Value function assessment to different RL algorithms for heparin treatment policy of patients with sepsis in ICU.

Heparin is a critical aspect of managing sepsis after abdominal surgery, which can improve microcirculation, protect organ function, and reduce mortality. However, there is no clinical evidence to support decision-making for heparin dosage. This paper proposes a model called SOFA-MDP, which utilizes SOFA scores as states of MDP, to investigate clinic policies. Different algorithms provide different value functions, making it challenging to determine which value function is more reliable. Due to ethical restrictions, we cannot test all policies on patients. To address this issue, we proposed two value function assessment methods: action similarity rate and relative gain. We experimented with heparin treatment policies for sepsis patients after abdominal surgery using MIMIC-IV. In the experiments, TD(0) shows the most reliable performance. Using the action similarity rate and relative gain to assess AI policy from TD(0), the agreement rates between AI policy and "good" physician's actual treatment are 64.6% and 73.2%, while the agreement rates between AI policy and "bad" physician's actual treatment are 44.1% and 35.8%, the gaps are 20.5% and 37.4%, respectively. External validation using action similarity rate and relative gain based on eICU resulted in agreement rates of 61.5% and 69.1% with the "good" physician's treatment, and 45.2% and 38.3% with the "bad" physician's treatment, with gaps of 16.3% and 30.8%, respectively. In conclusion, the model provides instructive support for clinical decisions, and the evaluation methods accurately distinguish reliable and unreasonable outcomes.

Dynamic Nomogram for Predicting the Risk of Perioperative Neurocognitive Disorders in Adults.

BACKGROUND: Simple and rapid tools for screening high-risk patients for perioperative neurocognitive disorders (PNDs) are urgently needed to improve patient outcomes. We developed an online tool with machine-learning algorithms using routine variables based on multicenter data. METHODS: The entire dataset was composed of 49,768 surgical patients from 3 representative academic hospitals in China. Surgical patients older than 45 years, those undergoing general anesthesia, and those without a history of PND were enrolled. When the patient's discharge diagnosis was PND, the patient was in the PND group. Patients in the non-PND group were randomly extracted from the big data platform according to the surgical type, age, and source of data in the PND group with a ratio of 3:1. After data preprocessing and feature selection, general linear model (GLM), artificial neural network (ANN), and naive Bayes (NB) were used for model development and evaluation. Model performance was evaluated by the area under the receiver operating characteristic curve (ROCAUC), the area under the precision-recall curve (PRAUC), the Brier score, the index of prediction accuracy (IPA), sensitivity, specificity, etc. The model was also externally validated on the multiparameter intelligent monitoring in intensive care (MIMIC) Ⅳ database. Afterward, we developed an online visualization tool to preoperatively predict patients' risk of developing PND based on the models with the best performance. RESULTS: A total of 1051 patients (242 PND and 809 non-PND) and 2884 patients (6.2% patients with PND) were analyzed on multicenter data (model development, test [internal validation], external validation-1) and MIMIC Ⅳ dataset (external validation-2). The model performance based on GLM was much better than that based on ANN and NB. The best-performing GLM model on validation-1 dataset achieved ROCAUC (0.874; 95% confidence interval [CI], 0.833-0.915), PRAUC (0.685; 95% CI, 0.584-0.786), sensitivity (72.6%; 95% CI, 61.4%-81.5%), specificity (84.4%; 95% CI, 79.3%-88.4%), Brier score (0.131), and IPA (44.7%), and of which the ROCAUC (0.761, 95% CI, 0.712-0.809), the PRAUC (0.475, 95% CI, 0.370-0.581), Brier score (0.053), and IPA (76.8%) on validation-2 dataset. Afterward, we developed an online tool (https://pnd-predictive-model-dynnom.shinyapps.io/ DynNomapp/) with 10 routine variables for preoperatively screening high-risk patients. CONCLUSIONS: We developed a simple and rapid online tool to preoperatively screen patients' risk of PND using GLM based on multicenter data, which may help medical staff's decision-making regarding perioperative management strategies to improve patient outcomes.

Two-stage hemoglobin prediction based on prior causality.

