Hemodialysis patient survival prediction method based on feature filling auto-encoder

Abstract

The invention discloses a hemodialysis patient survival prediction method based on a feature filling auto-encoder. The hemodialysis patient survival prediction method solves the problems of imbalance of short-term hemodialysis data and data feature missing. The method comprises the following steps of: firstly, inferring a value of a missing feature by utilizing an internal relationship between a manifold structure of a non-missing feature and a feature in a high-dimensional data space, then simulating missing data distribution in a deep learning framework through generation of a feature discarding mask, and designing a feature filling auto-encoder; the invention relates to a death rate prediction module, which performs unsupervised learning by using easily obtained short-term data, enables an encoder to reconstruct missing features and derive potential characterization, and finally realizes death rate prediction by using a classifier based on the potential characterization, and aims to provide more accurate death rate risk assessment for hemodialysis patients by using the short-term hemodialysis data.

Description

Technical Field

[0001] The present invention relates to the technical field of medical information prediction, and more specifically to a method for predicting survival of hemodialysis patients based on a feature-filled autoencoder. Background Art

[0002] Patients with end-stage renal disease (ESRD) face a significantly high mortality risk. The annual report of the US Renal Data System shows that from 2019 to 2021, under the influence of COVID-19, the all-cause mortality rate of adult ESRD patients increased from 15.9% to 18.8%, of which 11.6% of patients died within 6 months. Although the survival rate of ESRD patients has improved, the mortality rate of chronic kidney disease (CKD) patients within 18 months of progression to ESRD is 24.7%.

[0003] Hemodialysis (HD) is a common and key treatment for ESRD, but it is associated with many complications and adverse reactions that may seriously affect health; for example, physical stress, nutritional issues (such as malnutrition and protein energy wasting), environmental factors (such as temperature fluctuations), and common problems such as blood pressure changes, arrhythmias, and cardiovascular problems, all of which can lead to increased morbidity and mortality. Analyzing blood tests and recent responses during hemodialysis is essential to evaluate the effectiveness of hemodialysis treatment.

[0004] Against the backdrop of the deep integration of information technology and healthcare, the volume of electronic health records (EHRs) continues to grow. Currently, the use of machine learning technology for pattern recognition of EHRs has become a research hotspot. Studies have shown that by applying multiple machine learning models, data such as blood pressure and weight changes in hemodialysis patients can be used to predict blood pressure after dialysis. In addition, some studies have used parameters collected during treatment to estimate the probability of hypotension through machine learning. Other studies have used the WOA-CoxPH model to predict mortality based on the consistency index of eight key parameters. However, these methods mostly focus on clinical parameters during dialysis, and changes in these parameters over time may lead to instability of the methods. To address this problem, some studies have used blood pressure as sequence data and used multi-layer long short-term memory networks (LSTM) to significantly improve prediction accuracy. Similarly, some researchers have used convolutional neural networks (CNNs) combined with a two-stage training strategy to more effectively predict the risk of death in hemodialysis patients. .

[0005] Existing techniques LSTM and CNN are known for their effectiveness on time series data, but they rely on having a long-term, complete dataset of hemodialysis patients, which is often not feasible due to incomplete and irregular data input. Due to different patient conditions, not all biochemical markers can be measured consistently, resulting in significant data gaps that affect the selection of predictive models and their accuracy. Studies have shown that the risk of death in hemodialysis patients is highest early in the treatment process, with survival rates dropping from 91.9% in the first year to 28.5% in the fifth year. Due to changes in the patient's condition, long observation windows and possible hospital transfers are required, which adds challenges and makes it impossible to perform rapid risk assessment using time series-based models.

