal group was considered as negative samples. The results showed that the VSD screening model had a strong predictive ability, and an AUC value of 0.62 was obtained in the test set. Subsequently, to further improve the performance of the VSD screening model, we restricted the positive samples to VSD cases with defects greater than 3 mm, and the negative samples to the normal group, and performed a 5-fold cross-validation. The performance of various machine learning models is shown in Table 2. Arrange in descending order according to the AUC value in the test set. The best model is the GradientBoosting Classifier, which achieves an AUC value of 0.68 in the test set. After optimization techniques such as hyperparameter tuning, the accuracy can reach 0.71.

[0018] Table 2

[0019]

[0020]

[0021] Train setAUC:the averageAUC of the training set

[0022] TestsetAUC:the averageAUC of the testing set

[0023] TestsetAUC:3*STD:three times the standard deviation of the testing setAUC

[0024] In summary, the method for screening congenital heart disease in newborns based on tandem mass spectrometry in this embodiment, from the perspective of health economics, factors affecting the cost of CHD screening include the time, equipment and instruments of the screening personnel, the cost of professional training, and the cost of consumables. However, due to the gap in economic and medical resources, a large number of CHD patients have not been discovered or served. A study previously conducted in three regions with different socioeconomic conditions in China (Beijing, Shandong and Gansu) aimed to estimate the cost and health benefits of postpartum blood oxygen saturation screening. The results showed that the method was highly cost-effective in Beijing (metropolitan area), had a certain cost-effectiveness in Shandong (developed area), and had no cost-effectiveness in Gansu (underdeveloped area). This shows that the "dual index method" is only cost-effective in more developed areas with higher per capita GDP and more complete pediatric heart disease care services, which is similar to the research results in developed countries. In view of the differences in economic and medical resources, the screening system should be implemented from multiple perspectives in combination with the current level of economic development. The present invention is based on existing genetic and metabolic screening data for newborns, does not require additional testing, and saves a lot of manpower and material resources. Although the "dual indicator method" covers almost all serious congenital heart diseases (CCHD), this invention can effectively identify VSD, the largest type of CHD subtype, and shows high cost-effectiveness.

[0025] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A method for screening congenital heart disease in newborns based on tandem mass spectrometry, characterized in that: The steps include: Step 1: collect heel blood tandem mass spectrometry data of newborns who have undergone echocardiography after birth and preprocess the data; Step 2: extract features from the data collected and preprocessed in step 1, and divide the data into a training set and a test set according to the extracted features; Step three, use the sklearn package in Python to build a model. The model is trained on the training set divided in step two and evaluated on the test set. The evaluation indicator is the area under the ROC curve. The data to be screened is input into the model, and the data of congenital heart is screened out through the model.

2. The method for screening congenital heart disease in newborns based on tandem mass spectrometry according to claim 1, characterized in that: The specific method of preprocessing the data in step 1 is: matching the echocardiogram results and the heel blood tandem mass spectrometry data according to the personal information, establishing a database, and dividing it into a congenital heart disease group, a physiological change group, and a control group.

3. The method for screening congenital heart disease in newborns based on tandem mass spectrometry according to claim 2, characterized in that: The specific method of feature extraction in step 2 is: defining the normal group and the physiological change group as negative samples, defining the CHD group as positive samples, extracting 46 differentially expressed metabolites and clinical characteristics such as maternal age, birth weight and gestational age, and using 60% of the total samples as training sets and 40% as test sets.

4. The method for screening congenital heart disease in newborns based on tandem mass spectrometry according to claim 3, characterized in that: After the model is constructed in step three, a 5-fold cross-validation method is used to validate the model to ensure that the sample proportion of each category in each subset is consistent with the original data.

Images