Introduction: Perioperative hemoglobin (Hb) levels can influence tissue metabolism. For clinical physicians, precise Hb concentration greatly contributes to intraoperative blood transfusion. The reduction in Hb during an operation weakens blood's oxygen-carrying capacity and poses threats to multiple systems and organs of the whole body. Patients can die from perioperative anemia. Thus, a timely and accurate non-invasive prediction for patients' Hb content is of enormous significance. Method: In this study, targeted toward the palpebral conjunctiva images in perioperative patients, a non-invasive model for predicting Hb levels is constructed by means of deep neural semantic segmentation and a convolutional network based on a priori causal knowledge, then an automatic framework was proposed to predict the precise concentration value of Hb. Specifically, according to a priori causal knowledge, the palpebral region was positioned first, and patients' Hb concentration was subjected to regression prediction using a neural network. The model proposed in this study was experimented on using actual medical datasets. Results: The R 2 of the model proposed can reach 0.512, the explained variance score can reach 0.535, and the mean absolute error is 1.521. Discussion: In this study, we proposed to predict the accurate hemoglobin concentration and finally constructed a model using the deep learning method to predict eyelid Hb of perioperative patients based on the a priori casual knowledge.

Optimization of anesthetic decision-making in ERAS using Bayesian network.

Enhanced recovery after surgery (ERAS) can accelerate patient recovery. However, little research has been done on optimizing the ERAS-related measures and how the measures interact with each other. The Bayesian network (BN) is a graphical model that describes the dependencies between variables and is also a model for uncertainty reasoning. In this study, we aimed to develop a method for optimizing anesthetic decisions in ERAS and then investigate the relationship between anesthetic decisions and outcomes. First, assuming that the indicators used were independent, the effects of combinations of single indicators were analyzed based on BN. Additionally, the impact indicators for outcomes were selected with statistical tests. Then, based on the previously selected indicators, the Bayesian network was constructed using the proposed structure learning method based on Strongly Connected Components (SCC) Local Structure determination by Hill Climbing Twice (LSHCT) and adjusted according to the expert's knowledge. Finally, the relationship is analyzed. The proposed method is validated by the real clinical data of patients with benign gynecological tumors from 3 hospitals in China. Postoperative length of stay (LOS) and total cost (TC) were chosen as the outcomes. Experimental results show that the ERAS protocol has some pivotal indicators influencing LOS and TC. Identifying the relationship between these indicators can help anesthesiologists optimize the ERAS protocol and make individualized decisions.

Convolutional-de-convolutional neural networks for recognition of surgical workflow.

Computer-assisted surgery (CAS) has occupied an important position in modern surgery, further stimulating the progress of methodology and technology. In recent years, a large number of computer vision-based methods have been widely used in surgical workflow recognition tasks. For training the models, a lot of annotated data are necessary. However, the annotation of surgical data requires expert knowledge and thus becomes difficult and time-consuming. In this paper, we focus on the problem of data deficiency and propose a knowledge transfer learning method based on artificial neural network to compensate a small amount of labeled training data. To solve this problem, we propose an unsupervised method for pre-training a Convolutional-De-Convolutional (CDC) neural network for sequencing surgical workflow frames, which performs neural convolution in space (for semantic abstraction) and neural de-convolution in time (for frame level resolution) simultaneously. Specifically, through neural convolution transfer learning, we only fine-tuned the CDC neural network to classify the surgical phase. We performed some experiments for validating the model, and it showed that the proposed model can effectively extract the surgical feature and determine the surgical phase. The accuracy (Acc), recall, precision (Pres) of our model reached 91.4, 78.9, and 82.5%, respectively.

Learning to predict in-hospital mortality risk in the intensive care unit with attention-based temporal convolution network.

BACKGROUND: Dynamic prediction of patient mortality risk in the ICU with time series data is limited due to high dimensionality, uncertainty in sampling intervals, and other issues. A new deep learning method, temporal convolution network (TCN), makes it possible to deal with complex clinical time series data in ICU. We aimed to develop and validate it to predict mortality risk using time series data from MIMIC III dataset. METHODS: A total of 21,139 records of ICU stays were analysed and 17 physiological variables from the MIMIC III dataset were used to predict mortality risk. Then we compared the model performance of the attention-based TCN with that of traditional artificial intelligence (AI) methods. RESULTS: The area under receiver operating characteristic (AUCROC) and area under precision-recall curve (AUC-PR) of attention-based TCN for predicting the mortality risk 48 h after ICU admission were 0.837 (0.824 -0.850) and 0.454, respectively. The sensitivity and specificity of attention-based TCN were 67.1% and 82.6%, respectively, compared to the traditional AI method, which had a low sensitivity (< 50%). CONCLUSIONS: The attention-based TCN model achieved better performance in the prediction of mortality risk with time series data than traditional AI methods and conventional score-based models. The attention-based TCN mortality risk model has the potential for helping decision-making for critical patients. TRIAL REGISTRATION: Data used for the prediction of mortality risk were extracted from the freely accessible MIMIC III dataset. The project was approved by the Institutional Review Boards of Beth Israel Deaconess Medical Center (Boston, MA) and the Massachusetts Institute of Technology (Cambridge, MA). Requirement for individual patient consent was waived because the project did not impact clinical care and all protected health information was deidentified. The data were accessed via a data use agreement between PhysioNet, a National Institutes of Health-supported data repository (https://www.physionet.org/), and one of us (Yu-wen Chen, Certification Number: 28341490). All methods were carried out in accordance with the institutional guidelines and regulations.