[0006] Therefore, the present invention proposes a hemodialysis patient survival prediction method based on feature-filled autoencoder to solve the above problems. Summary of the invention

[0007] To solve the above problems, the present invention proposes a method for survival prediction of hemodialysis patients based on feature-filled autoencoders, which uses easily available short-term data to quickly assess the mortality risk of hemodialysis patients. The method utilizes the unsupervised learning of autoencoders to make it suitable for scenarios where limited samples are available. By training the autoencoder with unlabeled data, we capture the latent representation of the incomplete data from an adaptive feature extraction module, and then the classifier uses the latent representation to estimate the mortality risk of hemodialysis patients, providing valuable insights, especially in the early, high-risk stages of treatment.

[0008] The present invention provides the following technical solutions:

[0009] A method for predicting survival of hemodialysis patients based on feature-filled autoencoders comprises the following steps:

[0010] Step 1. The patients are divided into three groups: a survival group, including patients with no death record during the prediction period and subsequent transfer or continued dialysis records; a death group, consisting of patients with non-accidental death records during the prediction period; and an unknown group, with no death record during the prediction period and no subsequent transfer or dialysis records; an autoencoder for unsupervised representation learning is created, and the main components of the autoencoder for unsupervised representation learning include: an encoder and a decoder.

[0011] In step 1, the main components of the autoencoder for unsupervised representation learning: encoder and decoder are expressed as:

[0012] g = g(x)

[0013]

[0014]

[0015] In the formula, the input vector x is mapped to a hidden representation g by the encoder g, and the decoder f maps the hidden representation h back to the output vector Represents the loss function, which helps to improve the encoder's ability to effectively compress high-dimensional data into a low-dimensional feature space, thereby promoting unsupervised learning of data features.

[0016] Step 2: The adaptive mortality prediction model based on feature filling autoencoder (AMPAE) is used to predict death of hemodialysis patients with different degrees of feature missing. First, the missing data is solved by exploiting the underlying manifold structure and relationship between available features in high-dimensional space. The missing features are filled with the values ​​of the complete nearest neighbor features, and then unsupervised feature extraction is performed through the autoencoder. A loss function is customized by combining the enhanced reconstruction loss. and a regularization term To improve the performance of the autoencoder to promote semantic feature sparsity, and finally the model is trained. The training mainly consists of two stages. The adaptive feature extraction module is trained through unsupervised learning of the autoencoder, which makes use of unlabeled data. In the subsequent stage, the adaptive feature extraction module performs feature extraction on the labeled data, which is then fed into the classifier for supervised training. In addition, we implement an early stopping mechanism to prevent overfitting of the autoencoder. This method stops training based on monitoring the performance of feature extraction on known training set samples by a simple classifier; the model is trained for several rounds with randomly initialized weights, and the best performing autoencoder is selected as the feature extractor.

[0017] In step 2, a loss function is designed by combining the enhanced reconstruction loss and regularization loss To improve the performance of the autoencoder to promote the sparsity of semantic features, the expression is:

[0018]

[0019] Where α is a non-negative hyperparameter, and each input feature x in the autoencoder i，k are associated with a data credibility mask, expressed as:

[0020]

[0021] Random feature dropout mask, expressed as:

[0022]

[0023] Among them, the output is processed by the autoencoder as g(f(x i ⊙m i ), the enhanced reconstruction loss is expressed as:

[0024]

[0025] Where N is the sample size, ⊙ is the element-wise product, and β is a non-negative hyperparameter that amplifies the focus on reconstructing real data during training. When β = 0, this simplifies to the mean squared error loss, expressed as:

[0026]

[0027] Regularization loss Helping to control model complexity and reduce overfitting, this strategic mixture of loss components enables our adaptive feature extraction module to effectively capture and utilize high-quality features in datasets with missing data.

[0028] Step 3: Test the quality of the predicted samples, use five classification methods as benchmarks, and use F1 score and average precision (AP) as the main indicators to evaluate the performance of the model.