A Simple and Quick Screening Method for Intrapulmonary Vascular Dilation in Cirrhotic Patients Based on Machine Learning.

BACKGROUND AND AIMS: Screening for hepatopulmonary syndrome in cirrhotic patients is limited due to the need to perform contrast enhanced echocardiography (CEE) and arterial blood gas (ABG) analysis. We aimed to develop a simple and quick method to screen for the presence of intrapulmonary vascular dilation (IPVD) using noninvasive and easily available variables with machine learning (ML) algorithms. METHODS: Cirrhotic patients were enrolled from our hospital. All eligible patients underwent CEE, ABG analysis and physical examination. We developed a two-step model based on three ML algorithms, namely, adaptive boosting (termed AdaBoost), gradient boosting decision tree (termed GBDT) and eXtreme gradient boosting (termed Xgboost). Noninvasive variables were input in the first step (the NI model), and for the second step (the NIBG model), a combination of noninvasive variables and ABG results were used. Model performance was determined by the area under the curve of receiver operating characteristics (AUCROCs), precision, recall, F1-score and accuracy. RESULTS: A total of 193 cirrhotic patients were ultimately analyzed. The AUCROCs of the NI and NIBG models were 0.850 (0.738-0.962) and 0.867 (0.760-0.973), respectively, and both had an accuracy of 87.2%. For both negative and positive cases, the recall values of the NI and NIBG models were both 0.867 (0.760-0.973) and 0.875 (0.771-0.979), respectively, and the precisions were 0.813 (0.690-0.935) and 0.913 (0.825-1.000), respectively. CONCLUSIONS: We developed a two-step model based on ML using noninvasive variables and ABG results to screen for the presence of IPVD in cirrhotic patients. This model may partly solve the problem of limited access to CEE and ABG by a large numbers of cirrhotic patients.

A better method for the dynamic, precise estimating of blood/haemoglobin loss based on deep learning of artificial intelligence.

BACKGROUND: Dynamic and precise estimation of blood loss (EBL) is quite important for perioperative management. To date, the Triton System, based on feature extraction technology (FET), has been applied to estimate intra-operative haemoglobin (Hb) loss but is unable to directly assess the amount of blood loss. We aimed to develop a method for the dynamic and precise EBL and estimate Hb loss (EHL) based on artificial intelligence (AI). METHODS: We collected surgical patients' non-recycled blood to generate blood-soaked sponges at a set gradient of volume. After image acquisition and preprocessing, FET and densely connected convolutional networks (DenseNet) were applied for EBL and EHL. The accuracy was evaluated using R2, the mean absolute error (MAE), the mean square error (MSE), and the Bland-Altman analysis. RESULTS: For EBL, the R2, MAE and MSE for the method based on DenseNet were 0.966 (95% CI: 0.962-0.971), 0.186 (95% CI: 0.167-0.207) and 0.096 (95% CI: 0.084-0.109), respectively. For EHL, the R2, MAE and MSE for the method based on DenseNet were 0.941 (95% CI: 0.934-0.948), 0.325 (95% CI: 0.293-0.355) and 0.284 (95% CI: 0.251-0.317), respectively. The accuracies of EBL and EHL based on DenseNet were more satisfactory than that of FET. Bland-Altman analysis revealed a bias of 0.02 ml with narrow limits of agreement (LOA) (-0.47 to 0.52 mL) and of 0.05 g with narrow LOA (-0.87 to 0.97 g) between the methods based on DenseNet and actual blood loss and Hb loss. CONCLUSIONS: We developed a simpler and more accurate AI-based method for EBL and EHL, which may be more fit for surgeries primarily using sponges and with a small to medium amount of blood loss.