[0029] In step three, five classification methods are used as benchmarks, and F1 score and average precision (AP) are used as the main indicators to evaluate the performance of the model, including traditional algorithms such as logistic regression, Gaussian Naive Bayes (Gaussian NB), support vector machine (SVM), and deep learning models such as XGBoost and LightGBM classifiers. In order to offset the imbalance of class data, all models are trained using oversampling techniques. In view of the significant differences between the survival group and the death group, relying solely on the accuracy of model evaluation has proved to be insufficient. Therefore, F1 score and average precision (AP) are used as the main indicators to evaluate the performance of the model. The F1 score balances accuracy and recall, which is particularly effective in the case of class imbalance. The expression is as follows:

[0030]

[0031] Where TP is true positive, FP is false positive, and FN is false negative. F1 balances precision and recall, and is the harmonic mean of the two, which can avoid misjudging the model under extreme conditions. On the other hand, AP provides a weighted average of precision at each threshold, using the increase in recall compared to the previous threshold as the weight, expressed as:

[0032]

[0033] In the formula, R n and R n-1are the precision and recall at the nth threshold, respectively. This metric provides a nuanced view of the precision and recall at different thresholds, providing a more comprehensive assessment of the performance and result quality of the retrieval model. A precision-recall (PR) curve is used for detailed performance comparison; although both the receiver operating characteristic (ROC) curve and the PR curve can effectively evaluate the sensitivity and specificity of the model at different thresholds, the ROC curve may be misleading in the case of class imbalance, especially when there are few positive cases. However, the PR curve is still not affected by class imbalance. For missing data, we performed data filling for all data in different prediction windows to ensure the consistency of data processing methods between AMPAE and other models. Grid search is used to optimize the hyperparameters of all models in different prediction windows. The final performance evaluation is based on 5-fold cross validation. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0035] Figure 1 A flowchart of a method for predicting survival of hemodialysis patients based on a feature-filled autoencoder according to the present invention;

[0036] Figure 2 A flowchart of an AMPAE model for mortality prediction in an adaptive mortality prediction method based on a feature-filled autoencoder of the present invention;

[0037] Figure 3 This is a performance evaluation diagram of the model of the present invention; DETAILED DESCRIPTION

[0038] The present invention relates to a method for predicting survival of hemodialysis patients based on feature-filled autoencoder, which can effectively handle the problem of missing data and accurately predict the risk of death of hemodialysis patients. The specific implementation steps of the present invention are as follows:

[0039] Step 1: Data collection and preprocessing: Collect medical data of hemodialysis patients, including but not limited to age, gender, blood pressure during dialysis, hemoglobin level, serum ferritin level, etc.; preprocess the collected data, exclude cases with more than 50% missing data or no blood test or dialysis records during the observation period, average the blood test results during the observation period, and calculate the incidence of thrombolytic hypotension using 6 common definitions, record the maximum, minimum and average values ​​of systolic and diastolic blood pressure during dialysis, solve missing data by exploiting the underlying manifold structure and relationship between available features in high-dimensional space, fill in missing features with the values ​​of complete nearest neighbor features, and introduce a large amount of unlabeled data to help approximate the true distribution of the data, so that the model can discover more meaningful insights from the available data;

[0040] Step 2: Build an autoencoder model: Design and implement an autoencoder neural network, which includes two main components: encoder and decoder. The encoder maps the input data to hidden representation, and the decoder maps the hidden representation back to output data. The autoencoder is trained through unsupervised learning and uses unlabeled data to extract data features.

[0041] Step 3: Feature dropout mask generation: To simulate the distribution of missing data in deep learning frameworks, feature dropout masks are generated. These masks randomly deactivate certain features during the process of data input to the encoder based on the historical probability of feature loss.

[0042] Step 4: Adaptive feature extraction: Use the unsupervised learning ability of the autoencoder to extract latent representations from short-term data. Through feature loss masks, the autoencoder can accurately reconstruct missing features;

[0043] Step 5: Classifier training: Use the potential representation generated by the encoder to train a classifier, such as LightGBM, to predict the mortality rate of hemodialysis patients. During the training process, an early stopping mechanism is used to prevent overfitting.

[0044] Step 6: Model validation and evaluation: Model performance was evaluated through 5-fold cross validation, using F1 score and average precision (AP) as the main indicators. At the same time, PR curve was used to compare the performance of the models in detail.

[0045] Step 7. Feature importance analysis: Use the SHAP method to analyze the impact of different features on model prediction and determine the features that are closely related to the risk of death, such as age, total serum calcium, international normalized ratio, and C-reactive protein.

[0046] The effect of the present invention is further verified by the following experiments:

[0047] The method used in the present invention uses short-term hemodialysis data to evaluate the risk of death of hemodialysis patients, selects a data set with a prediction window of 30 days as the baseline data set, and selects two indicators, F1 and AP, to evaluate the performance of the model. Since the data set used is extremely unbalanced, the mortality rate of hemodialysis patients with an observation window of 30 days is only 6.0%, and it is unreliable to evaluate the model only by accuracy. F1 comprehensively considers the precision and recall rate, and can more comprehensively evaluate the performance of the model in the case of imbalanced categories, and has good stability. AP comprehensively considers the precision and recall rate, so it can more comprehensively evaluate the performance of the retrieval model and the quality of the results. This model performs well in evaluating the risk of death of hemodialysis patients in different time periods. When the evaluation window is 30 days, AMPAE's F1 is 0.53 and AP is 0.5, which has obvious advantages. AMPAE obtained the highest F1 and AP in all evaluation time lengths, and the comprehensive results are good and stable, which shows that the model proposed by the present invention is more effective in evaluating the risk of death of hemodialysis patients, and this advantage is more significant when the prediction window is small.

Claims

1. A method for predicting survival of hemodialysis patients based on feature-filled autoencoders, comprising the following steps: Step 1: Create an autoencoder for unsupervised representation learning, wherein the main components of the autoencoder for unsupervised representation learning include an encoder and a decoder; Step 2: Use an adaptive mortality prediction model based on feature-filled autoencoder to predict death for hemodialysis patient data with different degrees of feature missing; Step 3: Detect and predict sample quality. The detection and prediction sample quality uses five classification methods as benchmarks, and adopts F1 score and average precision (AP) as main indicators to evaluate the performance of the model.

2. The method for predicting survival of hemodialysis patients based on feature-filled autoencoder according to claim 1, characterized in that: The main components of the automatic encoder for unsupervised representation learning in step 1 include an encoder and a decoder, which can be expressed as: h=g(x) In the formula, the input vector x is mapped to a hidden representation h by the encoder g, and the decoder f maps the hidden representation h back to the output vector Represents the loss function, which helps to improve the encoder's ability to effectively compress high-dimensional data into a low-dimensional feature space, thereby promoting unsupervised learning of data features.

3. The method for predicting survival of hemodialysis patients based on feature-filled autoencoder according to claim 1, characterized in that: In step 2, the adaptive mortality prediction model based on feature filling autoencoder designs a loss function, which is combined with enhanced reconstruction loss and regularization loss To improve the performance of the autoencoder to promote the sparsity of semantic features, the expression is: Among them, α is a non-negative hyperparameter, each input feature x in the autoencoder i，k are associated with a data credibility mask, expressed as: The random feature dropout mask expression is: Output is processed by the autoencoder as g(f(x i ⊙m i ), the enhanced reconstruction loss is expressed as: Where N is the number of samples, ⊙ is the element-wise product, and β is a non-negative hyperparameter that amplifies the focus on reconstructing the real data during training; when β = 0, this simplifies to the mean squared error loss: Regularization loss Helping to control model complexity and reduce overfitting, this strategic mixture of loss components enables the adaptive feature extraction module to effectively capture and utilize high-quality features in datasets with missing data.

Images